Pedestrian detection algorithm of fusing color image edges and HOD (Histogram of Oriented Depths)

Abstract

The invention discloses a pedestrian detection algorithm of fusing color image edged and an HOD (Histogram of Oriented Depths), belongs to the field of pedestrian detection algorithms, and solves problems such as the problems that features extracted by existing algorithms on color images largely ignore structural properties of pedestrians, and the color images have inherent deficiencies for certain non-ideal conditions, and thus image extraction only on the color features is liable to cause poor robustness. According to the method, a color RGB image and a corresponding depth image thereof areused as an image pair, edge feature extraction is carried out on the color image through shearlet transform (ST), HOD feature extraction is carried on the depth image, two features are fused to be used as a new pedestrian feature descriptor, the same is input into an SVM classifier for training, a corresponding classifier is obtained, and then a test picture is input into the classifier to detecta pedestrian position.

Description

technical field [0001] The invention belongs to the field of pedestrian detection algorithms, in particular to a pedestrian detection algorithm for fusing color image edges and depth direction histograms. Background technique [0002] Pedestrian detection is the process of using computer vision to determine whether there are pedestrians in an image or video sequence and to give precise positioning. It has broad application prospects in traffic safety, video intelligent monitoring, intelligent robot, image retrieval and other fields. Due to the non-rigid nature of pedestrians and the differences in their clothing and body, the appearance of pedestrians varies widely, which brings great challenges to detection. In order to accurately detect the position of pedestrians, there are currently the following common detection methods: [0003] (1) Extract a single feature from a color image. Dalal et al. proposed to use the histogram of oriented gradient feature (Histogram of Orie...

AI Technical Summary

Problems solved by technology

[0007] The present invention overcomes the deficiencies in the prior art, and solves the problem that most existing pedestrian detection algorithms only focus on local characteristics while ignoring the overall structure, resulting in detection that is not robust to noise, occlusion, and illumination, and aims to provide a fusion of color image edges and The pedestrian detection algorithm of the depth direction histogram, the present invention combines the edge features and the depth histogram features extracted from the color image by shear wave as a new feature descriptor, combined with the SVM classifier, to detect pedestrians

Images