Vehicle detection method and device based on adaptive key point heat map

Abstract

The invention provides a vehicle detection method and device based on an adaptive key point heat map, and belongs to the technical field of computer vision and target detection and recognition. The method comprises the steps that a key point heat map is rendered through an adaptive variable-radius Gaussian kernel, the center point of a vehicle target is extracted, and then information such as thesize and the position of the vehicle target is obtained through regression according to the width and the height of the vehicle target; rendering of the key point heat map adopts a self-adaptive strategy, the Gaussian kernel radius changes along with the change of the target width and height, a boundary frame predicted under the condition of road congestion is more compact, the calculation coordinate range is reduced from the whole feature map into the target boundary frame, and network convergence can be accelerated; the method can effectively solve the problems of missing detection, false detection and low detection precision among densely shielded vehicles in a crowded road video scene, can meet the requirements for vehicle detection speed and accuracy in a crowded road monitoring video, and is of great significance to intelligent traffic management work.

Description

technical field [0001] The invention designs a vehicle detection method and device based on an adaptive key point heat map, and belongs to the technical fields of computer vision, target detection and recognition. Background technique [0002] In recent years, the traffic video surveillance system has been widely used in the field of intelligent transportation. Based on the method of deep learning to monitor and intelligently analyze the real situation of road traffic, it is very helpful for alleviating traffic congestion, emergency handling of traffic accidents, traffic route planning, and road vehicles. Things like breach management are critical. Vehicle detection is the basis of vehicle classification and vehicle tracking. How to realize real-time and accurate detection of vehicles in a congested environment is a challenging task. [0003] The existing vehicle detection methods based on deep learning are mainly divided into two categories: one is the vehicle detection me...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the problems existing in the prior art, the object of the present invention is to provide a vehicle detection method and device based on an adaptive key point heat map, which can effectively solve the problems of low vehicle detection accuracy and slow speed under road congestion conditions

Images