A smart camera performs real-time analysis to recognize scenic elements. Smart cameras
are useful in a variety of scenarios: surveillance, medicine, etc.We have built a real-time
system for recognizing gestures. Our smart camera uses novel algorithms to recognize gestures
based on low-level analysis of body parts as well as hidden Markov models for the moves that
comprise the gestures. These algorithms run on a Trimedia processor. Our system can recognize
gestures at the rate of 20 frames/second. The camera can also fuse the results of multiple
A "smart camera" is basically a video camera coupled to a computer vision system in a tiny
package. This communication begins stating the main differences between smart cameras and
standard smart vision systems. Smart camera architecture is described whereby a combination of
an on-board microprocessor and PLD’s allow for the embedding of image processing algorithms
in the camera. A static thresholding algorithm is presented which demonstrates the ability to
track non-uniformity in the inspection target. A multi camera inspection system application is
presented where a maximum of twenty smart cameras may be networked together to a single
host computer. Finally, a prediction is made of technological and applicational future evolution
on smart cameras.
The smart camera – a whole vision system contained in one neat housing – can be used
anywhere, in any industry where image processing can be applied. Companies no longer need a
cabinet in which to keep all their computing equipment: the computer is housed within the smart
camera. In the pharmaceutical industry and in clean rooms – when not even dust is allowed – this
can be a big advantage. A single square meter of space can be comparatively very expensive – if
there is no need for a component rack or cabinet, simply a smart camera, and then this could save
a lot of money. In particular, there would not be the usual cabling involved for other vision
systems, and set-up is simple. Later in this communication are stated some advantages of using
smart cameras or PC-based systems in vision applications.
In usual vision systems scenarios, only a small fraction of a picture frame will be the region of
interest (ROI). In fact, often no visual image of the ROI is even required. The object of a vision
system, after all, is to make a decision: "Is there a blob"? "Where is the blob"? "Is this a defect"?
What if all that pixel pre-processing and decision-making could be done within the camera? If all
the processing were done inside the camera, the blob analysis of a gigabit image might result in
only a few hundred bytes of data which need to be sent somewhere. Such compact packets of
data could be easily transmitted directly to a machine control without even passing through a PC.
Information should be processed where the information occurs – i.e. the brain should be behind
The answer to the problem stated earlier is a stand-alone, smart camera.To illustrates this, a smart
camera’s embedded system architecture is shown along with an example of a hardware
embedded image processing algorithm: “Static Gray Scale Thresholding”.
Many global companies including Glaxo and Ciba-Geigy, and motor companies such as Ford and
Volkswagen are users of smart cameras.
A small example of a vision system for web inspection is shown where twenty smart cameras are
Many changes are yet to come concerning smart cameras as in technology as well as in there
future applications. An overview, concerning these issues, is made closing this communication.
We have to recognize however that smart cameras will not be the answer for all vision
"smart camera" for an image capture system with
embedded computing can extract information from
images without need for an external processing unit, and
interface devices used to make results available to other
devices. – RETICON (first smart camera users)
2.1 OVERVIEW OF SMART CAMERA
A smart camera is an integrated machine vision system which, in addition to image capture
circuitry, includes a processor, which can extract information from images without need for an
external processing unit, and interface devices used to make results available to other devices.
A smart camera or "intelligent camera" is a self-contained, standalone vision system with built-in
image sensor in the housing of an industrial video camera. It contains all necessary
communication interfaces, e.g. Ethernet, as well as industry-proof 24V I/O lines for connection
to a PLC, actuators, relays or pneumatic valves. It is not necessarily larger than an industrial or
This architecture has the advantage of a more compact volume compared to PC-based vision
systems and often achieves lower at the expense of a somewhat simpler (or missing
altogether) user interface.
Although often used for simpler applications, modern smart cameras can rival PCs in terms of
processing power and functionalities. Smart cameras have been marketed since the mid 80s, but
only in recent years have they reached widespread use, once technology allowed their size to be
reduced while their processing power has reached several thousand MIPS (devices with 1GHz
processors and up to 8000MIPS are available as of end of 2006).
Having a dedicated processor in each unit, smart cameras are especially suited for applications
where several cameras must operate independently and often asynchronously, or when
distributed vision is required (multiple inspection or surveillance points along a production line
or within an assembly machine).
Fig 2.1 Early smart camera (ca. 1985, in red) with an 8MHz Z80 compared to a
modern device featuring Texas Instruments' C64 @1GHz
A smart camera usually consists of several (but not necessarily all) of the following components:
Image sensor (matrix or linear, CCD- or CMOS)
Image digitization circuitry
processor (often a DSP or suitably powerful processor)
program- and data memory (RAM, nonvolatile FLASH)
Communication interface (RS232, Ethernet)
I/O lines (often opt isolated)
Lens holder or built in lens (usually C, CS or M-mount)
Built in illumination device (usually LED)
Purpose developed real-time operating system (For example VCRT)
A video output (e.g. VGA or SVGA) may be an option for a Smart Camera.
