Pedestrian detection method and system based on improved YOLOv3 model

Abstract

The invention discloses a pedestrian detection method and system based on an improved YOLOv3 model, and the method comprises the steps: obtaining a monitoring scene video collected in real time in an automatic driving scene, and extracting an image frame from the video; and inputting the image frame into a pre-trained improved YOLOv3 model, judging whether the image frame contains pedestrians or not by using the improved YOLOv3 model, if so, performing marking by using a bounding box, and if not, not carrying out any processing. The method and system have the advantages of being capable of improving the recognition capability of the sheltered target, effectively improving the pedestrian detection precision, being lower in false detection rate of small targets and sheltered targets, being higher in speed, having certain practical application value, and having good generalization performance.

Description

technical field [0001] The invention relates to a pedestrian detection method and system based on an improved YOLOv3 model, belonging to the technical field of computer vision. Background technique [0002] In recent years, computer vision technology based on deep convolutional neural networks has made remarkable achievements, which have promoted unmanned driving, surveillance video, Internet video retrieval processing, human-computer interaction, virtual reality, health care, intelligent security, etc. and other technological developments. Pedestrian detection is an important research content in autonomous driving technology, and it is also one of the research hotspots in the field of computer vision. [0003] With the development of deep neural networks, especially convolutional neural networks in object detection tasks, and the emergence of large-scale pedestrian detection public datasets, pedestrian detection algorithms based on deep neural networks have been applied in...

AI Technical Summary

Problems solved by technology

Two-stage methods such as R-CNN and Faster R-CNN are difficult to implement due to the difficulty of guaranteeing real-time performance, while one-stage algorithms such as YOLOv2 and YOLOv3 have outstanding performance in terms of real-time detection after continuous improvement, but Requires a large GPU memory, consumes a lot of resources, and has certain difficulties in deployment

Images