Color Feature Assisted Intelligent Vehicle Target Tracking Method

Intelligent vehicle target tracking is the key technology of vehicle environment perception,but only relying on a single feature information can not achieve ideal tracking effect for intelligent vehicle target tracking system.As an important part of intelligent vehicle,vision sensor can provide a lot of assistant information.Target color feature is the most direct description of target appearance characteristics.Using target color feature can assist the data association process,so as to improve the performance of the whole target tracking system.Therefore,it has theoretical significance and practical value to introduce color features into the process of target tracking.In this thesis,the target color feature is quantized into color eigenvector,and the effectiveness of color eigenvector as assistant feature in intelligent vehicle target tracking is studied.Firstly,the region of interest of the target is obtained,and the background interference is removed by reducing the region of interest.Then,the region of interest is downsampled to reduce the amount of calculation.Then,the largest value of color in the color histogram is extracted,and the target color eigenvector is obtained.In order to verify the feasibility of using target color as assistant feature,this thesis collects the target data of vehicles and pedestrians in different scenes,analyzes the influence on the color components of each channel when the share changes by narrowing the region of interest,and then analyzes the change of the color eigenvector of vehicles and pedestrians with time.Through the corresponding experiments and analysis,it can be verified that it is effective to use color eigenvector as assistant feature.In order to effectively improve the accuracy of data association,the color feature is introduced into the state vector and measurement vector of the target in this thesis,and a tracking model of the target is established with color features.Firstly,the tracking gate is set up with the one-step prediction position of the target as the center in the process of tracking,and the target measurements are filtered,and then the candidate measurements are associated with the target trajectory.In order to make effective use of color features in the association process,the concept of association probability vector is introduced in this thesis,this vector represents the position,wide,height and color of the target respectively.In the association process,the association probability is calculated according to the components of the association probability vector.The association probability vector is used to participate in the data association,and then the target state is fused and estimated according to the association probability vector norm.Through the corresponding simulation experiments and analysis,it shows that adding color features can effectively improve the target tracking accuracy of intelligent vehicles.According to the color feature assisted target tracking method,the intelligent vehicle target tracking software is developed,and the real vehicle experiment is carried out in the actual road scene.The experimental results show that the target tracking method improves the accuracy of data association and the performance of intelligent vehicle target tracking system.