Research On Key Techniques For Performance Evaluation Of Optical Target Tracking

Due to the lack of accurate evaluation model and complete test image set,the performance of tracking algorithm is difficult to measure,and there are no clear indicators and data.This leads to the lack of quantitative analysis of the results of the tracking algorithm and the lack of objective understanding of the performance of different algorithms.In addition,due to the lack of a general theoretical method,it is more difficult to choose the appropriate tracking algorithm in a specific application scenario,the tracking algorithm often need to be repeated the test.The object of this paper is suitable for the target tracking algorithm under infrared and visible light.Considering the characteristics of optical target tracking,the target itself,the background of the target and the tracking jamming are studied.The target in common cases is generated by simulation,and the test image set is established.The tracking status of the tracking algorithm is evaluated by analyzing in turn the situation that one or more different targets may exist in the tracking process under the combined action of single and multiple influencing factors.Therefore,the index,parameter system and framework of performance evaluation are established.In order to solve the problem of selecting the optimal tracking algorithm in a given scenario,it is necessary to test the performance on a large scale test sequence.In this paper,three methods are used to reduce the number of test sequences.The first is to sort test sequences according to the difficulty of tracking sequence,the second is to select representative sequences uniformly in the test sequence set by orthogonal experimental design,so as to reduce the amount of test data,the third is to screen test sequences using a uniform design experimentation.Experimental results show that the proposed performance evaluation method of the target tracking algorithm can effectively evaluate the performance of the algorithm and greatly improve the testing efficiency of the tracking algorithm.