Pedestrian detection method

Abstract

A pedestrian detection method provided by the invention comprises the following steps: S100, training SVM (Support Vector Machine) classifiers, and training the two SVM classifiers by taking a human head and two legs as positive samples and taking a pedestrian background as a negative sample; S200, constructing an image pyramid by using the to-be-detected image; S300, sliding the window, and intercepting a target window on each scale of the image pyramid; S400, extracting HOG features from the target window; S500, sending the HOG features extracted in the step S400 into an SVM classifier, andjudging whether the target window is the head of a pedestrian, the two legs of the pedestrian or the pedestrian background in a classified mode; And S600, combining pedestrians according to the head and leg results detected in the step S500, and judging whether the to-be-detected image contains the pedestrians or not. Therefore, the pedestrian detection rate is improved by detecting the combination strategy of the human head and the two legs, and the problem that the pedestrian detection rate is not high is solved through the pedestrian detection and recognition method based on the SVM and theHOG.

Description

technical field

[0001] The invention relates to the technical field of intelligent identification, in particular, a pedestrian detection and identification method based on SVM and HOG, which is applicable to the fields of intelligent assisted driving, intelligent monitoring, pedestrian analysis, intelligent robots and the like. Background technique

[0002] In recent years, with the rapid development of intelligent detection, pedestrian detection has also entered a relatively rapid development stage, but there are still many problems to be solved, especially in terms of performance and speed.

[0003] At present, pedestrian detection technology can be roughly divided into two categories: 1. Based on background modeling, use the background modeling method to extract the foreground moving target, perform feature extraction in the target area, and then use the classifier to classify to determine whether pedestrians are included , the main problems in background modeling at pres...

Images