基于神经网络的智能驾驶模式识别研究

为满足智能驾驶汽车高级驾驶辅助系统(ADAS)功能研发和验证的需求,提高ADAS功能的准确性,设计了一款基于神经网络的智能驾驶模式识别程序,该程序由数据采集、目标检测、场景识别预测3个模块组成。数据采集模块利用ESR毫米波雷达、前置摄像头对交通环境及周围车辆的数据信息进行采集;目标监测模块通过控制算法选择判断触发各类ADAS功能场景的最可疑目标;场景识别处理模块以汽车制造商提供的大量自然驾驶数据的场景挖掘结果为依据,利用神经网络学习各类ADAS场景的特征行为,并通过约束条件对各类ADAS功能场景的识别结果进行实时判定。通过开放道路试验进行验证,结果表明,该程序的场景识别结果准确率可达到99.86%。

Research on Intelligent Driving Pattern Recognition Based on Neural Network

In order to meet the requirements of functional development and verification of ADAS of intelligent driving vehicle,and improve the accuracy of ADAS function,an intelligent driving pattern recognition program is designed based on neural network.The program consists of data acquisition module,target monitoring module,scene recognition&prediction module.The millimeter wave radar ESR and forward camera are used in the data acquisition module to collect data of traffic environment and surrounding vehicles.The target monitoring module uses the control algorithm to select the most doubtful target which can trigger various ADAS functional scenarios.Based on the scene mining results of a large amount of natural driving data provided by automobile manufacturers,the scene recognition processing module uses the neural network to learn the characteristic behaviors of various ADAS scenes,and conducts real-time judgment of the recognition results of various ADAS functional scenes through constraints.Open road test is used to verify the results,which shows that the scene recognition results of this program have high accuracy of up to 99.86%.