Pedestrian recognition method based on light field camera and HOG and SIFT mixed features

Abstract

A pedestrian identification method based on a light field camera and HOG and SIFT mixed features comprises the following steps: (1) using the light field camera to shoot a large number of pedestrian images as a positive sample of an experiment; (2) shooting a large number of non-pedestrian images by using a light field camera as a negative sample of an experiment; (3) obtaining an original 2D image and a depth map of field of the shot image by using Lytro Desktop software; (4) performing ROI region extraction on the original 2D image by using a method of combining HOG + SIFT mixed features andSVM, and judging whether pedestrians are included or not; and (5) identifying the depth map of the pedestrian judged in the step (4) by using the method of combining the HOG + SIFT mixed features with the SVM again, and judging whether the deceptive pedestrian is included or not. According to the invention, the characteristics that the light field camera can record the scene 4D information and acquire the depth image are utilized to eliminate false identification caused by pedestrian traffic, so that the pedestrian identification capability is enhanced, and the working cost is low.

Description

technical field [0001] The invention belongs to the field of computer vision and pattern recognition, in particular to a pedestrian recognition method based on the mixed features of HOG and SIFT of a light field camera. Background technique [0002] With the development of computer artificial intelligence, many tasks that require a lot of manpower can now be handed over to computers, and computer vision is a recent research hotspot. Computer vision is a discipline that uses cameras and computers to detect and identify targets to obtain the data and information we need. Pedestrian recognition is an important part of the computer vision research field, and it has played a very important role in video surveillance, assisted driving, intelligent human-computer interaction and other fields. [0003] Pedestrian recognition is first and foremost to identify pedestrians. However, in real life, pedestrians are a kind of non-rigid body targets, which are not like vehicles. No matter ...

AI Technical Summary

Problems solved by technology

Although these algorithms have obvious advantages, the problem of pedestrian recognition cannot be solved by only using traditional two-dimensional cameras.

[0005] For this problem, we can theoretically use the stereo vision measurement method to solve it, but the stereo vision method requires the coordinated processing of multiple multi-sensors, thus increasing the experimental cost

Images