Image Recognition For Total Suspended Particulate (TSP) Based On Fractal Theory

Total Suspended Particulate (TSP) means the solid and liquid grain things, whose aerodynamics equivalent diameter is below 100 microns. It is one of the most important reasons of air pollution. How to identify TSP effectively in the atmosphere, and to adopt necessary protection, has been an important topic that the environment personnel, both in and abroad, studies all the time.This thesis takes the digital image of the TSP as the research object. First of all, the thesis looks back the development history of the digital image, and summarized the mainly study contents and its application fields. And then it introduces the fractal theory, a few kinds of definition on fractal dimension, multi-scale and multiplex fractal theory.On this foundation, the thesis briefly summarizes some general methods to compute the fractal dimension, including differential box-counting approach, approach in the frequency field, and approach based on fractal Brown Model. After introducing the basic operating for mathematics morphology and the approach based on it to estimate fractal dimension, the thesis has improved the method based on mathematics morphology with variable template, which was enlightened by mathematics morphology with fixed template, to estimate fractal dimension. That kind of method has been applied to the digital image of TSP. The fractal characteristic has been one of the most important gist for image recognition, that is applying the fractal theory to recognize the digital image of TSP.A recognition software has been designed and implemented with Visual C++ as the tool.The software is a single document file program based on MFC. Firstly, it reads the digital image of TSP, which acquires from CCD image sensor. After pre-processing with medium filtering and threshold value selection by adaptation, the TSP could be recognized with the method to estimate the fractal dimension based on mathematics morphology, according to the fractal characteristic of the source image. And finally the amount of TSP in the image will be counted.Finally, by comparing fractal theory with wavelet and chaos, whose research objects are non-linear problems, discussing their relation, the thesis has settled a good academic foundation for further research.