Target detection and tracking method and target detection and tracking device

Abstract

The present application provides a target detection and tracking method and a target detection and tracking device. The target detection and tracking method includes: judging whether the current frame collected by the collecting device is the first frame for determining the tracking target; if not, performing particle filter resampling on the current frame Carry out feature extraction according to the resampling position of each particle; Calculate the similarity of the feature of each particle and the feature model of the tracking target; The position of the tracking target of the frame is sent to the acquisition device for the acquisition device to adjust the monitoring angle. The present application realizes target detection and tracking under the premise of linkage with equipment.

Description

technical field [0001] The present application relates to the field of image processing, and in particular, to a target detection and tracking method and a target detection and tracking device. Background technique [0002] Video tracking algorithms have many applications in surveillance video intelligence. Currently, commonly used video tracking algorithms include discriminative and generative methods. The discriminative method is also called tracking-by-detection. With the in-depth application of neural network algorithms in the field of vision, people have found that it can well detect or identify objects such as people or cars in dynamic scenes, and can more accurately achieve the tracking requirements in intelligence, thus deriving many Video tracking algorithm based on deep learning target detection. [0003] Commonly used target detection algorithms generally have the problem of missed detection and false detection of detection results, and it is difficult to adapt...

AI Technical Summary

Problems solved by technology

[0003] Commonly used target detection algorithms generally have the problem of missed detection and false detection of detection results, and it is difficult to adapt to complex and changeable scenarios; while target detection and recognition algorithms based on neural networks require high computing power to meet the requirements. Therefore, the use of neural networks for detection and tracking at the same time has high requirements for hardware equipment. Ordinary monitoring equipment is often difficult to meet the requirements in consideration of real-time and practicality, and often focuses on algorithm effects and ignores the linkage with actual equipment.

Images