Statistical Studies Of Visual Art Based On Crop Pests

The article takes computer vision technology as research foundation, hasconducted the research in view of pests automatic counting method. Rice is the mostimportant food crop. Hence the production of rice plays a decisive part in agriculturalproduction and national economic development all over the country. Rice planthopperis a major pest to rice, however, in recent years, there’s a aggravating tendency in ricepests outbreak, which has posed a serious threat to the rice production. Currently, theroutine method of rice pests monitoring relies mainly on the sex attractant or blacklight lamps for trapping the pests. The trapped pests are taken back, counted bytechnical staff the following day which is of high labor intensity, low efficiency andnon-real time. Because of serious pests outbreak, the method can’t satisfy the currentdemand of rice pests monitoring. Therefore, it is obvious that only by studying oneffective automatic counting technology, can we get accurate information of pests intime and then the scientific decision be made to for integrated pests prevention. In thispaper, the automatic counting of rice planthopper is studied on basis of computervision and image processing technology. The results proved that this method wasfeasible to do the pest counting about rice planthopper.The main contents of the research included the following five aspects:(1) The determination of the research project. In greenhouse rice fields, use theyellow sticky cards with green plaid lines to trap rice planthoppers and take themback, then photograph it by digital camera and use image processing to achievethe goal of automatic counting. In order to analyze the accuracy of counting,compare the labor statistics and automatic counting;(2) Image enhancement. By comparing the median filter and mean filter, concludedthat the median filter can protect image details. Through the experimental analysis,using3×3square window can get a satisfying results;(3) Image segmentation. A new method was proposed for background based onIterative threshold method, improved Otsu method and the minimum error rateBayesian decision theory. Firstly, remove the yellow background of the sticky cards by using global threshold method, the green grid lines and rice planthopperswere taken out; secondly, in the HSV space, calculate the optimal threshold withthe improved Otsu, then estimate the mean, variance and proportion of the greenlines and planthopper in S channel; finally, with minimum error rate Bayesianstrategy to separate the green lines and planthopper to complete the segmentationof background. For adhesion planthopper, a segmentation based on range convertis adopted which help to segment the conglutination planthopper;(4) Automatic counting. Using a Breadth-first labeling algorithm which is based onregion growing. Compared with the traditional algorithm, this method can onlyscan the image once with depth-first search algorithm. Also, this method isn’taffected by the area and shape of the connected component which possess a betterrobustness;(5) Design of the automatic counting system about rice planthopper. The systembased on computer vision is designed and developed by using MATLAB language.The software of the system is composed of file managing, image enhancement,background segmentation, adhesion planthopper segmentation, automaticcounting and so on.