Real-time Inspection of Metal Laminates by means of CNN's

摘要

Analog CNN array computer arises as an alternative to traditional digital processors in many industrial inspection like visual quality control of metal laminates, capable of make in a single chip Tera equivalent operations per second. A 4096 analog CNN processor array is able to perform complex space-time image analysis, being much faster than a camera-computer system in continuous inspection applications. Both chips have been implemented in CMOS technology and they are managed by a 32-bit high-performance low-cost micro-controller that closes the pan, tilt, lighting, focus and zoom loops required in the implementation of the active vision strategies. Several convolution masks for the Cellular Processors has been selected to detect particular changes in the texture, size, direction or orientation of the image entities, reprogramming "on the fly" the pixel resolution or shape when necessary. Laboratory results present these Cellular Processors and multiple resolution imager circuits as a promising architecture for visual inspection of industrial processes in real time. The traditional image processing techniques (filtering, segmentation, interpretation) require a lot of computational effort due to data on each pixel are computed in a sequential way and the path of information is an analog/digital converter. The delay accumulation create in this process is unacceptable in real time image processing because of the information flow created in the usual vision tasks (e.g. automatic industrial inspection, vision problems in robotics, pattern analysis, etc.). Thus, the use of a massive parallel architecture working with analog signals avoid the previous problems. This is just the basis idea of Cellular Neural Network (CNN's): An array of analogic dynamic processors which cells interact directly within a finite local neighborhood. The local CNN conectivity allow its realization as VLSI chips that can operate at a very high speed and complexity: 0.3TeraXPS performance for a 10Xl0mm2 chip using a 2-um technology in a robust implementation.