Confidence Evaluation and Classifier Update Method of DCF Tracking Based on Neural Network

Abstract

The invention relates to a neural network-based DCF tracking confidence evaluation and classifier updating method, which belongs to the technical field of computer vision. First design and train a small-scale convolutional neural network of the response graph analysis network. In correlation filtering tracking, when the classifier is convolved with the features of the search region, the resulting response map is fed into this network, and the output is used as the tracking confidence score of the classifier for this frame. When the score is lower than the preset low confidence threshold, it is considered that the target has been severely disturbed, and the update is stopped to prevent the target model from being polluted, and the updated learning rate and time interval are adaptively adjusted by the confidence score. When the frames have a high confidence score, the target appearance is considered to be in a highly similar state, and the update interval is increased to alleviate the overfitting phenomenon. The method of the invention can significantly enhance the adaptability of correlation filter tracking to interference factors such as illumination changes, occlusion, out of view, etc., and improve space and time efficiency.

Description

technical field [0001] The present invention relates to a DCF (Discrimitive Correlation Filters, discriminative correlation filter) tracking confidence evaluation and classifier update method based on neural network, in particular to a DCF tracking confidence evaluation and classifier based on response graph analysis convolutional neural network The updated method belongs to the technical field of computer vision. Background technique [0002] Moving object tracking technology is one of the important research fields of computer vision, and has been widely used in the fields of automatic driving, human-computer interaction, security monitoring and so on. [0003] At present, the challenge of moving target tracking technology is: how to maintain accurate and real-time tracking under the influence of complex interference factors such as changes in background lighting conditions, occlusion, fast motion, out of view, posture shape changes, and size changes. [0004] In the prior...

AI Technical Summary

Problems solved by technology

[0008] The purpose of the present invention is to solve the problem of high interference and low efficiency when dealing with moving target tracking tasks due to the lack of tracking confidence evaluation measures in the existing correlation filter tracking method in the field of computer vision, which affects the recognition accuracy and real-time performance. Technical problem, a neural network-based DCF tracking confidence evaluation and classifier update method is proposed

Images