Sky background infrared imaging multi-target tracking method

Abstract

The invention relates to a sky background infrared imaging multi-target tracking method, comprising the steps that A, detecting the state of a current frame every k frames; if the target is in the detection state, obtaining centroid coordinates of all targets; if the state is a tracking state, updating the tracker template to obtain a target position; B, obtaining three matching states during state detection; C, initializing the position of the tracker when the target is matched with the tracker; D, when the tracker is distributed but the detector does not detect the target, updating each filter template to obtain the current frame position, adding 1 to the value of the second variable, and deleting the tracker when the value of the second variable reaches a threshold value; E, initializing the position of the tracker when a new target is detected but the tracker is not allocated; F, correcting the predicted position to obtain a final target tracking position; G, displaying a trackingresult; and H, if the frame is the last frame, ending, otherwise, returning to the step A to execute. According to the sky background infrared imaging multi-target tracking method, multiple targets inthe sky background can be rapidly detected and captured, and false targets can be eliminated, and interested targets can be identified.

Description

technical field [0001] The invention relates to a computer vision target tracking method, in particular to a sky background infrared imaging multi-target tracking method. Background technique [0002] With the development of science and technology, people have conducted extensive and in-depth research on the theory of target detection and tracking, and achieved remarkable results. However, most of these methods are dedicated algorithms researched and developed for specific applications and specific scenarios. For real-time target detection and tracking in multi-target dense scenes, existing methods still have great limitations. Especially for the stable tracking of multiple targets and point targets in complex backgrounds, there is still a lack of efficient and adaptable general-purpose technology. Therefore, developing a multi-target tracking and detection technology with good real-time performance and high robustness is still a huge challenge in the field of computer visi...

AI Technical Summary

Problems solved by technology

[0004] The current target tracking methods mainly include (1) based on regional information, such as the template matching method, which is simple, accurate, and fast, but it cannot adapt to complex environments such as severe deformation of the target, and it is easy to cause target loss in this case; (2) Based on the model information, the geometric model of the target is established, and then the model is searched. This method is also difficult to solve the occlusion problem, and the lack of color information in the infrared environment has weaker anti-occlusion capabilities; (3) Based on the Bayesian framework , that is, on the basis of capturing the initial state of the target and the target features through feature extraction, a space-time combined target state estimation is performed, which can be used for target position estimation in the case of being occluded, but the algorithm complexity is high; (4) based on Deep learning methods have good robustness, but they are prone to data loss problems, and the network training speed is difficult to meet the real-time requirements; (5) Based on correlation filtering, this type of method is generally relatively fast. Among them, based on kernel correlation filtering (Kernelized Correlation Filters, KCF) target tracking has the characteristics of fast and high accuracy. Compared with the tracking algorithm based on Structured output tracking with kernels (STRUCK) and Tracking-Learning-Detection , TLD) frame tracking and other algorithms, its tracking speed is increased by nearly 10 times, compared with OTB50 (Object tracking benchmark, the first 50 tracking video competition sequences) accuracy is 43.1% minimum output mean square error (Minimum Output Sum of Squared Error, MOSSE) filtering algorithm, and has extremely high accuracy, and the accuracy can reach 73.2% in the case of using HOG features

Images