Sa Gas Sensor Films Array-based Image Recognition System And Its Fpga Implementation

The development of highly sensitive, selective, reliable, and compact sensing systems to detect a wide range of volatile organic compounds is of great significance to olfactory analysis based on the utilization of electronic nose, while the discrimination of compounds at low concentrations or within a given chemical class remains challenging goals.A new and simple approach to array detection of odorants using the colorimetric response from a library of immobilized vapor-sensing dyes has been discovered recently by Professor Kenneth S. Suslick at the University of Illinois at Urbana-Champaign. The colorimetric array shows good linear response to single analytes, as well as unique response for analyte mixtures. This technique, which is called "smell-seeing" by its inventors, is very similar to using litmus paper to determine if a solution is acid by seeing if the paper goes from blue to pink. Accordingly, the image recognition system based on the sensor array has been designed in this thesis for the purpose of molecular recognition of volatile organic compounds vapor. The main research contents include:1. As gas sensing films have been prepared by self-assembly technology due to its comparatively simple structure and easiness to control, fabricate and characterize, the gas molecule could be detected by the responsive sensor array in order to obtain the responsive image. Color change patterns obtained from the array is researched to give striking visual identification and even weakly-ligating vapors could be differentiated.2. Using a sensor array of color-responsive dyes with enormous discriminatory power, the responsive image could be acquired with a CMOS image sensor. The image is decomposed to RGB color components by the image sensor and the signal could be sent to a microprocessor for further processing.3. Image recognition algorithms have been designed for detection of different odorants according to the different colors of colorimetric array. The feature information of responsive image is transformed into the feature vector of color histogram. Then, the algorithm of color histogram and the algorithm of color coherence vector are designed and compared with each other in detail. 4. An embedded image recognition architecture based on MicroBlaze embedded system has been designed in Xilinx Spartan-3 FPGA to implement the algorithms of image recognition. Both of these two designed algorithms can be implemented in real-time. The algorithm of color histogram can be implemented faster while the algorithm of color coherence vector provides higher recognition efficiency. Experimental results demonstrate that both of the two algorithms above could be analyzed effectively to represent different volatile organic compounds according to their different responsive images in this system.In summary, current research and analysis of the image recogniton system according to smell-seeing technology in this thesis is focused on how to make the whole system device smaller, less expensive, and more sensitive. High-performance sensors and systems of electronic nose are required that can identify and measure various kinds and quantities of odorants, and these electronic noses have high potential in numerous medical, industrial and commercial applications such as environmental monitoring, quality control, food processing and medical diagnosis.