Pedestrian detection method based on aggregation channel characteristics

Abstract

The invention discloses a pedestrian detection method based on aggregation channel characteristics, and belongs to the technical field of digital image processing. The method is used for training a plurality of models (a big window model and a small window model), so as to solve a problem of false detection and detection leakage caused by the application of a single model. Moreover, the method greatly improves the confidence level of a window detected by the big window model in a process of window inhibition through a non-extreme value inhibition algorithm, and a non-extreme value inhibition threshold value is updated according to the sliding window scale ratio of the big and small models, thereby inhibiting a body part which is wrongly detected as a small target. Compared with a pedestrian detection method which employs a single model and is based on the aggregation channel characteristics, the method can effectively reduce the false detection rate and the detection leakage rate, andcan meet the real-time requirements.

Description

technical field [0001] The invention belongs to the technical field of digital image processing, and in particular relates to a pedestrian detection method based on aggregation channel features. Background technique [0002] With the rapid development of economy and science and technology, people's requirements for quality of life are getting higher and higher, and intelligent life has become a way of life that people pursue. Among them, intelligent transportation and intelligent monitoring are important components, and they are also a major research hotspot in the field of computer vision today. Pedestrians are one of the main targets in the field of intelligent transportation and intelligent monitoring. The development of its detection technology, especially the pedestrian detection technology based on image processing, is closely related to the realization of intelligence in this field. [0003] At present, pedestrian detection methods based on image processing are mainl...

AI Technical Summary

Problems solved by technology

[0009] The model matching method based on the sliding window can only detect targets larger than the model scale. In applications that need to detect small targets, if the model scale used is large, it will cause a large number of missed detections.

[0010] Based on the above problems, simply reducing the scale of the model will cause a large number of false detection problems due to the reduction of the number of features contained in the small-scale model. At the same time, simply using a small-scale model will also cause a large pyramid feature scale and slow detection speed. Difficult to meet real-time requirements

Images