Abrupt Motion Tracking Method Of Ochotona Curzoniae Based On Particle Filter

As a predominant rodent species on the Qinghai-Tibet Plateau,the normal population quantity of the Ochotona curzoniae is the key to maintaining the ecosystem balance.However,when its population quantity exceeds a certain range,rodent damage will occur,which is one of the main reasons for accelerating grassland degradation.In order to achieve the non-contact,dynamic and continuous monitoring of the Ochotona curzoniae,a machine vision-based intelligent monitoring system can be used to monitor Ochotona curzoniae,and targeted measures can be taken to prevent and control the dangers of Ochotona curzoniae.In the intelligent monitoring system,target tracking is used to track the Ochotona curzoniae,thus laying a foundation for subsequent quantitative statistics and behavioral analysis.However,the motion of the Ochotona curzoniae is random and unpredictable,and sometimes smooth motion is accompanied by abrupt changes;in addition,there is a low contrast between target and background color,which increases the difficulty of tracking.1.Considering that the motion model of the particle filter tracking method ignores acceleration information with a poor abrupt motion tracking performance of Ochotona curzoniae,and that there is a linear relationship between acceleration variance and acceleration's absolute value as well as between acceleration increment and position increment,a acceleration variance factor can be introduced into the motion model of the particle filter tracking method,so as to track the abrupt motion of the Ochotona curzoniae.Given that the particle filter tracking method takes color features as target representation and that it has a low robustness when used to track targets with a low contrast between target color and background color.Aiming at this problem,multi-feature fusion can be performed on the observation model,and color information can be fused with gradient information;in terms of weight allocation,Pearson correlation coefficient can be used to adaptively change the weight of each feature.The experimental results show that the tracking accuracy of the particle filter tracking method can be improved by modifying its motion model,and can improve the anti-interference ability of the tracking method by multi-feature fusion on the observation model.2.When traditional tracking methods are used for target tracking,usually single tracking method is used to track the target.With regard to this problem,when themotion form discrimination method is introduced to the tracking method in this thesis and different tracking methods are adopted for different motion forms,the tracking of the Ochotona curzoniae can be refined.The experimental results also show that,after discrimination of motion forms,adopting different tracking methods for different motion forms increases the accuracy of overall tracking.