Surrounding vehicle behavior recognition method based on smartphone and RNN

Abstract

The invention discloses a surrounding vehicle behavior recognition method based on a smartphone and RNN, and belongs to the field of intelligent drive. The method comprises the steps of a, an offlinetraining link, wherein typical surrounding vehicle behaviors are concluded and divided, vector encoding is carried out on relative characteristics and vehicle behaviors of a tracked vehicle and vehicle owner by using vehicle driving data collected by the smartphone, and the relative characteristics and vehicle behaviors of the tracked vehicle and vehicle owner serve as a training set of RNN parameter learning; b, an online testing line, wherein based on a real-time traffic scene, a main vehicle combines the driving data of the tracked vehicle with the driving data of an own vehicle through 4Gcommunication to form a new characteristic matrix as the input of the trained RNN to recognize a behavior mode which surrounding vehicles belong to. According to the surrounding vehicle behavior recognition method based on the smartphone and RNN, the smartphone is taken as hardware of data collection and vehicle communication and has the advantages of feasibility and convenience; the characteristic of the RNN good at processing multi-dimensional matrix operation are used, the relative characteristics of the own vehicle and surrounding vehicles are enriched, the recognition rate is improved, and it is ensured that the behavior recognition has higher real-time performance.

Description

technical field [0001] The invention belongs to the field of vehicle intelligent driving, and in particular relates to a surrounding vehicle behavior recognition method based on a smart phone and RNN. Background technique [0002] In recent years, vehicle behavior recognition is shifting from fixed location monitoring system vehicle behavior recognition to dynamic surrounding vehicle behavior recognition based on driving vehicles. The key to behavior recognition is to learn the behavior pattern of the vehicle, establish a behavior recognition model, and then perform vehicle behavior recognition through the trained vehicle behavior recognition model, and even predict vehicle behavior. [0003] In order to provide a training set for the surrounding vehicle behavior recognition model, it is first necessary to use smartphone sensors to collect vehicle driving status information, and establish a vehicle-to-vehicle communication group through the 4G network. At present, there are...

AI Technical Summary

Problems solved by technology

[0005] In terms of vehicle behavior recognition modeling, traditional methods generally use HMM, that is, hidden Markov model. Although in terms of model establishment, HMM can well meet the requirements of vehicle behavior modeling for sequence models, but the classification ability of HMM is relatively low At the same time, the observation sequences used as model input can cover very limited vehicle driving characteristics, so the misrecognition rate of HMM is still high

Images