2.2 Smart Cameras vs. Standard Smart Vision Systems
The question often comes up as to what is the most appropriate approach to take in implementing
a vision system - using a smart camera or using some sort of PC-based approach. There is no
question that as the microprocessors, DSPs and FPGAs are getting faster and, therefore, more
capable, smart cameras are getting smarter. Therefore, they are a challenge to more ''traditional''
approaches to vision systems. Significantly, however, ''traditional'' approaches are also taking
advantage of the advances and so, too, are faster and smarter.
Traditional approaches usually mean a PC-based implementation. This could be either using a
camera with the capability to interface directly to the PC (IEEE 1394/Fire wire, Camera Link,
LVDS, USB, etc.), or a system based on a frame grabber or other intelligent image processing
board or vision engine that plugs into the PC. In this latter case, more conventional analog
cameras are used as the input device.
A smart camera, on the other hand, is a self-contained unit. It includes the imager as well as the
''intelligence'' and related I/O capabilities. Because this format resembles the format of many
intelligent sensors, these products are often referred to as ''vision sensors.'' However, a vision
sensor often has a limited and fixed performance envelope, while a smart camera has more
flexibility or tools, inherently capable of being programmed to handle many imaging algorithms
and application functions. A PC-based vision system is generally recognized as having the
greatest flexibility and, therefore, capable of handling a wider range of applications.
2.3.1 PC-Based Vision Systems Advantages
The PC-based vision systems advantages include:
1) Flexibility –
The PC offers greater flexibility in the number of options that can be selected. For
example one can use a line scan versus an area scan camera with the PC. One can use
third party software packages with the PC approach (smart cameras tend to be single
2) Power –
PC's tend to offer greater power and speed due in large part to the speed of the Intel
processors used internally. This power in turn means that PC's are used to handle the
''tougher'' applications in vision systems.
2.3.2 Smart Cameras Advantages
The smart cameras advantages include:
Smart cameras are generally less expensive to purchase and set up than the PC based solution,
since they include the camera, lenses, lighting (sometimes), cabling and processing.
2) Simplicity –
Software tools available with smart cameras are of the point-and-click variety and are easier to
use than those available on PC's. Algorithms come pre-packaged and do not need to be
developed, thus making the smart camera quicker to setup and use.
3) Integration –
Given their unified packaging, smart cameras are easier to integrate into the manufacturing
4) Reliability –
With fewer moving components (fans, hard drives) and lower temperatures, smart cameras are
more reliable than PC's.
In general the performance of the smart camera will continue to increase. This will mean that the
smart camera will be used for more difficult applications, slowly displacing the PC approach.
Fig 2.2. Standard Smart Vision System
Fig 2.3. Smart Camera System.
2.4 Smart Camera Architecture
The smart camera presented in this communication reduces the amount of data generated to the
‘data of interest’ by making use of embedded image processing algorithms. The data of interest
might be, for example, defective areas of the product being inspected. Multiple cameras can
route their data to a single frame grabber and computer due to the reduction of data stream, thus
dramatically reducing system cost and increasing inspection bandwidth capability. This smart
camera also makes use of an on-board microprocessor for communication with the inspection
systems’ host computer and for internal control functions.
The following block diagram illustrates the camera architecture.
Fig 2.4. Smart Camera Architecture Block Diagram
A detailed explanation of the camera architecture follows, starting with the image sensor.
2.4.1 Image Sensor Basics
In this smart camera, a CCD (Charge Coupled Device) image sensor converts photons (light) into
electrons (charge). When photons hit an image sensor, the sensor accumulates electrons. This is
called charge integration. The brighter your light source, the more photons available for the
sensor to integrate, and the smaller the amount of time required to collect a given amount of light
Finally, the sensor transfers its aggregate charge to readout registers, which feed each pixel’s
charge from the image sensor into an output node that converts the charges into voltages. After
this transfer and conversion, the voltages are amplified to become the camera’s analog output.
2.4.2 Analog to Digital Conversion Electronics
The analog output of the CCD is converted to a digital output for further processing. The camera
presented here sub-divides the CCD analog output into eight channels of 256 pixel elements
each. Analog to digital conversion is performed at a 20 MHz data rate for each channel thus
yielding an effective camera data rate of 160 MHz. The digital data is then passed along to the
image processing electronics for processing and analysis.
2.4.3 Image Processing Electronics
Image processing is performed by embedded algorithms on a per channel basis. The following
block diagram illustrates the basic processing architecture for each channel.
Fig 2.5. Image Processing Architecture Block Diagram
The processing algorithm is embedded in the processing PLD. The microprocessor has a
bidirectional path being able to randomly access the algorithm parameters, as well as program a
new algorithm into the PLD as required by the user. Raw pixel data and associated timing and
control signals are also connected to input pins into the processing PLD,.
