The Investigation Of Lure And Monitor System For Bemisia Tabaci Based On Computer Vision

Bemisia tabaci is the worldwide pest which has bad effect on vegetables and flowers.When they enter to our country,many of the places experience explosive disaster,causing enormous economic loss to farmers.Traditional method for bemisia tabaci monitoring mainly depends on artificial sampling and investigation,which is time-consuming,arduous,less accuracy and less reliability.With the widely used of computer vision and image recognition technology in the field of agricultural plant diseases and insect pests,the growing number of pest monitoring and forecasting systems appear at this historic moment.But for tiny pests like bemisia tabaci,common monitoring system is not useful.This research mainly focuses on three parts:whiteflies image automatic acquisition device,image recognition and segmentation algorithm,to develop a whiteflies iure and monitor system based on computer vision.The specific research contents and results as below:(1)The first part discusses the development of whiteflies iure and monitor system.Aiming at whiteflies' individual life habits and characteristics,this part analyzes some existing pest iure devices,then chooses the reasonable trap tool(yellow trap board)and the image acquisition device(scanner),and designs an automatic whiteflies image acquisition device.This device can adjust height of the bracket and base according to the growing of plants,change trap board automatically,position the trap board precisely and collect image.This device also can meet the requirement of continuous acquisition of whiteflies image without the guard of anyone.(2)The second part studies the whiteflies segmentation method based on the color threshold.According to image pre-processing techniques and the statistical analysis data of the image of whiteflies,the obvious color characteristics of whiteflies can be found.The RGB color image segmentation method which use B=60 as optimal threshold value can be chosen to separate whiteflies target from the trap plate background,miscellaneous worm and impurities.The segmentation accuracy can reach to 97.8%.(3)For the segmentation of mutual adhesion,the improved watershed algorithm,k-means clustering segmentation algorithm and the improved code method have been used.The improved watershed algorithm has effective segmentation results,but it has some over-segmentation phenomenon,the segmentation accuracy is 87.4%;For K-means clustering segmentation algorithm,due to too much target image,the accuracy of the clustering center is not precise,which makes the segmentation effect is not ideal;For the improved code method,it can position the inflection point positioning of whiteflies adhesion target accurately,and get an ideal separation effect.But for the whiteflies targets which wings are spread excessively,it has 27%possibility of wrong segmentation.So,the accuracy of code method is 94.9%.(4)The whiteflies are very tiny,which make the large obstructions in the image,such as leaves,large moths block off a lot of whiteflies goals.This sitution brings negative impact on the accuracy of the count results.This study estimates the amount of whiteflies in the shelter,which could enhance the accuracy of the whiteflies' counting.In conclusion,this bemisia tabaci iure and monitoring system can reach to more than 96%accuracy of the counting.It can basically reflect the density of whiteflies in the field,and can provide the data support for the telemetry,prevention and control of whiteflies.Besides,automated image sampling process can liberate the labor force and improve the working efficiency.This system surely has certain research value and practical significance.