Unmanned aerial vehicle target tracking method and system based on color histogram similarity

Abstract

The invention belongs to the technical field of computer vision, and discloses an unmanned aerial vehicle target tracking method and system based on color histogram similarity, and the method comprises the steps: extracting HOG, CNN and CN features of a to-be-tracked target as a feature subset for the initialization of a position filter, and increasing the number of samples through the overturningand the training of depth features through different sample labels; performing target search by using the trained filter to obtain a target position and a target scale; adding the current frame picture and the tracking result into a sample space and updating a sample space model; performing update discrimination by using the color correlation discrimination model and the primary and secondary peak discrimination model, and updating the position filter when the color histogram correlation of the predicted target and the previous frame target is relatively high and the difference between the primary and secondary peaks in the confidence map is obvious; and finally, updating the scale filter, and outputting a complete target position. According to the invention, complex conditions such as rapid target movement, large-amplitude deformation and shielding during target tracking of the unmanned aerial vehicle can be effectively handled.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to a method and system for tracking an unmanned aerial vehicle target based on the similarity of color histograms. Background technique [0002] At present, the existing technologies commonly used in the industry are as follows: [0003] At present, UAVs have the characteristics of rapid movement and high flexibility. They are suitable for completing tasks such as monitoring, investigation, and material transportation in complex scenarios. They are widely used in military, industrial, and civilian applications. UAVs equipped with cameras can obtain ground pictures and video information, and obtain timely and accurate information from them to complete subsequent targeted tasks. In the process of air-to-ground observation, targets on the ground or on the water usually require UAVs to keep paying attention to them during flight. Usually, the moving target is in th...

AI Technical Summary

Problems solved by technology

[0005] (1) The existing correlation filter tracking method is based on the entire target area template for tracking and model update, which leads to unsatisfactory processing effect on the occlusion problem; constructing positive and negative samples through the target area circular matrix leads to limited training samples and leads to overfitting The risk of false negative samples reduces the robustness to the background clutter problem

[0006] (2) In the existing deep learning tracking methods, the acquisition of the deep model requires effective learning of a large number of labeled training data, and the training process is expensive in terms of space and time

[0007] (3) At present, there are few target tracking technologies used on UAVs, and due to the hardware performance limitations of the onboard computer on UAVs, most of the tracking methods actually used have low tracking accuracy. easy to lose target

[0009] (1) The most direct way to solve the problem of fitting and false negative samples caused by limited training samples is to expand the training samples, but the conventional sample construction method has a high probability of constructing false samples and will further increase the computational complexity. The difficulty in solving this problem lies in finding a reasonable and less computationally intensive sample construction method

[0010] (2) In order to take advantage of the advantages of deep learning under the limitation of hardware conditions, the method of extracting deep features can be used, but the features extracted by different layers of networks contain different levels of information and have different feature resolutions, which will add a lot to the algorithm. Large amount of calculation and computational complexity, the difficulty of solving this problem lies in how to simplify the calculation under the premise of ensuring the effect of deep features

[0011] (3) After occlusion occurs, the algorithm will extract and learn the features of the occluder as target features, resulting in the model being no longer accurate. The difficulty in solving this problem lies in how to let the algorithm recognize that the target is occluded and then give up the occlusion. update study

Images