For storage and subsequent readout, algorithm processed data is output along with a write control
signal to FIFO memory. multiplexing of the data is achieved by using FIFO memory.
Readout control is accomplished by the microprocessor/FIFO readout control card whose
architecture is shown in the figure 2.6.
Fig 2.6. Microprocessor/FIFO Readout Control Circuit Board Block Diagram
The Microprocessor/FIFO readout control circuit board acts as the master controller for the smart
camera. FLASH memory is used to store microprocessor code and PLD algorithm code. In
system programmability is achieved because all PLD devices in the image processing section of
the camera are SRAM based.
2.5 Image Processing Algorithms
Many types of image processing algorithms can be embedded within the camera, since the video
processing modules are completely in-system programmable. As an example, a static grey scale
thresholding algorithm is presented below.
2.5.1 Static Grey Scale Thresholding
In static thresholding, an upper and a lower bound are established around what is considered a
normal value. Image data that falls within the boundary window is considered normal non-
interesting data. Image data that falls either above or below the boundary window is considered
data of interest. Considering we are dealing with an 8-bit digital camera, the normal, upper and
lower boundary values are seen to be digital numbers (DN) on a scale of 0 to 255 “Gray scale”.
Imagine that a product is being inspected for defects and the grey scale level of non-defective
product is 85 DN, and the upper and lower boundary values have been set to +/- 15 DN. All
image data that fell within the bounds of 70 DN to 100 DN would be considered non-interesting
and would not be transmitted out of the camera.
Image data that fell below 70 DN and above 100 DN would be considered interesting and would
be transmitted out of the camera. Substantial data reduction is achieved since only some of the
data will fall outside of the established boundaries.
It is important to note that all of the ‘data of interest’ is transmitted out of the camera and thus
data reduction is achieved where all of the grey scale information is preserved. This type of
algorithm is illustrated by the image shown in figure 2.7.
Fig 2.7. An Example of Static Thresholding
For later display and analysis each pixel must be given an address such that an image can be
reconstructed by the frame grabber since an algorithm of this type produces noncontinuous data.
The static thresholding algorithm requires three parameters as follows an upper bound, a lower
bound, and a centre value. The determination of the centre value is essential to this type of
algorithm, and the acceptable band between the upper and lower bound. The static thresholding
algorithm is expressed as follows:
IF (PIXEL GRAY IS > (CENTER + UPPER)) OR (PIXEL GRAY IS < (CENTER – LOWER) )
2.5.2 Embedded Image Processing Algorithms
The algorithms are embedded in hardware with a PLD/microcontroller combination and operate
at a 20MHz data rate per channel. The effective processing rate is 40MHz because each image
processing PLD can process two channels of image data. With dedicated DSP controllers such
data processing rates could be difficult to achieve. Microcontroller also can directly control the
algorithm without host computer intervention, since it has access to the image data.
2.6 Smart Camera System
A vision system for web inspection is presented below where a maximum of twenty 2048 pixel
high sensitivity line scan smart cameras are networked together to a single host computer and
frame grabber. A block diagram of the system is shown in figure 2.8
Fig 2.8. System Block Diagram
The system shown consists of up to twenty 2048 pixel high sensitivity line scan smart cameras
housed within a camera enclosure mounted above the web. Transmissive illumination is provided
since illumination source is mounted beneath the web. Routed through two cabinets are the data,
control, and power lines to/from the cameras. The system makes use of fibre optic technology for
transmission of data and control signals thus allowing the inspector station to be located remotely
at a distance of up to 100m.
Today embedded technology produces Intelligent,
Multitasking, Compact electronic products with networking
3.1 Introduction to embedded systems:
An embedded system is a special-purpose computer system designed to perform one or a few
dedicated functions, often with real-time computing constraints. It is usually embedded as part of
a complete device including hardware and mechanical parts. In contrast, a general-purpose
computer, such as a personal computer, can do many different tasks depending on programming.
Embedded systems control many of the common devices in use today.
Since the embedded system is dedicated to specific tasks, design engineers can optimize it,
reducing the size and cost of the product, or increasing the reliability and performance. Some
embedded systems are mass-produced, benefiting from economies of scale.
Physically, embedded systems range from portable devices such as digital watches and MP4
players, to large stationary installations like traffic lights, factory controllers, or the systems
controlling nuclear power plants. Complexity varies from low, with a single microcontroller
chip, to very high with multiple units, peripherals and networks mounted inside a large chassis or
In general, "embedded system" is not an exactly defined term, as many systems have some
element of programmability. For example, Handheld computers share some elements with
embedded systems — such as the operating systems and microprocessors which power them —
but are not truly embedded systems, because they allow different applications to be loaded and
peripherals to be connected.
Embedded computing systems
• Computing systems embedded within electronic devices
• Hard to define. Nearly any computing system other than a desktop computer
• Billions of units produced yearly, versus millions of desktop units
• Perhaps 50 per household and per automobile.
Fig 3.1 An Embedded System
3.2 Examples of embedded systems
Embedded systems span all aspects of modern life and there are many examples of their use.
