This paper is devoted to creation of novel CMOS APS imagers with focal plane parallel image preprocessing for smart technical vision and electro-optical systems based on neural implementation. Using analysis of main biological vision features, the desired artificial vision characteristics are defined. Image processing tasks can be implemented by smart focal plane preprocessing CMOS imagers with neural networks are determined. Eventual results are important for medicine, aerospace ecological monitoring, complexity, and ways for CMOS APS neural nets implementation.
To reduce real image preprocessing time special methods based on edge detection and neighbored frame subtraction will be considered and simulated. To select optimal methods and mathematical operators for edge detection various medical, technical and aerospace images will be tested. The important research direction will be devoted to analogue implementation of main preprocessing operations (addition, subtraction, neighbored frame subtraction, module, and edge detection of pixel signals) in focal plane of CMOS APS imagers. We present the following results: the algorithm of edge detection for analog realization, and patented focal plane circuits for analog image reprocessing (edge detection and motion detection).