Research On Feature Classification And Recognition Of Grassland Pests Based On Machine Vision

Prairie is the main base of animal husbandry and breeding in China,and it is also the main source of green food in our country.In recent years,as a result of grassland pest propagating,grassland is seriously affected by the erosion of pest,so that an amount of grazing is damaged every year and the deterioration of grassland is on the verge of death or destruction,which seriously affected the animal husbandry and the production of meat product in China.In the meanwhile,the deterioration of grassland and strong sandstorm have seriously affected the people who live in the grassland and those who live in the surrounding cities.Although the government has paid more attention to the restoration of grassland ecosystems,which mitigates the trend of the deterioration of grassland which is caused by locust in the grassland,yet other pests in the grassland cause the deterioration of grassland by gnawing grassland more and more seriously.According to analysis,the weakness of monitoring and early warning ability is the main reason of causing the serious diseases and pests of grassland.The research of this paper is based on the machine vision of grassland pests feature classification and recognition,which is of great significance for the reconstruction and greening of grassland and creating a good ecological environment.The research of this paper belongs to the field of machine vision and pattern recognition.In fact,it is a kind of recognition system on grassland diseases pests which is based on machine learning.In this paper,the images of pests were obtained by artificial shooting and choosing part of the pictures as sample images.Five kinds of pests images were involved in the research,and each type of pest has fifty different pictures with different photographing angle and different backgrounds.The pictures of each type of pest were drew 20% at random as training sample and the rest pictures were used as testing sample.In this paper,the typical pest of grassland,Haplotropis,was used as a case.Firstly,I segmented the picture of Haplotropis by the method of OTSU and got the binary image.Then I processed the binary image using the method of Mathematical Morphology.Finally,I cascaded the processed image with original image in order to get the color segmentation image,and in this way,image segmentation operation was finished.Secondly,I described the segmented color image of Haplotropis as features and then formed these feature to feature vector to train the two classifier of SVM and BP neural network.The research used two classifiers on pest image recognition.Finally,I used two classifiers to recognition the pest image and compared two classifiers,in this way,I got that the SVM classifier has a relatively good recognition results for grassland pests recognition.The research is a new type of grassland pests identification system,and it can effectively solved the disadvantages of causing serious pollution to the ecological environment by pesticide spraying to kill pests.It has great significance for the restoration of grassland ecological system and protect the environment.