Telecommunications systems employ numerous embedded systems from telephone switches for
the network to mobile phones at the end-user. Computer networking uses dedicated routers and
network bridges to route data.
Consumer electronics include personal digital assistants (PDAs), mp3 players, mobile phones,
videogame consoles, digital cameras, DVD players, GPS receivers, and printers. Many
household appliances, such as microwave ovens, washing machines and dishwashers, are
including embedded systems to provide flexibility, efficiency and features. Advanced HVAC
systems use networked thermostats to more accurately and efficiently control temperature that
can change by time of day and season. Home automation uses wired- and wireless-networking
that can be used to control lights, climate, security, audio/visual, surveillance, etc., all of which
use embedded devices for sensing and controlling.
Transportation systems from flight to automobiles increasingly use embedded systems. New
airplanes contain advanced avionics such as inertial guidance systems and GPS receivers that
also have considerable safety requirements. Various electric motors — brushless DC motors,
induction motors and DC motors — are using electric/electronic motor controllers. Automobiles,
electric vehicles, and hybrid vehicles are increasingly using embedded systems to maximize
efficiency and reduce pollution. Other safety systems such as anti-lock braking
system (ABS), Electronic Stability Control (ESC/ESP), traction control (TCS) and automatic
Medical equipment is continuing to advance with more embedded systems for vital signs
monitoring, electronic stethoscopes for amplifying sounds, and various medical imaging (PET,
SPECT, CT, MRI) for non-invasive internal inspections.
In addition to commonly described embedded systems based on small computers, a new class of
miniature wireless devices called motes are quickly gaining popularity as the field of wireless
sensor networking rises. Wireless sensor networking, WSN, makes use of miniaturization made
possible by advanced IC design to couple full wireless subsystems to sophisticated sensor,
enabling people and companies to measure a myriad of things in the physical world and act on
this information through IT monitoring and control systems. These motes are completely self
contained, and will typically run off a battery source for many years before the batteries need to
be changed or charged. An Embedded System is a combination of Hardware and Software that
may have some mechanical components to perform specific tasks.
Fig 3.2 PC Engines' ALIX.1C Mini-ITX embedded board with an x86 AMD Geode LX 800 together with Compact
Flash, miniPCI and PCI slots, 22-pin IDE interface, audio, USB and 256MB RAM
3.3 Characteristics of embedded systems:
1. Device programmability or manageable-
The functioning of a specifically designed hardware part on a smart device can easily be
changed by simply changing the software associated with it.
An embedded system employed in a modern refrigerator performs door sense and
temperature sense at the same times, which are two functions at the same time.
3. Real time response-
It is the ability of an embedded system to respond to ambient conditions suddenly. That
is, a smart TV adjusts picture quality suddenly in response to sudden environmental
4. Embedded systems are designed to do some specific task, rather than be a general-
purpose computer for multiple tasks. Some also have real-time performance constraints
that must be met, for reasons such as safety and usability; others may have low or no
performance requirements, allowing the system hardware to be simplified to reduce costs.
5. Embedded systems are not always standalone devices. Many embedded systems consist
of small, computerized parts within a larger device that serves a more general purpose.
For example, the Gibson Robot Guitar features an embedded system for tuning the
strings, but the overall purpose of the Robot Guitar is, of course, to play music.
Similarly, an embedded system in an automobile provides a specific function as a
subsystem of the car itself.
6. The program instructions written for embedded systems are referred to as firmware, and
are stored in read-only memory or Flash memory chips. They run with limited computer
hardware resources: little memory, small or non-existent keyboard and/or screen.
CAMARAS AS EMBEDDED SYSTEMS
A video camera coupled to a computer vision system is
a tiny package of embedded systems.
Co working of hardware and software engineers is the
basis of Embedded Application Development.
4.1 Overview of Smart Cameras in embedded systems
The system also provides input and output to the plant control system. Data from the cameras is
acquired by the frame grabber, assembled into images, and then transferred to the host computer
real time display on the defect monitor, and stored to a database residing on the file server via an
Ethernet connection. Subsequent analysis of the data is performed at the analysis workstation
with analysis software that allows extraction of data from the database for creation of reports. All
system and analysis software is multithreaded and provides real time data access and display. Via
a modem connection the system is also operable remotely. To ensure smooth and constant
illumination of the web the system software also controls the illumination source with a fuzzy
logic control scheme.
Recent technological advances are enabling a new generation of smart cameras that represent a
quantum leap in sophistication. While today's digital cameras capture images, smart cameras
capture high-level descriptions of the scene and analyze what they see.
These devices could support a wide variety of applications including human and animal
detection, surveillance, motion analysis, and facial identification.
Video processing has an insatiable demand for real-time performance. Fortunately, Moore's law
provides an increasing pool of available computing power to apply to real-time analysis. Smart
cameras leverage very large-scale integration (VLSI) to provide such analysis in a low-cost, low-
power system with substantial memory. Moving well beyond pixel processing and compression,
these systems run a wide range of algorithms to extract meaning from streaming video.
