Thermal infrared human body target recognition method based on deep abstract features

Abstract

The invention relates to the technical field of infrared recognition, in particular to a thermal infrared human body target recognition method based on deep abstract features. Comprising the following steps: acquiring a static image or a sequence image by using a thermal infrared imaging device; highlight areas are obtained through image segmentation, the highlight areas which do not meet the basic features of the human body targets are filtered out through human body form priori knowledge, and candidate human body targets are obtained; calculating the multi-scale monotonic wave representation of the candidate human body target; calculating the depth abstract features of the candidate human body targets; and performing classification judgment on the candidate targets based on the depth abstract features and a shallow classifier. According to the method, the classification separation of the complex features of the thermal infrared image region is realized, and the feature representation definition is enhanced; automation of data essential feature extraction is realized; noise in the feature components is eliminated to a large extent; the data quality of the deep abstract features is ensured; the requirement for the number of training samples during classifier training is lowered, namely, the over-fitting risk is lowered, and the classification recognition precision is improved.

Description

technical field [0001] The invention relates to the technical field of infrared identification, in particular to a thermal infrared human body target identification method based on deep abstract features. Background technique [0002] Infrared vision is traditionally mainly used in the military field, but in recent years, with the development of thermal infrared imaging technology and the performance improvement of infrared thermal imaging equipment, the price has dropped, and the accompanying promotion of artificial intelligence technology, the human body recognition technology based on thermal infrared thermal imaging is increasingly Expand to intelligent transportation, security monitoring, intelligent robots and many other fields. [0003] The outstanding advantage of thermal infrared imaging is that the imaging of environmental objects does not depend on the reflection of visible light by the object, but on the amount of thermal radiation from the environmental objects....

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to provide a thermal infrared human body target recognition method based on deep abstract features, which is used to solve the problem that the existing technology is very dependent on massive training data, and the interpretability of automatic feature acquisition is not strong, and the network structure design Technical issues lacking a relatively clear guiding theory

Images