Target Tracking Method Based On Multi-feature Fusion Of Effective Blocks

At present,target tracking is a research hotspot in the field of computer vision.It is widely used in dynamic video monitoring,drone detection,intelligent detection and virtual reality.Its essence is to accurately locate the target in a continuous video image.Although there have been great progress in target tracking research at home and abroad,the existence of many uncertain factors such as occlusion,illumination changes and scale changes makes target tracking still a challenging and worthy research topicFor the current moving target method,the overall information in the target box is considered when tracking the target,but the shape of the tracked target is variability,resulting in the problem that the tracking method is not robust and accurate.This paper takes advantage of the local structure of the tracked target and treats the tracking target as a block particle.In the sequence Monte re-vote,the optimal position of the tracking target is calculated.And based on the local features of the image,the HOG feature has good adaptability to the target's subtle deformation,illumination changes,etc.,but if the target undergoes large deformation and occlusion,error tracking or leakage will occur;Important Perceptual Feature Color features are a global feature based on pixel points that are insensitive to target rotation,translation,and scale changes,but color features do not describe the local features of the target well and are not suitable for illumination changes.To this end,the two features are combined to describe the characteristics of the target model.When the global feature of the target is obtained,the local features of the target can also be obtained,and the accuracy of the target detection is improved.Therefore,we adopt a multi-feature linear fusion method based on HOG features and color features.Experiments in Chapter 4 show that tracking experiments are performed on 12 sets of published test video sequences.Compared with many excellent target tracking algorithms,the tracking algorithm of this paper achieves a target overlap rate of 0.67 and achieves the lowest average tracking error.14.83,the algorithm can perform target tracking stably and accurately,and the algorithm has better robustness.