Because they push the design space in so many dimensions, smart cameras are a leading edge
application for embedded system research.
4.2 Detection and Recognition Algorithms
Although there are many approaches to real-time video analysis, we chose to focus initially on
human gesture recognition—identifying whether a subject is walking, standing, waving his arms,
and so on. Because much work remains to be done on this problem, we sought to design an
embedded system that can incorporate future algorithms as well as use those we created
exclusively for this application.
Our algorithms use both low-level and high-level processing. The low-level component identifies
different body parts and categorizes their movement in simple terms. The high-level component,
which is application-dependent, uses this information to recognize each body part's action and
the person's overall activity based on scenario parameters.
Human detection and activity/gesture recognition algorithm has two major parts: Low-level
processing (blue blocks in Figure 4.1) and high-level processing (green blocks in Figure 4.1).
Figure 4.1 : Algorithm.
A) Low-level processing
The system captures images from the video input, which can be either uncompressed or
compressed (MPEG and motion JPEG), and applies four different algorithms to detect and
identify human body parts.
The first algorithm transforms the pixels of an image like that shown in Figure 4.2 a, into an M ¥
N bitmap and eliminates the background. It then detects the body part's skin area using a YUV
color model with chrominance values down sampled
Figure 4.2 a
Next, as Figure 4.2 b illustrates, the algorithm hierarchically segments the frame into skin tone
and non-skin-tone regions by extracting foreground regions adjacent to detected skin areas and
combining these segments in a meaningful way.
Figure 4.2 b
The next step in the process, shown in Figure 4.2 c, involves linking the separate groups of
pixels into contours that geometrically define the regions. This algorithm uses a 3 ¥ 3 filter to
follow the edge of the component in any of eight different directions.
Figure 4.2 c
To correct for deformations in image processing caused by clothing, objects in the frame, or
some body parts blocking others, an algorithm fits ellipses to the pixel regions as Figure 4.2d
shows to provide simplified part attributes. The algorithm uses these parametric surface
approximations to compute geometric descriptors for segments such as area, compactness
(circularity), weak perspective invariants, and spatial relationships.
Figure 4.2 d
Each extracted region modeled with ellipses corresponds to a node in a graphical representation
of the human body. A piecewise quadratic Bayesian classifier uses the ellipses parameters to
compute feature vectors consisting of binary and unary attributes. It then matches these attributes
to feature vectors of body parts or meaningful combinations of parts that are computed offline.
To expedite the branching process, the algorithm begins with the face, which is generally easiest
B) High-level processing
The high-level processing component, which can be adapted to different applications, compares
the motion pattern of each body part—described as a spatiotemporal sequence of feature vectors
—in a set of frames to the patterns of known postures and gestures and then uses several hidden
Markov models in parallel to evaluate the body's overall activity.
We use discrete HMMs that can generate eight directional code words that check the up, down,
left, right, and circular movement of each body part.
Human actions often involve a complex series of movements. We therefore combine each body
part's motion pattern with the one immediately following it to generate a new pattern. Using
dynamic programming, we calculate the probabilities for the original and combined patterns to
identify what the person is doing. Gaps between gestures help indicate the beginning and end of
A quadratic Mahalanobis distance classifier combines HMM output with different weights to
generate reference models for various gestures. For example, a pointing gesture could be
recognized as a command to "go to the next slide" in a smart meeting room or "open the
window" in a smart car, whereas a smart security camera might interpret the gesture as
suspicious or threatening.
To help compensate for occlusion and other image-processing problems, we use two cameras set
at a 90-degree angle to each other to capture the best view of the face and other key body parts.
We can use high-level information acquired through one view to switch cameras to activate the
recognition algorithms using the second camera. Certain actions, such as turning to face another
direction or executing a predefined gesture, can also trigger the system to change views Soft-
We can use MatLab to develop our algorithms. This technical computation and visualization
programming environment runs orders of magnitude more slowly than embedded platform
implementations, a speed difference that becomes critical when processing video in real time. We
can therefore port our MatLab implementation to C code running on a very long instruction word
(VLIW) video processor, which lets us make many architectural measurements on the
application and make the necessary optimizations to architect a custom VLSI smart camera.
At the development stage, we can evaluate the algorithms according to accuracy and other
familiar standards. However, an embedded system has additional real-time requirements:
The system must process a certain amount of frames per second to properly analyze motion and
provide useful results. The algorithms we use as well as the platform's computational power
determine the achievable frame rate, which can be extremely high in some systems.
The amount of time it takes to produce a result for a frame is also important because smart
cameras will likely be used in closed-loop control systems, where high latency makes it difficult
to initiate events in a timely fashion based on action in the video field.
