- This article is filed under:
- » View All
Year 2002 Smart Cameras
by Vic Wintriss, President - Wintriss Engineering Corporation Posted 03/10/2000
If you were to try to imagine what industrial cameras will be capable of by the year 2002, what would you say? Are pixel resolutions of 2,000 x 2,000 at 40 frames per second achievable? How about 14-bit images? More? In my view, even numbers as powerful as these will be considered conservative by then. After all, technological progress has proceeded at an exponential rate ever since the development of the modern digital computer. I am convinced this exponential trend will continue for several years to come.
Such imaging prowess, however, will not come without its challenges. To understand the problem it will pose for manufacturers, next imagine that one of these powerful new cameras has been installed in an industrial environment. Consider that it is collecting 14-bit, 2,000-x-2, 000-pixel images at the rate of 40 frames per second. That's 2.24 gigabits/second, or 280 megabytes/second (2,000 x 2,000 x 14 x 40). Even in 2002, a 2-gigabit serial interface will be considered fast, and processing 280 megabytes of pixel data per second in a PC will not be easy. The chart below illustrates how this increase in the rate of pixel generation will outrun channel and PC capabilities within a few years
A good question to ask at this point is, "Why would anyone want to dedicate such tremendous serial bandwidth for sending frames of raw pixel data to a computer?" In usual machine-vision scenarios, only a small fraction of a picture frame comprises the region of interest (ROI). In fact, often no visual image of the ROI is even required. The object of a machine-vision system, after all, is to make a decision: "Is there a blob"? "Where is the blob"? "Is this a defect"?
This raises another question: 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.
The answer to the problems stated earlier is a stand-alone, smart camera which can internally perform at least 20 BOPS (billion operations/second) of effective pixel processing power. This camera would have discrete digital outputs for applications where fast, machine-control response is required. High-speed Ethernet connectivity should also be available for transmission of pre-processed picture data directly to existing network assets. Each camera must have synchronization inputs for applications that require multiple, synchronized images.
Specifications for a smart, stand-alone camera that will meet year 2002 requirements are listed below.
- 5000 - 10,000 pixel Line-Scan Sensor
- 50 Mega-pixels per second pixel rate
- S/N greater than 50dB
- Variable line length from 256 to 10,000 pixels/line
- Variable line rates:
3750 lines/second at 10,000 pixel resolution
150,000 lines/second at 256 pixel resolution
- Sync modes:
External encoder locked
- Exposure Control:
1 microsecond to 1 second
On-Board Pipelined Processing Features:
- Corrected video (digital gain and offset per pixel)
- 10 bit look up table providing 1,2,4 or 8 bits per pixel
- Data Packing 1,2,4 or 8 pixels per byte
- Multi-Class Streak Encoder
- 32Mbyte Video Buffer
Data Output modes:
- Raw Video via Ethernet
- Packed Video Data 1,2,4,8 bits per pixel
- Real-time Streak Data (Xstart, Xstop, Line Number)
- Real-time Raw Video via 400 MHz HotLink Fiber Channel (ANSI X3.230)
I/O Control and Data
- Ethernet (100 Mbit, 100 Base TX)
- Serial (RS-422)
- Digital Inputs (TTL Level)
- Digital Outputs (TTL Level)
- Analog Control Outputs
A Practical Web Inspection Application
GAF Materials Corporation needed to devise a way to inspect fiberglass moving at 1900 feet per minute for the presence of streaks, holes and clumps. A series of ten, synchronized cameras with the characteristics listed above could be used for this application with the following system specifications:
|Maximum image width:
Maximum image length:
Maximum image acquisition rate:
Maximum image processing rate:
Maximum line scan rate:
Maximum web speed:
Minimum defect size:
|50,000 pixels (10 cameras)
400 Megapixel per second
400 Megapixel per second
530 K lines/second
6,000 feet per minute
A ten-camera system is arranged as shown below.
An example of a large web operation is shown below.
A major stumbling block in the implementation of smart camera systems is the technical expertise needed to program these complex devices. Programming is normally required in both the host computer (typically a PC) and within the camera. For applications where pixel processing can be accomplished in a PC, a number of third-party vendors provide software packages for image processing. VisionBlox by Integral Vision has developed an excellent set of extremely powerful image-processing tools that can be dragged and dropped into a Visual Basic or Visual C++ program running on a PC. Using VisionBlox, an experienced programmer can develop a typical application in less than a day. National Instruments also has a vision system processing package. For real-time, pipeline-processing applications where pixel processing must be done in the camera, a graphical user interface (GUI) for the PC which communicates with a real-time operating system such as VxWorks running within the camera will be necessary. Wintriss Engineering has developed such a system called the WebRanger for web inspection applications. A typical WebRanger operator screen is shown below.
Wintriss Engineering Corporation manufactures a line of smart, pipeline-processing cameras which match most of the requirements described earlier for the ideal year 2002 smart camera. These cameras can crunch large amounts of pixel data within the camera and send the results over the Internet. They can also send real-time data over a Fibre Channel interface. An illustration of the Wintriss camera is shown below.
The Machine Vision Industry is rapidly moving away from the video camera/frame grabber systems of the twentieth century to a new breed 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. Until that time, however, it is my belief that the FPGA/DSP smart camera approach described here will be the gold standard for line scan cameras.