Automatic video tracking of Chinese mitten crabs based on the particle filter algorithm using a biologically constrained probe and resampling - Computer Vision Research Group（计算机视觉实验室）

In order to know the living habits of Chinese Mitten Crab automatically, we proposed a new computer vision algorithm to track the trajectory of crab in this paper. Particle Filter tracking algorithm is used to predict the target's location and size from the captured video sequences. We proved it is suitable to handle non-Gaussian movement, which meet the movement characteristics of crab movement. In the feature match stage of motion tracking, we propose using multi-features which contains color and contour features to match the target for increasing the tracking stability and robustness. By using the weighting method, the multi-features particle filter (MFPF) combines the color and contour features which could effectively solve the tracking failure problem due to similar color interference, strong light disturbance and deformation of the target. To determine the robustness and accuracy of our algorithm, we have used about 10 minutes video sequences to test our algorithm. The experimental results show that the MFPF is feasible and can be used as an efficient tool to get the pathway of crabs under water to a certain extent with solving the problem of the previous methods.