Research On Feature Identification Of Wheat Impact Acoustic Signals Based On Support Vector Machine

Food is the necessity for humans to survive in the world, and food safety or not directly affects political safety, economic development and even social stability. However, crops are gathered, they usually sprout, mold and are infested by insects for humidity, temperature and even store method. So there are a great deal of insect kernels, moldy kernels and sprout kernels in grain storehouse, which cause the decrease of the food quality. Therefore, it is vitally import to detect and classify the damaged kernels in grain.It is so hot that the techniques which can combine pattern recognition and acoustic detection by computer technology achieve the efficient and low-cost automatic detection.In this article, the detection equipment for wheat impact acoustic signals is used to collect the sound of wheat kernel impact acoustic for undamaged wheat kernels and damaged wheat kernels, and then classify the signals automatically by both binary tree support vector machine and the support vector machine optimized by artificial bee colony, respectively. The results indicate the classification accuracy rates improve dramatically, which provide a novelty classification method for wheat kernels.This dissertation mainly includes the following aspects:(1) Elaborate research background and significance of grain detection, and review the present state of impact acoustic detection technology in the world.(2) Design a detection system of wheat kernel impact kernels from both the hardware and software. And introduce the equipment of the system, recording software and setting methods of parameters during the procedure of collecting signals of wheat impact acoustic.(3) Introduce the relative theory of support vector machine which includes statistics learning theory, optimum classification plane, standard support vector machine, classification methods and the hot questions about the research field.(4) Propose the classification method of wheat impact acoustic signals which is based on the binary tree support vector machine, the separability measure of inter classes and implementation procedure algorithm. And extract the features of wheat impact acoustic signals. And study the selection of penalty factor, kernels function about every sub-classifier of support vector machine and the effect for the final classification result which is caused by them. Lastly, analyze the classification results between binary tree support vector machine and traditional support vector machine. It is proved that the method can obtain an ideal result of classification.(5) Propose the classification method which is based on Artificial Bee Colony. First of all, introduce the fundamental principle of Artificial Bee Colony. And show the implementation procedure of penalty factor of support vector machine and the parameter of kernel function which are optimized by Artificial Bee Colony. Besides, decompose wheat impact acoustic signals into8layers by Daubechies wavelet basis function and extract the energy of each layer as the feature vector. In the end, analyze the results of classification and the complexity of the algorithm. It indicates that the Artificial Bee Colony can achieve the optimization for the parameters of support vector machine and improve the performance of its classification.