基于深度学习的无人机视频车辆行为分析

Deep Learning-based Vehicles Behavior Recognition for Unmanned Aerial Vehicle Data

本文提出一种基于深度神经网络的车辆行为分析框架用于交通参数的智能化提取。该框架先基于利用单镜头多盒探测器(SSD)检测实现车辆检测与识别,然后在视频序列中对车辆进行线性跟踪,根据追踪结果对车辆轨迹进行建模,设计基于LSTM的深度学习模型对车辆行为进行分类。本文以深圳市龙井地铁站T型路口的无人机数据作为研究对象,进行实验。结果表明,提出的方法能够提高车辆识别与跟踪算法的精度和效率,并获得准确的车辆分类结果及轨迹识别。

This paper proposes a vehicle behavior analysis framework based on deep neural network for intelligent extraction of traffic parameters.The framework is based on the use of single-lens multi-box detector(SSD)detection to achieve vehicle detection and recognition,and then linearly track the vehicle in the video sequence,model the vehicle trajectory according to the tracking result,and design a deep learning model based on LSTM Vehicle behavior is classified.This paper takes the UAV data from the T-junction of Shenzhen Longjing Metro Station as the research object and conducts experiments.The results show that the proposed method can improve the accuracy and efficiency of vehicle identification and tracking algorithms,and obtain accurate vehicle classification results and trajectory recognition.