Moving to an embedded platform also meant that we have to conserve memory. Looking ahead
to highly integrated smart cameras, we want to incorporate as little memory in the system as
possible to save on both chip area and power consumption. Gratuitous use of memory also often
points to inefficient implementation.
Our development strategy calls for leveraging off-the-shelf components to process video from a
standard source in real time, debug algorithms and programs, and connecting multiple smart
cameras in a networked system. We use the 100-MHz Philips TriMedia TM-1300 as our video
processor. This 32-bit fixed- and floating-point processor features a dedicated image coprocessor,
a variable length decoder, an optimizing C/C++ compiler, peripherals for VLIW
concurrent real-time input/output, and a rich set of application library functions including
MPEG, motion JPEG, and 2D text and graphics.
4.5 Testbed Architecture
Our testbed architecture, shown in Figure 4.3, uses two TriMedia boards attached to a host PC
for programming support. Each PCI bus board is connected to a Hi8 camera that provides NTSC
composite video. Several boards can be plugged into a single computer for simultaneous video
operations. The shared memory interface offers higher performance than the networks likely to
be used in VLSI cameras, but they let us functionally implement and debug multiple-camera
systems with real video data.
Fig 4.3 Testbed architecture
4.6 Experiments and Optimizations
As data representation becomes more abstract, input/output data volume decreases. The change
in required memory size, however, is less predictable given the complex relationships that can
form between abstract data. For example, using six single precision, floating-point parameters to
describe 100 ellipses requires only 2.4 Kbytes of memory, but it takes 10 Kbytes to store
information about two adjoining ellipses.
Based on these early experiments, we optimize our smart camera implementation by applying
techniques to speed up video operations such as substituting new algorithms better suited to real
time processing and using TriMedia library routines to replace Clevel code.
4.7 Algorithmic changes
We originally fit super ellipses (generalized ellipses) to contour points, and this was the most
time-consuming step. Rather than trying to optimize the code, we decided to use a different
algorithm. By replacing the original method developed from principal component analysis with
moment-based initialization, we reduced the Levenberg- Marquardt fitting procedure, thus
decreasing the execution time.
After converting the original Matlab implementation into C, we performed some experiments to
gauge the smart camera system's effectiveness and evaluate bottlenecks. The unoptimized code
took, on average, 20.4 million cycles to process one input frame, equal to a rate of 5 frames per
We first measure the CPU times of each low-level processing step to determine where the cycles
were being spent. Microsoft Visual C++ is more suitable for this purpose than the TriMedia
compiler because it can collect the running time of each function as well as its sub functions'
Figure 4.4a shows the processing-time distribution of the four body-part-detection algorithms
Figure 4.4b shows the memory characteristics of each low-level processing stage.
Figure 4.4 a, 4.4b
4.8 Control-to-data transformation
Increasing the processor's issue width can exploit the high degree of parallelism that region
extraction offers. Using a processor with more functional units could thus reduce processing time
during this stage. However, contour following, which converts pixels to abstract forms such as
lines and ellipses, consumes even more time. The algorithm also operates serially: It finds a
region's boundary by looking at a small window of pixels and sequentially moving around the
contour; at each clockwise step it must evaluate where to locate the contour's next pixel. While
this approach is correct and intuitive, it provides limited ILP.
We evaluate all possible directions in parallel and combined the true/false results into a byte,
which serves as an index to look up the boundary pixel in a table. We also manipulate the
algorithm's control-flow structure to further increase ILP. These optimizations double the
contour-following stage's running speed.
OF SMART CAMERAS
Application of smart cameras takes place where
volume, pricing or reliability constraints forbid use of bulkier
devices and PC's.
5.1Fields of application:
Smart cameras can in general be used for the same kind of applications where more complex
vision systems are used, and can additionally be applied in some applications where volume,
pricing or reliability constraints forbid use of bulkier devices and PC's.
Typical fields of application are:
Vision-based smart environments.
Surveillance and tracking applications.
Applications based on fusion of visual and other sensory data.
Multi-view vision for Human-Computer Interfaces (HCI).
3D scene analysis with distributed sensors.
Distributed multimedia and gaming applications.
Automated inspection for quality assurance (detection of defects, flaws, missing parts...)
Non contact measurements.
Part sorting and identification.
Code reading and verification (barcode, Data Matrix, alphanumeric etc.)
Web inspection (inspection of continuously flowing materials such as coils, tubes, wires,
extruded plastic) for defect detection and dimensional gauging.
detection of position and rotation of parts for robot guidance and automated picking
unattended surveillance (detection of intruders, fire or smoke detection)
biometric recognition and access control (face, fingerprint, iris recognition)
visual sensor networks
Developers can purchase smart cameras and develop their own programs for special, custom
made applications, or they can purchase ready made application software from the camera
manufacturer or from third party sources. Custom programs can be developed by programming
in various languages (typically C or C++) or by using more intuitive, albeit somewhat less
flexible, development tools where existing functionalities (often called tool or blocks) can be
connected in a list (a sequence or a bidimensional flowchart) that describes the desired flow of
operations without any need to write program code. The main advantage of the visual approach
Vs. programming is in a much shorter and somewhat easier development process, available also
to non-programmers. Other development tools are available with relatively few but
comparatively high level functionalities, which can be configured and deployed with very
Smart cameras running software tailored for a single specific application are often called "vision
The term "smart camera" for an image capture system with embedded computing was used by
Reticon for one of their products, by 1977.
