Livestock tracking method and device

Abstract

The invention discloses a livestock tracking method and device. The livestock tracking method comprises the following steps: S1, receiving a monitoring video frame by frame, and performing target detection on a current video frame through a target detection network to obtain a tracking target of the current video frame; s2, when the current video frame is an initial video frame, distributing a livestock label to a tracking target of the current video frame; s3, when the current video frame is an intermediate video frame, matching a tracking target of the current video frame with a tracking target of a previous video frame according to a predefined multi-threshold step-by-step matching strategy, and when matching succeeds, enabling the tracking target of the current video frame to inherit a livestock label of the tracking target of the previous video frame; and S4, repeating the steps S1 to S3 until the current video frame is a termination video frame. According to the method, the situation that multiple livestock rapidly move or even disappear in an actual scene and the livestock are seriously adhered can be fully considered, and the livestock can be stably and accurately tracked.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a livestock tracking method and device. Background technique [0002] In order to realize the healthy breeding of livestock such as pigs, computer vision technology is gradually applied to recognize the behavior of livestock to pay attention to the living conditions of livestock. The process of target behavior recognition can be divided into target recognition, target tracking and behavior recognition, among which target tracking as a key link has important research significance. The more representative target tracking algorithms proposed so far are Deepsort algorithm and IOU-Tracker algorithm. [0003] The Deepsort algorithm uses recursive Kalman filtering to predict the position of each target, as well as frame-by-frame data association. In terms of target feature extraction, the image information of the current frame is extracted, and then the residual network trained...

AI Technical Summary

Problems solved by technology

[0004] For the above algorithm, the Deepsort algorithm needs to extract the image information of the target when tracking, and it is not suitable for livestock such as pigs with similar body sizes. In the case of fast movement of multiple livestock and serious adhesion between livestock, it is easy to cause the label exchange of livestock, and The Deepsort algorithm uses the residual network trained by ReID to extract target features, which is far slower than the target tracking algorithm that does not require image information; the IOU-Tracker algorithm uses a single-threshold matching method, which is easy for livestock to stick together. Matching errors are caused, and the IOU-Tracker algorithm does not predict the target position. When the livestock moves too fast or the livestock is lost, the livestock cannot be tracked

[0005] Therefore, the target tracking methods currently proposed are not yet perfectly suitable for livestock targets, and it is difficult to track livestock stably and accurately

Images