Target tracking method based on fast tensor singular value decomposition feature dimensionality reduction

Abstract

The invention discloses a target tracking method based on fast tensor singular value decomposition feature dimensionality reduction. The target tracking method comprises: extracting multiple featuresfrom each frame of video data, and constructing a tensor structure; performing singular value decomposition on the constructed tensor; and training related filters by using the features after dimension reduction, and tracking the target. According to the method, the number of features can be effectively reduced, the tracking speed is increased, and compared with a traditional vector-based principal component analysis feature dimensionality reduction mode and other modes, the structure information of the features is better reserved; the tensor singular value decomposition has invariance to therotation of the feature to enhance the robustness of the tracker to the target rotation.

Description

technical field [0001] The invention relates to a target tracking method based on fast tensor singular value decomposition feature dimensionality reduction, and belongs to the technical field of video target tracking. Background technique [0002] Target tracking is of great significance to the development of robotics, unmanned aerial vehicle, automatic driving, navigation and guidance and other fields. For example, in the process of human-computer interaction, the camera continuously tracks the human body's behavior, and through a series of analysis and processing, the robot can understand the human body's posture, movement, and gestures, so as to better realize the friendly communication between humans and machines; During the target tracking process of the UAV, the visual information of the target is continuously obtained and transmitted to the ground control station, and the video image sequence is analyzed through an algorithm to obtain the real-time position informatio...

AI Technical Summary

Problems solved by technology

[0003] The target tracking process involves a series of challenges such as illumination changes, scale changes, in-plane rotation, out-of-plane rotation, occlusion, deformation, motion blur, fast motion, background spots, and low resolution. In recent years, the "correlation filter" class of target tracking methods Not only the tracking speed is fast, but also the tracking accuracy is good, but with the continuous increase of various features, the tracking speed of the correlation filter drops seriously

[0004] In recent years, image features used in correlation filtering have been increasing, such as color name features, gradient histogram features, and deep features of deep convolutional neural networks. These features have greatly improved tracking accuracy, but make correlation filtering The tracking speed of the device drops rapidly

Images