Research And Implementation Of Vision Ranging Technology For Automatic Guided Vehicle

The automatic guided vehicle uses sensing equipment such as electromagnetic induction coil,laser radar,industrial camera and supporting automatic guidance system to realize unmanned automatic guidance and operation according to preset or independently planned driving path.The automatic guidance system needs to solve the problems of environmental information perception,spatial positioning and navigation,distance measurement during vehicle operation.At present,the automatic guidance system based on laser radar or stereo binocular camera has the defects of high construction cost and low information perception.The automatic guidance system based on computer vision has the advantages of low cost,strong environmental adaptability and high scalability.Aiming at the positioning and navigation task of automatic guided vehicle in complex working environment,a visual ranging system based on computer vision and spatial geometric relationship is studied and implemented.It provides real-time visual information for automatic guided vehicle by using object detection and object tracking technology,and measures the linear distance between automatic guided vehicle and destination or obstacle by using spatial ranging technology.The accuracy and time performance of the object detection and tracking algorithms on the automatic guided vehicle platform are optimized through the lightweight modification of the backbone network,scene clustering optimization and balanced video frame filtering mechanism.The field of view angle and ground hypothesis are introduced to simplify the visual ranging method and expand the universality of the visual ranging method in a variety of terrain scenes.The vision ranging system of automatic guided vehicle based on computer vision and spatial geometric relationship can sense and track the target load and obstacles in the scene in real time.The ranging error within 5 m of the simulated logistics storage site is less than0.2 m.Compared with the scheme using depth estimation model,the system can reduce the calculation time cost by more than 95%,and compared with the scheme using laser radar,the system realizes real-time tracking and ranging of moving objects.