Vehicle-mounted thermal imaging pedestrian detection RoIs extraction method based on head feature points

Abstract

The invention discloses a vehicle-mounted thermal imaging pedestrian detection RoIs extraction method based on head feature points. The method comprises the following steps: extracting head center points of marked pedestrians in training data frame by frame, and counting the head center points of the pedestrians in the training data frame by frame according to a confidence coefficient of 95% to obtain a frame head region; training data pedestrian head feature points are extracted frame by frame, and a head feature point RoIs model is defined and trained; cutting out a current image frame by utilizing the frame head area, then calculating to obtain a double-threshold segmentation binary image of the current image frame, and combining the two to obtain a current frame head feature point area; sampling and screening a current frame head feature point area to obtain a current frame head feature point; calculating to obtain a current frame head feature point RoIs by using the trained head feature point RoIs model, and combining the current frame head feature point RoIs with the current frame double-threshold segmentation RoIs to generate a current frame RoIs; and measuring and adjustingthe current frame RoIs. According to the method, RoIs loss caused by factors such as shielding, segmentation adhesion and segmentation fracture can be effectively avoided, and the pedestrian detection performance is improved.

Description

technical field [0001] The present invention relates to the field of advanced assisted driving and pedestrian detection thereof, and more specifically, to a method for extracting RoIs for vehicle-mounted thermal imaging pedestrian detection based on head feature points. Background technique [0002] On-board thermal imaging pedestrian detection is one of the core technologies of the vision-based assisted driving system. The main goal is to detect pedestrians on the road ahead of the vehicle, and feed back the detection results to the driver, reminding the driver to make necessary reactions in time to reduce the number of people and vehicles. A collision occurs. Vehicle-mounted thermal imaging pedestrian detection includes two important steps: RoIs extraction and RoIs classification: the RoIs extraction step gives the possible position and scale of pedestrians in the image in the form of a rectangular frame, and the RoIs classification step judges whether there is a pedestria...

AI Technical Summary

Problems solved by technology

[0008] Existing technology 1 uses the SURF algorithm to extract RoIs for a long time, which cannot meet the real-time requirements, and lacks RoIs post-processing and tuning steps, and is susceptible to background noise interference, resulting in inaccurate positioning; prior art 2 extracts RoIs quickly, but requires a lot of ( >1000) RoIs can achieve a high recall rate, which has extremely high requirements on the classifier, and it is difficult to guarantee real-time requirements; although the existing technology 3 extracts RoIs quickly and the number of extracted RoIs is small, it cannot avoid occlusion, segmentation and adhesion. RoIs loss problem caused by segmentation fracture

Images