The goal of this panel is to debate on the current state of the teaching resources in Computer Vision, its strengths and weakness in meeting the current needs of research, teaching and industry. The panel members will act as catalysts to stimulate discussions and are hopeful to generate some recommendations.
The curriculum - The faculty - The CV student's goals can be categorized into seeking successful career opportunities within academic, industrial and research environments. Naturally, keeping in mind the greatest career opportunity lies in the industrial arena. The industry's goals are to engage in successful implementation of MV solutions to enhance productivity, process control and increase the total quality by employing capable MV graduates.
2. Definition
Do we have a clear definition for some of the following questions? What is computer vision ? machine vision ? image processing? What is the primary discipline of the student enrolled in the MV/CV program? CV Hardware? CV Software? IP Algorithms?
3. Issues
My primary observation with graduating student recruits has been the lack of systems approach to deal with machine vision design. It will be beneficial to understand that a successful vision system will have, as a minimum, the following building blocks:
Does the student know the principles behind image formation? A CCD camera, linescan or a reascan, when to use what? What is a TDI? What about other forms of imaging? Has the student been asked to setup the camera, optics and lighting to capture an image data to process? If no, why not? Does the student know the issues associated with real-time processing and the need for efficient algorithms, both in terms of time, performance and the economics. Do we expose the students to the exploding hardware tools and how? (Elementary schools are teaching kindergardners the use of PCs with no awareness of basic addition and subtraction). The vision hardware today performs many of the fundamental algorithms in silicon at extremely high speed. The students must be exposed to the rapidly changing IP hardware. Do we relate the theory of algorithms to practice? Do we point out the relationship between complex named algorithms and their use in real-world applications? Do we explain the useful algorithms and their behavior? For example, the Sub-Pixel Processing Algorithm used to increase the measurement accuracy.
Do we define that the MV/CV is more than a few algorithms? It encapsulates multi-disciplinary components and know-how to successfully complete a practical system. The key areas including electro-optics, illumination, physics, hardware, system software, image processing algorithms, user interface and control; and in some cases mechanical engineering. How does one simplify the problem within the above areas? Do we teach the need for calibration procedures both in 2D and 3D processing?
Basic skills of communication - understanding the problem, interacting with the domain expert, relating to the available tools and implementing a solution by building on numerous small and yet manageable building block solutions.
4. Strategies
Recognize the weakness of the current course work and its objective. Realize there are not many cookbook choices as teaching aids in MV/CV Identify the student's career need and recommend different avenues to master CV and PR. For those who are attracted to an industrial role, insure the curriculum includes a simple practical problem solving exercise and a review and critic an existing MV system. Migrate the topics dealing with basic computer science and/or mathematics to the appropriate prerequisites courses. No need to talk about arrays, linked lists, etc. in MV syllabus.
5. The Reward
If we succeed in addressing the above issues, we may witness the rapid growth of MV applications by the turn of the century and create a greater career opportunity to MV/CV graduates. The teaching faculty may be exposed to more of the challenges found in real world MV problems than abstract algorithms. The industry will experience greater opportunity to succeed in the global market place by increased productivity and quality of its goods with the use of automated MV systems in their manufacturing environments.