Research On Identification Of Stored-Grain Insects’ Activity Based On Video Object Tracking Algorithm

The prevention and control of stored-grain insects is an important work of grain storage management.In addition to early monitoring and detection,the fumigation experiment of stored-grain insects is also an important part.The insecticidal experiment is an experiment to test the resistance and dosage of fumigation by using insecticides to kill insects under specific experimental conditions.This helps to understand the resistance of various insects and guide timely and appropriate insecticidal operations.The results of the current insecticidal experiment in China mainly rely on artificial observation and recording,not only a waste of time and labor costs,but also does not obtain the specific situation of continuous changes in insect status in the process of insecticidal,which is not conducive to a more detailed and scientific assessment of insecticidal effect.A multiple object tracking algorithm is proposed to evaluate the survival state of stored-grain insects automatically in response to this situation,which tracks the movements of stored-grain insects in the target area,recording the velocity and displacement of each stored-grain insect,and analyzed the moving speed and displacement of tracked insects during the time period of a certain length.A grading method is proposed to evaluate the activity degree of stored-grain insects to assist the analysis.The main research work completed in this paper is as follows:1.Establish a data set:Taking the Tribolium castaneum Herbst insects as the research object,shot multiple sets of video,and establish a video data set.After manual labeling,data expansion and data preprocessing,the data sets are applied to the follow-up target tracking algorithm being evaluated and experimented.2.An object detection algorithm for stored-grain pests and a multitarget tracking algorithm based on video data sets are designed to locate,detect and track single stored-grain insects,and record each velocity and displacement of insects.Specific research content includes:feature extraction convolutional neural network design of stored grain insects,target detection and location,data association of detection and tracking,tracking strategy based on Mean Shift,and evaluation results of the algorithm.3.A method for analyzing and discriminating the activity degree of stored-grain insects based on the results of target tracking is proposed:This article records the movement data of each insect during the video object tracking process.Since the movement data can reflect the activity degree of stored-grain insects,Therefore,we propose two ways to determine the degree of activity,and experimentally verify the feasibility of discrimination:One is to use a 5-minites video dataset that contains death and survival insects,to verify the feasibility of extreme activity;the second is to collect insects video data from all alive to all deaths within 20 days,to record the change of the average activity of insects,and use actual changes as a control to verify the validity of the discrimination.