Research On Obstacle Detection In Front Of Intelligent Vehicle Based On Lidar And Machine Vision

In recent years,the use of multi-sensor data fusion technology in intelligent vehicle detection of obstacle information in driving environment,mostly under good daylight conditions and other conditions,but for fog and night obstacle detection is less studied,and there are problems such as low accuracy and reliability of obstacle identification.In this paper,based on the data fusion of lidar and machine vision,obstacle detection in foggy or night traffic environment is carried out.The low contrast image enhancement method and more reasonable data fusion algorithm are studied to improve the accuracy and robustness of intelligent vehicle obstacle detection.The main contents of this paper are as follows:Firstly,based on the detection of obstacles by lidar when vehicles are moving straight,the method of setting vehicle lateral distance threshold to determine whether the target vehicle is located in the same lane in the detection cycle is proposed,and the validity of the target is verified by Kalman filter method and the decision of the target is made by life cycle method.Based on machine vision obstacle detection,the AdaBoost algorithm is proposed for the detection of obstacles.The original image is trained,the weight of positive and negative samples is redistributed,and many weak classifiers are obtained.Finally,strong classifiers are combined to achieve better recognition results.Secondly,the car body coordinate system is established,the radar and camera are calibrated separately,and any point in space is transformed into the image coordinate system.In order to improve the image recognition and the integrity of the target contour,the median filter is used to denoise the original image and the improved Single scale retinex algorithm is proposed to enhance the image.Thirdly,Dempster-Shafer evidence theory is used to fuse the data collected by radar and camera.The least square method is used to synchronize the radar and camera in the time of acquisition data,and the synchronization in space is realized through the calibration of sensors.The accuracy of D-S evidence theory is verified by the recognition of vehicles and pedestrians in real driving environment.The validity of D-S evidence theory is verified by comparing with single sensor in the integrity of target image contour extraction.Finally,a test platform for intelligent vehicle obstacle detection is built,and a single sensor and data fusion technology for obstacle detection in front of the vehicle is experimentally studied.When detecting radar obstacles,a set of radar data is collected and processed by the test platform to verify whether intelligent cars can distinguish vehicles in different lanes.Write Python data fusion obstacle avoidance program and import it into intelligent car for obstacle avoidance test.