| The prevention of insect pests for grain storage has always been the key research field of the food storage industry, while the hidden pests’ detection and prevention shall be the difficult point in this field. In this paper, aiming at the defeats and shortages of the traditional food pest detection method, the biophoton analytical technology(BPAT) is combined with the pattern recognition methods, and the pattern classification algorithm is designed to conduct the classified study to see if the wheat is infected. Moreover, the classification algorithm is optimized further.The emphasis of this study is the classifier design and optimization algorithm design based on the ultra-weak photon radiation signals. The research contents and innovations include:(1) Feature extraction is conducted for the ultra-weak photon emission signal of wheat. It mainly include: noise reduction treatment for the signals measured for the normal wheat and infected wheat. 6 statistical characteristics and 20 histogram features are extracted from the noise reduction data to form the characteristic space.(2) The classifier design based on the KNN algorithm and BP neural network algorithm: different parameters and different feature vectors are employed respectively to distinguish the normal wheat and infected wheat.(3) The traditional BP neural network has such defeats as the slow convergence speed, easy to fall into local minimum, etc. Aiming at improving it, the genetic algorithm(GA) is adopted to optimize the weight and threshold of the BP neutral network. And so the optimized GA-BP classifier is designed. The GA-BP classifier is adopted to conduct the classification experiment for normal wheat and infected wheat, and good classification results has been achieved.The research results proves that it is feasible to apply BPAT and pattern recognition method in the detection of hidden pests. The research achievements of this paper will partly lay the experimental and theoretical foundation for the further study. |
PDF Full Text Request