Pedestrian detection method for crossroad scene

Abstract

The invention discloses a pedestrian detection method for a crossroad scene, and the method comprises the following steps: S1, preparing and arranging a training sample set of a crossroad scene, the training sample set comprising a positive sample set and a negative sample set; S2, establishing a background model of the crossroad scene by using an average background model algorithm; S3, training by using the training sample to obtain a classifier of each part; S4, combining the classifiers of each part by using a behavior knowledge space algorithm to form a final classifier; and S5, inputtinga to-be-detected picture into the model, and completing the detection and recognition of the pedestrian in the to-be-detected picture. According to the method, through fusion of the algorithm based onthe overall characteristics and the algorithm based on multiple parts, and combining the preprocessing of background modeling by using the similarity of scenes, a rapid, feasible and good-effect scheme is provided for pedestrian detection in a crossroad scene.

Description

technical field [0001] The invention relates to a pedestrian detection method, in particular to a pedestrian detection method for a crossroad scene, and belongs to the field of computer vision technology and image recognition. Background technique [0002] With the rapid development and continuous improvement of the Internet, artificial intelligence technology and intelligent hardware, the types and quantities of image and video data that record people's daily life are also increasing, which makes computer vision technology in the analysis of people's daily life. The role played in the process is getting bigger and bigger, and because of this, the current research on computer vision technology is becoming more and more popular. [0003] Computer vision technology, as the name suggests, is a natural science that studies how to use machines to "see" things. More precisely, it is a science that uses cameras, cameras and computers to track, identify and detect pictures and vide...

AI Technical Summary

Problems solved by technology

It generally includes: the detection accuracy of pedestrians with different clothing, appearance, and posture is not high enough; the distribution of the extracted features in the feature space is not compact enough; the performance of the classifier is greatly affected by the training samples, that is, the general applicability of the scene is not high enough ; Negative samples during offline training cannot cover all real application scenarios, etc.

Images