3D Information Acquisition Of Cabbage Butterfly On Orthogonal Binocular Vision

With the rapid development of agriculture,the importance of insect behavior analysis has become increasingly prominent.Insect behavior is one of the important research topics in plant protection.In order to study the behavior of insects accurately,it is necessary to observe and record their flight trajectory and motion parameters quantitatively and precisely in three dimensions(3D).According to the wind tunnel experimental environment and the characteristics of the cabbage butterfly,the common pest insect,was selected as the research object.The goal of this research was to analyze frames extracted from videos using Kernelized Correlation Filters(KCF)to plot the 3D trajectory of cabbage butterfly,and then calculating four common used parameters,velocity,acceleration,time and flight distance.Considering the experimental environment with wind tunnel and the movement characteristics of cabbage butterfly,the image acquisition system with two CCD cameras was designed to capture flight videos of cabbage butterfly.Cameras were mounted on two separate cantilever supports,which could insure the optical axis of each camera orthogonality.This study can obtain the insect trajectory and kinematic parameters more accurately,which can meet the application requirements of insect behavior analysis.The research content and innovation of the thesis are as follows:(1)A quadrature binocular vision insect video capture system was designed and used in this study.After analyzing several parameters of wind tunnel,such as installation location,size of the wind tunnel,insect behavior,et al.,and the requirement for high quality of videos,camera model was built with a quadratic binocular vision bracket,completed the installation of the camera pallet,through the test to determine the way to trigger synchronization.The test results showed that the system can meet the demand for the collection of flying butterfly,video smoothing and no frame dropping.The success rate can reach 99.93% to 100%.(2)Two camera calibration methods,orthographic binocular vision and parallel binocular vision methods were compared.Camera internal and external parameters were needed for the parallel binocular vision calibration.The orthographic binocular vision calibration method was not complicated and only need the internal parameters for camera calibration.According to the research requirements and the external factors such as the test site,Zhang Zhengyou’s checkerboard calibration method was used and two cameras were calibrated to get the internal parameters.Side-view and top-view camera calibration error are(0.52;0.50)and(0.42;0.52),respectively.(3)The 2D trajectory of cabbage butterfly was tracked by KCF to obtain the two dimensional coordinates of the centroid of target in two videos,respectively.The target position detection was achieved by finding the maximum output response of the KCF.Then the target models were updated online.Finally the 3D trajectory was calculated according to the matching relation in the corresponding frames of two angles in the video.The parameters of velocity and acceleration were calculated according to 3D coordinate sequence of the centroid of target.The fastest,slowest and average flight speed of cabbage butterfly were 0.6810m/s,0.1542m/s and 0.4331m/s,respectively.And the maximum and minimum accelerations were-6.21m/s2 and 12.00m/s2,respectively.(4)By comparing the measured distance of the test ball with the Euclidean distance obtained by the algorithm proposed in this paper,the results showed that the relative error of the system was less than 3.44%.The method proposed in the thesis could capture a single insect flight trajectory and motion parameters accurately.