Front vehicle behavior recognition method based on machine vision

Abstract

The invention relates to a front vehicle behavior recognition method based on machine vision. The method comprises the steps: firstly carrying out front vehicle detection through employing an improvedtiny-YOLOv3 algorithm; carrying out tracking of a target vehicle based on a KCF algorithm, a Kalman filtering algorithm and a Hungary matching algorithm after the detection of the target vehicle; andfinally constructing an LSTM according to a front vehicle detection and tracking result, and performing behavior recognition on the front vehicle. According to the improved tiny-YOLOv3 algorithm, a feature extraction network of the tiny-YOLOv3 algorithm is improved into an Inception module, a feature pyramid network is constructed, and a detection scale is increased to three. The front vehicle behavior recognition method based on machine vision is applied to the intelligent vehicle, can recognize various driving behaviors of the front target vehicle, and has good real-time performance and accuracy.

Description

technical field [0001] The invention belongs to the technical field of machine vision, and relates to a machine vision-based front vehicle behavior recognition method, which is applied to an intelligent vehicle, can identify various driving behaviors of a front target vehicle, and has good real-time performance and accuracy. Background technique [0002] With the development of intelligent technology and people's higher and higher requirements for travel methods, intelligent vehicles have become one of the current research hotspots. The detection, tracking and behavior recognition of the vehicle in front by the intelligent vehicle can provide the intelligent vehicle with real-time road environment information and identify the driving intention of the target vehicle, so that the decision-making control layer of the intelligent vehicle can make appropriate planning and avoidance for the current road environment. It is of great significance to reduce the occurrence of traffic a...

AI Technical Summary

Problems solved by technology

In order to increase the accuracy of vehicle behavior recognition, the present invention improves the tiny-YOLOv3 algorithm in the vehicle detection step, improves the feature extraction network of the original algorithm to an Inception module, expands the network width and improves the ability to extract features, and builds a feature pyramid network at the same time , using the up-sampling method to fuse high-level features and low-level features, increase the scale of the improved detection network to 3, and allocate more suitable detection scales to smaller detection targets; in order to solve the problem of poor tracking effect when the vehicle in front is blocked, in The vehicle tracking step is based on the KCF algorithm (Kernel Correlation Filtering Algorithm). If occlusion occurs, the Kalman filtering algorithm is used to predict the position of the blocked vehicle, and the Hungarian matching algorithm is used to realize long-term tracking of multiple vehicle targets in complex environments

Images