Far infrared pedestrian detection method for changed scenes

Abstract

The invention discloses a far infrared pedestrian detection method for changed scenes. A sample extension target data set is screened out of auxiliary data on the basis of the Boosting-style inductive transfer learning algorithm DTL Boost. At first, a sample importance measurement model based on the k-nearest neighbor is utilized for evaluating the similarity between the auxiliary data and the target data, and corresponding initial weights are distributed for different samples in the auxiliary data. In the training process, the prediction inconsistency degree of member classifiers is explicitly defined, iterative updating is carried out on the current weights of the auxiliary data and the target data sample through the prediction error rate of the current member classifiers, a sample extension training set with the positive transfer ability is screened out of the auxiliary data, and the different member classifiers are encouraged to learn different parts or aspects of the target data. In this way, an integrated classifier with the stronger generalization ability is obtained through training, and the robustness of pedestrian detection in the new scene is enhanced.

Description

Technical field [0001] The invention relates to the technical field of pedestrian detection, in particular to a far-infrared pedestrian detection method oriented to changing scenes. Background technique [0002] There are usually inevitable differences in data distribution between training data and test data. This is one of the main reasons why most far-infrared pedestrian detection methods based on machine learning do not perform well when scene factors change greatly. Specifically, the traditional machine learning algorithms used in general pedestrian detection schemes meet the following basic assumptions by default: training data and test data are independent and identically distributed, that is, training data and test data usually come from similar or even the same scenes. When this basic assumption is not met, such far-infrared pedestrian detection schemes based on traditional machine learning algorithms are often difficult to successfully apply. [0003] However, in pedestri...

AI Technical Summary

Problems solved by technology

The advantage of the above method is that there is no need to mark the target pattern in the new scene in advance, but it may face two important problems: (1) The samples in the extended sample set may contain the noise of the data label, and it is difficult to guarantee the update by directly using the extended sample set with noise. (2) When the generalization ability of the pedestrian classifier is poor, it is usually unable to process or collect pedestrian patterns that cannot be detected correctly in new scenes, because such pedestrian patterns are not included in the auxiliary data In the middle, that is, the information cannot be obtained through learning, and it is difficult to ensure that the updated classifier can effectively identify various pedestrian patterns in new scenes

[0006] Migration learning tries to quickly and reasonably associate existing knowledge with new similar problems by "inferences from one instance". At present, the research on using transfer learning to solve the problem of far-infrared pedestrian detection in changing scenes is still in its infancy.

Images