Research And Implementation Of Automated Driving Lane Change Decision Method Based On Image Semantic Segmentation

Along with the advancement of technology,in addition to improving the comfort and handling of vehicles and the safety of the vehicle itself,more and more companies have carried out research on autonomous driving technology.Compared with the current radar in the environmental perception is not sensitive to color information,object classification is not accurate and has a certain limitation on the ability to understand the environment,and with the continuous development and improvement of image processing technology,the use of vision-based perception Environmental information around the vehicle has gradually become a key technology for high-level autonomous driving.It has stronger environmental understanding,character and color resolution than non-visual methods,and automatic decision making based on the information about the car.The driving function plays a decisive role in making decisions by visually sensing the surrounding environment,such as decision-making in the direction of the lane change of the vehicle,and automatic parking.Therefore,this thesis makes an deeply study on the method of automatic driving lane change based on image semantic segmentation,and designs and implements the prototype system to realize the automatic decision of automatic driving lane change direction.The main work of this thesis is as follows:(1)A method for generating a vehicle environment map based on image stitching is proposed.Using multiple standard cameras to capture images of different angles around the vehicle,extract feature points of the image overlap region,and match the extracted feature points,determine the coordinate transformation matrix H of the image,and transform the image into the same coordinate system,and finally the generation of the vehicle environment map is realized by image stitching technology.(2)A method of acquiring variable lane travel area based on image semantic segmentation is proposed.Firstly,the image semantic segmentation method is used to identify the category of each pixel in the vehicle environment image,so as to perceive the environmental information around the vehicle body,such as vehicles,roads,pedestrians,etc.At the same time,the lane line detection algorithm is used to detect the lane line,and the left and right lanes are divided.The environmental information in each lane is determined according to the image semantic segmentation recognition result,and finally the lane change decision is made according to the extracted environment information to determine the variable lane travel area.(3)The prototype system,autonomous vehicle change lane decision-making,is designed and implemented,and the system performance test is carried out.The test results show that the proposed method can accurately realize the automatic decision-making of vehicle automatic driving lane change direction.