Single target tracking method based on Siamese network

Abstract

The invention discloses a single target tracking method based on a Siamese network and belongs to the technical field of target tracking. The method comprises the following steps that firstly, a neural network part of a Siamese network is constructed, and the weight of the Siamese convolutional neural network is trained; in the training process, the method is carried out based on a neural networkmodel of an embedded loss function, meanwhile, all layers of features are fused, a random gradient descent algorithm is used for carrying out loss optimization, then classification and regression results are obtained through an RPN, and finally follow-up frame tracking is carried out according to the classification and regression results. The method is advantaged in that the tracking target can bebetter detected and positioned, an image detection method can be effectively fused to target tracking, a video is replaced by an image frame mode, and training cost and the calculation overhead are reduced, therefore, tracking processing efficiency is improved, and the effect of distinguishing similar objects is more obvious.

Description

technical field [0001] The invention belongs to the technical field of target tracking, and in particular relates to a single target tracking technology based on Siamese network prediction technology. Background technique [0002] With the rapid development of hardware, software and artificial intelligence, object tracking has become one of the hot spots in the field of computer vision research and has been widely used. The tracking and focusing of the camera and the automatic target tracking of the UAV all require the use of target tracking technology. In addition, there are specific object tracking, such as human body tracking, vehicle tracking in traffic monitoring systems, face tracking and gesture tracking in intelligent interactive systems, etc. Simply speaking, target tracking is to establish the positional relationship of the object to be tracked in a continuous video sequence, and obtain the complete trajectory of the object. Given the coordinate position of the o...

AI Technical Summary

Problems solved by technology

[0009] (1) Lack of necessary template updates;

[0010] (2) Since the window width remains unchanged during the tracking process, when the target scale changes, the tracking will fail;

[0011] (3) When the target speed is fast, the tracking effect is not good;

[0012] (4) The histogram feature is slightly deficient in the description of the target color feature, and lacks spatial information

Images