The Detection And Recognition Of Stored Grain Pests Based On Image Processing

Grain is the basic safeguard of national security, and is also the foundation of human survival and development. As a country’s major strategic issue, grain security always related to China’s national economic development, social harmony and stability. With the continuous development of science and technology, grain production and grain storage quantity have been greatly improved. The damage of grain pests during grain storage is a key factor. Under the background of engineering application, conducting the accurately detect and correct recognition to the images from the system becomes an important basis to make a decision about protecting the grain. It will help us protect national food security by science and green Initiative. It’s a relatively complete process from the image preprocessing, feature extraction and optimization to the image’s recognition process. This paper has completed the following tasks in the framework of image recognition processing:(1) In the aspect of image enhancement, we use the multifunctional grain situation detection system developed by Zhengzhou Xinsheng electronic technology Co. Ltd. to extract the image of grain pests, and then make the image denoised. Because the image has the trait about low resolution and clean background, so we adopt the contrast enhancement to highlight the research object and select median filter to remove noise. Experiments show that the way can meet the system requirements.(2) In the aspect of image segmentation, this paper briefly introduces the threshold segmentation and edge segmentation. This article focuses on image edge detection which based on cellular automata(CA). The parameters and model building are given by the CA’s discrete of the time, space and state. Simulation verifies that test performance by CA is better than that of Prewitt operator, so make two programs for the following experiments.(3) About the feature detection and the optimization of features, Using negative pressure to extract the grain pests which in observation desk with strong light irradiation, the images have less obvious in color(mainly show brown) and texture characteristics. Five kinds of common grain pests(corn weevil, lesser grain borer, flour beetle, saw-toothed grain beetle, rusty grain beetle) are chosen to make a study by the 33-dimensional geometric characteristics and morphological characteristics. Principal component analysis is used to on the extracted features for further dimension reduction and optimization. Ultimately, short time-consuming and high classification success rate features are selected, provides the basis of the optimal combination.(4) In the aspect of recognition and classification, using radial basis function(RBF) neural network based on the Gaussian kernel and support vector machine(SVM) to recognize and classify the images about grain pests respectively. The results show that, the SVM to grain insect image has a better recognition result, which make the detection of image edge by cellular automata.Based on the research content above, and the background of engineering application, this paper puts forward a method which based on cellular automata and SVM to classify the grain pests.