5.2 Applications of Smart Camera Networks
Self localizing smart camera networks can serve as an enabling technology for a wide range of
higher level applications. Here we focus on two applications where the images from the camera
systems are used to derive information about the geometric structure of the environment.
5.2.1 Visual Hull Reconstruction
Multi camera systems are commonly used to derive information about the three dimensional
structure of a scene. One approach to the reconstruction problem which is particularly well suited
to the proposed self localizing smart camera network is the method of volume intersection which
has been employed in various forms by a number of researchers. This method can be used to
detect and localize dynamic objects moving through the field of view of the smart camera
network. Here a set of stationary cameras are used to observe one or more objects moving
through the scene. Simple background subtraction is employed to delineate the portions of the
images that correspond to the transient objects. Once this has been accomplished one can
interrogate the occupancy of any point in the scene, P, by projecting it into each of the images in
turn and determining whether or not it lies within the intersection of the swept regions. This
process can be used to produce an approximation for the 3D structure of the transient objects by
sampling points in the volume. The results of such an analysis are shown in Figure5.1.
Figure 5.1(a) Background image of a scene 5.1(b) Image with object inserted
5.1c) Results of the background subtraction operation 5.1(d) Results of applying the volumetric
reconstruction procedure to the difference images derived from the three smart camera nodes.
In this application the ability to rapidly localize a set of widely separated cameras is a distinct
advantage. Other implementations of this reconstruction scheme involve complex, time
consuming calibration operations. This implementation, in contrast, could be be quickly
deployed in an ad-hoc manner and would allow a user to localize and track moving objects such
as people, cars or animals as they move through the scene.
CONCLUSION AND SOME
A Musial fountain dances according to tunes of a music-
this is the smart world created by embedded technology.
In future, Smart Cameras will be able analyze and
understand scenes they observe in a cooperative manner.
6.1 Optimization results and Conclusion
The combination of these methods radically improves CPU performance for the application.
Optimization boosts the program's frame rate from 5 to 31 frames per second. In addition,
latency decreases from about 340 to 40-60 milliseconds per frame. We can add HMMs and other
high-level processing parts, and that makes the program now runs at about 25 frames per second.
Our board-level system is a critical first step in the design of a highly integrated smart camera.
Although the current system is directly useful for some applications, including security and
medicine, a VLSI system will enable the development of high-volume, embedded computing
Because digital processors and memory use advanced small-feature fabrication and the
sensor requires relatively large pixels to efficiently collect light, it makes sense to design the
system as two chips and house them in a multichip module. Separating the sensor and the
processor also makes sense at the architectural level given the well understood and simple
interface between the sensor and the computation engine.
The advantages of leveraging existing sensor technology far outweigh any benefits of using
pixel-plane processors until they become more plentiful. However, attaching special-purpose
SIMD processors to the multiprocessor can be useful for boundary analysis and other operations.
Such accelerators can also save power, which is important given the cost and effort required to
deploy multiple cameras, especially in an outdoor setting. High-frame-rate cameras, which are
useful for applications ranging from vibration analysis to machinery design, will likely require
many specialized processing elements that are fast as well as area efficient.
In contrast to common digital cameras, as known for example for taking holiday pictures, Smart
Cameras are equipped with a computing unit that is used for the analysis of captured
images. Recent advances in the research field of Computer Vision allow for the detection of
human faces and gestures and for the measurement of distances between objects. In future, Smart
Cameras will be able analyse and understand scenes they observe in a cooperative manner.
Distributed Smart Cameras can be connected to each other by ad-hoc or infrastructure networks.
This Distributed System does not rely on a central control console and has therefore no single
point of failure - an important aspect when safety and robustness are concerned.
This paper describes a scheme for determining the relative location and orientation of a set of
smart camera nodes and sensor modules. The scheme is well suited for implementation on
wireless sensor networks since the communication and computational requirements are quite
minimal. Self localization is a basic capability on which higher level applications can be built.
For example, the scheme could be used to survey the location of other sensor motes enabling a
range of location based sensor analyses such as sniper detection, chemical plume detection and
Further, the ability to automatically localize a set of smart cameras deployed in an ad-hoc manner
allows us to apply a number of multi-camera 3D analysis techniques to recover aspects of the 3D
geometry of a scene from the available imagery. Ultimately we envision being able to construct
accurate 3D models of extended environments based on images acquired by a network
inexpensive smart camera systems.
