Research On Obstacle Detection And Ranging Method For Unmanned Vehicles

With the rapid development of the world economy,automobiles have become an indispensable means of transportation in modern life,and the problems of traffic safety have become a problem that cannot be ignored.Therefore,research on unmanned vehicles has become more and more important.Unmanned vehicles can alleviate traffic congestion,increase traffic safety,and reduce air pollution.To save people time and improve the quality of life,one of the key technologies of unmanned vehicles is the detection and ranging of obstacles in front of unmanned vehicles.In the current vehicle obstacle detection system,common sensors include highfrequency radar(millimeter wave),ultrasonic,infrared laser,radar,etc.Each sensor adapts to different scenes,in order to adapt to various scenarios,current car manufacturers and Assisted driving system providers generally combine the information collected by various sensors to make decisions,and customize specialized hardware devices to achieve vehicle detection and distance measurement.However,such platform construction costs are high and the popularity is low.Current vehicle ranging techniques are susceptible to external conditions such as the environment,especially when the vehicle is in a complex and variable traffic scene or a bad weather environment,and it is difficult to obtain a satisfactory recognition rate and robustness.Because the vision-based front vehicle detection system has low cost and rich information,and the rapid development of computer vision in recent years provides a technical basis,it has attracted more and more researchers in related fields.In view of the above problems,this paper studies the methods of vehicle detection and ranging in front of obstacles.In terms of vehicle detection,the vehicles in front are initially positioned and accurately positioned to adapt to a variety of complex environments,including rainy days,urban roads,sudden changes in light,etc.In vehicle ranging,traditional methods of vehicle ranging often Based on the characteristics of the vehicle itself,but when the environment changes,the characteristics of the vehicle itself are ignored,affecting the ranging result.Therefore,this paper adopts the monocular vision method based on license plate feature information to measure distance,which avoids the influence of ranging results due to environmental changes.The main contents of this paper are the following four aspects.(1)Vehicle image preprocessing.The image of the vehicle collected by the camera will cause image noise due to weather,camera shake and other reasons.The image is grayed out,image grayscale enhanced,image filtered,image segmented and mathematically processed by digital image preprocessing.The contrast analysis experiments were carried out to select the image processing method suitable for vehicle detection and positioning.(2)Lane detection.The premise of vehicle detection and positioning is to detect the lane line on the road where the vehicle is traveling.In this paper,the Hough transform is used to detect the lane line,and the traditional Hough transform method is improved.The inner lane based on linear slope constraint is proposed.The line detection method can overcome the influence of uneven light and effectively eliminate the interference of objects with similar linear characteristics on the road and the lane line,such as the isolation belt,railing and building on the road.(3)Vehicle detection and positioning.In the vehicle detection and positioning process,the region of interest of the vehicle is first extracted by means of artificial delineation;then the vehicle is initially positioned based on the partial model,and the vehicle is divided into upper and lower parts as candidate parts,and multiple types of features are used.The hybrid image template of the descriptor is modeled,and then the candidate part is identified by template matching to detect the vehicle.Finally,the vehicle is accurately positioned by using the multi-feature vehicle detection algorithm suitable for the vehicle image,which has high precision and complexity.Low advantage.Through the preliminary positioning and precise positioning methods,the expected effect of vehicle detection and positioning is achieved.(4)Firstly,the geometric space coordinate transformation is used to derive the conversion relationship between the image and the world coordinate system,and then the small hole imaging model is established.Because the format and size of China's license plate are more uniform,this paper adopts a monocular vision vehicle ranging method based on license plate feature information.The method first locates the license plate and detects its four corners,and then performs distance estimation based on the rectangular measurement theory.The camera's internal parameters are obtained using the current popular checkerboard technique,and the distance between the world coordinate system and the camera coordinate system is calculated using the P4 P algorithm,thereby calculating the distance between the two vehicles.The static vehicle ranging experiment of simulating unmanned vehicles on campus is carried out.After the obtained vehicle image is processed by algorithm,the experimental results show that the relative error of the proposed method is within the allowable range,and the accuracy of visual ranging is verified.