Design And Implementation Of Image Analysis System Based On The Identification Of Pests

In order to study high performance of pest recognition methed,this paper uses digital image processing techniques and pattern recognition techniques to design a system of pest recogntion based on image processing techniques with Visual C++as software.The main contents are as follows:(1)Pest image pre-processing.This paper describes how to proceed location uniform, image scaling,smooth de-noising,threshold segmentation,mathematical morphology operation,contour extraction and contour filling to pest image,and give the results of processing. Using the inertial principal axis of the pest to achieve location uniform;using bicubic interpolation to achieve image scaling;according to the particularity of pest images,this paper gives improved methods of median filtering and threshold segmentation,mathematical morphology operation and contour extraction,then fills contour to the pest imgae,realize the image preprocessing of pest.(2)Feature extraction.In this paper,according to the characteristics of pest image,we extract two types of features:global shape features of pest and SIFT local features.Shape features include two categories:geometric features and moment features.Shape features include regional perimeter,regional area,horizontal axis length and so on,moment features include form parameter,eccentricity.Then introduce how to extract SIFT features,and extract SIFT features of pests.(3)Feature selection、classification and recognition.This paper introduces the fuzzy clustering method based on fuzzy mathematics theroy and describe how to select feature with this method.Then introduce how to design the classification in this paper,we design it to two layers,first grade classify the pest by pest global characteristic,second grade classify the pest by SIFT local features,and make the output of fisrt grade classification as the input of the second grade classification.(4)System design and realization.This paper describes every function of this system,the system’s workflow,and how to design this recognition system in detail,then give the results after recognition of six types of pests.At the end of this paper, it summarizes the content of the research and points out the area where improvement is needed.