Real-time compression tracking method of multi-characteristic transfer learning

Abstract

The invention discloses a real-time compression tracking method of multi-characteristic transfer learning. The method comprises the following steps: generating a training sample, and respectively extracting a positive sample and a negative sample for training a classifier from a current frame and a video frame which previously tracks an object; extracting the characteristics of an object and a background in the training sample, projecting an extracted high-dimension characteristic to a low-dimension characteristic of a compression domain by use of two complementary sparse mapping matrixes, and generating two balanced characteristics for indicating the object and the background; constructing and updating the classifier, and training a naive Bayes classifier for classifying samples to be detected by use of the characteristic of the compression domain; and employing a secondary object search strategy. According to the invention, by use of the coarse-to-fine secondary search strategy, the quantity of generated scanning windows is reduced, the quantity of the samples to be detected is reduced, the object search process is accelerated, the scanning window with the maximum response value is taken as a tracked object, and the training sample and the classifier are updated based on this.

Description

technical field [0001] The invention belongs to the technical fields of computer vision and machine learning, in particular to a real-time compression tracking method for multi-feature transfer learning. Background technique [0002] The target tracking system generally consists of three parts: (1) the target appearance model, which is used to evaluate the similarity between the candidate area and the target; (2) the motion model, which is used to model the motion state of the target in a continuous period of time ; (3) The search strategy, which is used to search out the most likely target area in the current frame; among these three components, the target appearance model is an indispensable part. [0003] Offline tracking learns the appearance model of the target offline. This method needs to collect a large number of samples of the target in advance for the training of the classifier, and the training process takes a long time and cannot adapt to the change of the target...

AI Technical Summary

Problems solved by technology

[0003] Offline tracking learns the appearance model of the target offline. This method needs to collect a large number of samples of the target in advance for the training of the classifier, and the training process takes a long time and cannot adapt to the change of the target appearance; online tracking updates the target in real time. The appearance model can adapt to changes in the appearance of the target. During the training process of the classifier, the tracked target is usually used as a positive sample, and negative samples are selected around the positive sample. This method obtains fewer training samples and requires repeated training. and testing to improve the accuracy of the classifier; multi-instance tracking selects multiple positive samples and multiple negative samples for the training of the classifier. The above method is easy to introduce background information when the target is partially occluded, resulting in incorrect update of the classifier. , eventually lead to tracking drift or even loss of target

Images