Real-time significance pedestrian detection method based on detail discrimination

Abstract

The invention discloses a real-time significance pedestrian detection method based on detail discrimination. The method includes: carrying out parameter pre-training in an existing image data set; reading video data in the monitoring video and converting the video data into a processible image format; performing feature extraction on the image in the step B; and migrating the pre-trained model parameters to a YOLO neural network model, performing network model training on a pedestrian detection data set, and performing pedestrian detection by using the trained network model according to the pedestrian feature H and the saliency feature. Parameter pre-training is carried out on an existing public image data set and is introduced into a YOLO neural network model; pedestrian features are distinguished according to detection significance area features and emphasis details, then pedestrians are detected in real time; and compared with the method that a YOLO model is purely utilized, the multi-scale prediction generalization ability of the model is effectively improved, the detection rate of the system is greatly increased, and the omission ratio is reduced.

Description

technical field [0001] The invention relates to the field of pedestrian detection, in particular to a real-time significant pedestrian detection method based on detail discrimination. Background technique [0002] Pedestrian detection is an important research direction in the field of computer vision, which detects whether there are pedestrians in the image by analyzing the image. It has a wide range of applications in the fields of vehicle assisted driving, intelligent monitoring, and intelligent robots. For example, the vehicle auxiliary driving system can use pedestrian detection technology to detect pedestrians in front of the vehicle and remind the driver to avoid; the intelligent monitoring system can use pedestrian detection technology to detect pedestrians in the monitoring screen, so as to analyze the behavior of pedestrians and track suspicious persons; Intelligent robots can use pedestrian detection technology to find pedestrians in the surrounding environment, s...

AI Technical Summary

Problems solved by technology

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.

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 present: (1) must adapt to changes in the environment (such as changes in illumination caused by changes in image chromaticity); (2) camera shakes cause image shakes (such as movement when holding a camera to take pictures); (3) Ghost areas often appear in object detection. Ghost areas mean that when an originally stationary object starts to move, the background difference detection algorithm may mistakenly detect the area covered by the original object as moving. This area It becomes a Ghost. Of course, when the original moving object becomes stationary, it will also introduce the Ghost area. The Ghost area must be eliminated as soon as possible during the detection.

[0004] In the development of pedestrian detection from the two stages of feature extraction and metric learning in traditional methods to the end-to-end learning process based on deep learning, there are currently the following main problems in pedestrian detection technology: (1) Pedestrians have different postures and clothing, Complex background, different pedestrian scales and different lighting environments; (2) the distribution of the extracted features in the feature space is not compact enough; (3) the performance of the classifier is greatly affected by the training samples; (4) when offline training The negative samples of cannot cover all real application scenarios

Early pedestrian detection uses carefully designed manual features (SIFT, HOG, etc.) to describe pedestrian features, and then classifies them through a Support Vector Machine (SVM) classifier. This process belongs to shallow learning, and the description of pedestrian features Limited ability, often can not achieve the desired effect

Images