6.2 Looking Ahead
The vision industry is rapidly moving away from the video camera/frame grabber systems of the
twentieth century to a new generation of smart-camera-based systems for the 21st century. These
21st century smart-camera systems will perform real-time, pixel-data extraction and processing
operations within the camera at extremely high speeds and at a cost, which is considerably less
than required today for comparable capabilities. Eventually, complete vision-processing-systems-
on-a-sensor-chip will be available.
Components in smart cameras will undoubtedly change due to the push from semiconductors and
new microprocessors coming onto the market. The trends for the next year will be towards
megapixel sensors, higher resolution, faster processing power, and colour. A typical standard
CCD based camera has a matrix of 480,000 pixels, however the new megapixel cameras offer at
least 1,000x1,000 pixels – or 1 million pixels. Some manufacturers already offer cameras of 2
6.2.1 Future Applications
To date, exploitation of smart camera technology has been mainly for industrial vision systems,
but a crossover is just starting to take place. Smart camera technology will begin to enter new
applications, for example, in the security and access control markets, in the automotive industry,
for collision avoidance, and even – one day – for the toy industry.
Even our automobiles may soon be outfitted with miniature eyes. Built into a cruise control
system, for instance, such a camera would suddenly alert the driver if it noted a rapidly
decelerating vehicle. The cameras could also take the place of the rear view and side-view
mirrors, thereby eliminating dangerous blind spots and - in the event of an accident – recording
the seconds prior to a collision.
Another example would be with intelligent lifts. An office block, with many lifts and floors, may
see a lot of people travelling up and down between floors, particularly at high traffic times such
as early morning or end of the working day. At the moment, lifts are called by somebody
pressing a button and putting in a request for the lift to stop at a particular floor. Connected with
smart camera technology, lifts could be routed on demand, working intelligently, stopping only
when there was a pre-set number of passengers waiting at a floor – and missing out a floor if too
many people were waiting to meet the maximum capacity of the lift.
Looking into the future, we can foresee an infinite number of applications for the smart camera;
in fact, as many as there are potential image processing uses.
Distributed Smart Cameras can be used for a wide range of applications. To name a few, the
Our society is aging constantly. This leads to a large number of elderly people that need to be
cared for. Many elderly move to residential homes because they are frightened of accidents that
might occur at home and leave them helpless without the possibility to call for help. Smart
Cameras can detect accidents on their own and inform either relatives or a mobile nursing
2.Surveillance of large, safety-critical areas (airports, train stations):
Common surveillance systems are prone to errors since human staff in central control rooms can
usually not analyse all video data captured by the surveillance cameras. Critical events may pass
unnoticed and cause great harm that could have been avoided by the use of Smart Cameras.
Smart Cameras are able to detect critical events (e.g. persons leaving suspicious luggage or
entering forbidden areas) on their own and raise an alarm. Notifications about suspicious
incidents can be sent directly from a Smart Camera to a PDA carried by security staff nearby.
Common sensors used for automation often rely on wired, electromechanical devices. The use of
wireless, contactless vision-based devices can help to reduce costs, since maintenance costs of
common sensors are high.
Internet online shops offer the possibility to analyze the customers' behavior in detail. Each click
and length of stay on pages can be analyzed in detail and thereby the customers’ satisfaction and
the shop's conversion can be increased. For shopping malls and retail stores, this is currently not
possible. Smart Cameras can help to analyze customers' behavior in detail while preserving the
customers' privace by not transmitting video data but operating figures (people counter, duration
The DISC team has been awarded a prize for this innovation. The project is funded by the
german federal ministry of economics and technology (EXIST programme). The excellence-
oriented EXIST programme line "Transfer of Research" promotes especially sophisticated
technology-based business start-up projects in the pre-start-up phase.
. Wayne Wolf, Burak Ozer, Tiehan Lv, “Smart Cameras as High- Performance Embedded
. Burak Ozer, Wayne Wolf, "A Hierarchical Human Detection System in Compressed and
. Tiehan Lv, Burak Ozer, Wayne Wolf, "Smart Camera System Design," Invited Paper,
International Packet Video Workshop, Pittsburgh, April 2002.
 Advanced Imaging Europe Magazine: The intelligent camera, October (2002) 12 – 16
 Wintriss Engineering Corporation: Year 2002 Smart Cameras
 Smart Cameras vs. PC-based Machine Vision Systems, http://www.coreco.com/,
 Longbottom, D.: Latest Developments in Sensor and Camera Technology, Alrad Instruments
Ltd., White paper, http://www.ukiva.org/IPOTMV02Latest.pdf,
 Smart Cameras - A complete vision system in a camera body, Vision Components,
 Lehotsky, D. A.: Intelligent High Sensitivity CCD Line Scan Camera with embedded Image
Processing algorithms, DALSA INC, (2002)
 Siemens R&I: The Camera that Grew a Brain,http://w4.siemens.de/FuI/en/archiv/zeitschrift/
 Camillo J. Taylor Department of Computer and Information Science University of
 Babak Shirmohammadi Department of Computer and Information Science University of