Research On Target Tracking Algorithm In Complex Condition

In the field of computer vision,target tracking has always been a basic and important research topic,and has a wide range of applications.With the continuous development and maturity of artificial intelligence technology in recent years,target tracking has more important application value.At the same time,It also puts forward higher requirements.Usually in the video scene,there are many interference factors such as occlusion,illumination variation,and background clutters.These extremely complex factors have also made target tracking face great difficulties and challenges.Therefore,this article studies the target tracking technology in complex situations.(1)This paper focuses on the research of target tracking algorithms in complex situations.For complex situations such as interference,occlusion,and target transformation,the tracking algorithm using only a single feature cannot effectively describe the target,and the anti-interference ability is poor.In this paper,a multi-feature fusion mechanism is adopted,and HOG features with good characterization proven by a large number of experiments and different depth features extracted from different layers of the deep network are used for feature fusion,which effectively improves the effect of target tracking.(2)For the case where the tracking effect is poor during the target tracking process,the cumulative error of continuous low-quality tracking may eventually lead to tracking failure.An adaptive model update mechanism is proposed,and the low-quality tracking situation is re-tracked by updating the model.When updating the model,fully consider the target's time-series feature information and current target feature information,build a reliable updated target model to re-tracking the target,and the overall tracking effect has been effectively improved.(3)For tracking situations under the influence of complex backgrounds,the algorithm may not be able to distinguish the target from the surrounding background clearly,resulting in unreliable target tracking effects.When this happens,the tracking response map is analyzed,and the speed information of the tracking target and the trajectory credibility are used for collaborative evaluation to remove the obvious drift points in the response map to further improve the accuracy of the tracking results.