Two-dimensional false pedestrian identification method based on light field imaging in combination with LBP and SVM

Abstract

A two-dimensional false pedestrian identification method based on light field imaging combined with LBP and SVM comprises the following steps: (1) using a light field camera to shoot a large number ofpedestrian images and non-pedestrian images as a positive sample and a negative sample of an experiment; (2) obtaining an original 2D image and a depth image of the shot image through Lytro Desktop software; (3) performing pedestrian ROI region extraction on the original 2D image by using an LBP + SVM method, and judging whether the image is a pedestrian or not; and (4) identifying the depth image of the image judged as the pedestrian in the step (3) again by using the LBP + SVM method, and judging whether the pedestrian is a two-dimensional false pedestrian or not. According to the invention, the characteristic that the light field camera can obtain a depth image is utilized to eliminate false identification caused by two-dimensional false pedestrians.

Description

technical field [0001] The invention belongs to the field of computer vision and pattern recognition, in particular to a two-dimensional false pedestrian recognition method based on light field imaging combined with LBP and SVM. Background technique [0002] Recognizing and locating objects is an important research content in the field of computer vision. As a branch of object detection, pedestrian recognition is a special case of object detection. Man is a special kind of object, which not only has universality but also has various particularities. Therefore, pedestrian recognition has broad scientific research value and application prospects. [0003] Pedestrian recognition is primarily to identify pedestrians. However, in real life, pedestrians have different clothing, shape changes, backgrounds, and lighting conditions. This will pose severe challenges to pedestrian recognition, and there are widespread two-dimensional false Printing human body photos or character post...

AI Technical Summary

Problems solved by technology

[0003] Pedestrian recognition is primarily to identify pedestrians. However, in real life, pedestrians have different clothing, shape changes, backgrounds, and lighting conditions. This will pose severe challenges to pedestrian recognition, and there are widespread two-dimensional false Printing human body photos or character posters will have a fatal impact on pedestrian recognition, which will greatly cause misidentification and reduce the accuracy of recognition

[0004] At present, the following two methods are generally used for two-dimensional false pedestrian recognition: 1. The combination of multi-sensors is used to realize, but multi-sensors are expensive, cost-effective and there are problems of mutual compatibility between multi-sensors; 2. The application of structured light to realize Carry out two-dimensional false pedestrian recognition, but structured light is a kind of intrusive lighting, which may cause unnecessary harm to the human body

[0005] As a new type of camera, the light field camera has great development potential and broad application prospects. So far, no light field camera has been applied to the research of two-dimensional false pedestrian recognition.

Images