Monitoring Of Citrus Fruit Fly Base On Computer Vision Technology

As rural labor to urban migration of the population in young adults,the disadvantages of insufficient rural labor force increasingly highlighted.Now the rural surplus population has enough to prop up the traditional agriculture need heavy labor.In reaction to the phenomenon,how to make full use of advanced automation technology and computer vision technology in order to reduce the burden of farmers become the relationship to the national economy and people's livelihood.The computer vision technology is mainly used in Agricultural situation monitoring and agricultural product quality,higher than the experts of artificial prediction precision.In the rural labor force is far not enough to support the traditional Agricultural situation recognition,more saving time,person and more precise machine vision technology used more widely in the field of agriculture.This article mainly studied in the following aspects:(1)Verify the SIFT algorithm in the fruit fly insect state performance analysis,th is study mainly to a small fruit fly insect pest could SIFT orange distinguish from t hree kinds of fruit fly pests,using SIFT algorithm to distinguish small orange fruit fl y pests accuracy,can better the orange small fruit fly distinguish from other fruit fly pests.(2)Flies morphological characteristics of template matching: this project by use o f the morphological characteristics of a particular joint characteristics of fruit flies,th rough regional roundness and size to identify specific joints,to distinguish between fr uit flies and other insects and template matching.(3)Found in experiments,three kinds of fruit fly with the stripes on the body an d the ratio of shield area have bigger difference,can be as the basis of to distinguis h three types of fruit fly insect pest,this experiment on CCS platform has developed a set of fruit fly shield area and stripe extraction algorithm,this algorithm through t he color space transformation,histogram equalization,image corrosion method completed the extraction of dong area,connected component labeling algorithm,to shield ar ea through the center of gravity is extracted and the vertical striped area search algor ithm to extract,longitudinally striped belt will extract the stripe with pixels and shiel d for the pixel ratio,the ratio of 0 to orange small fruit fly,the ratio falls in the ra-nge of [0.0856,0.3770)for melon fly,ratio above 0.3770 for pumpkin fruit fly,teste d the algorithm to the distinction between the three kinds of fruit fly pests had a go od recognize rate.(4)The physical design of the orchard pest monitoring node: the main introductio n of the real fly tracking algorithm on the platform,as well as the actual number of flies in the algorithm,in the real fly tracking and counting algorithm to consider th e main aspects of the following algorithms:Threshold segmentation algorithm is: through the analysis of experimental compa rison,can get the effect of two-peak method is the best,so this project in fruit flies detection is adopted in the two-peak method threshold segmentation method.Foreground detection algorithm: through the analysis of experimental comparison,can get background updating background difference method,the effect of the detecti on algorithm is optimal,so the project is in the fruit fly detection and background difference.A new mass detection algorithm: this project USES is through the distance as th e size of the old and new judgment of briquette,at the same time through the conto ur detection to detect briquette.Target area segmentation algorithm: by comparison with experimental image seg mentation algorithm based on gray level information and region growing image segm entation algorithm based on YUV channel,based on the YUV channel image segmen tation,image information loss is less,can obtain good effect.Before the target recogni tion,therefore,this project adopts the image segmentation based on YUV channel is to extract the target image area.