Research On Visual Tracking System Of Truck's Loading Port Based On Artificial Intelligence

With the progress of industrial technology and the development of economy,higher requirements are brought for the efficient operation ability of powder material truck under the requirement of high efficiency.When loading powder materials,it is necessary to ensure that the feeding port of the truck and the hopper are fully aligned,so as to achieve efficient loading.At present,the method of manual control of aligning was adopted on construction site,which has problems such as complicated operation,low docking efficiency,poor reliability and low intellectualization,and cannot satisfied the requirements of highly efficient loading.How to realize the detection and tracking of the feeding port in the process of loading has become the urgent task in related industry.In order to solve the problems existing in traditional truck loading,a visual tracking system of truck's loading port based on artificial intelligence was proposed.The overall scheme of the system was designed by analyzing the system requirements.The embedded hardware platform was built based on ARM,which took image processing module as the core and contained image acquisition module,storage module,communication module and display module,In which the image processing module was designed by the use of quad-core processors H5 with Cortex-A7 CPU architecture.CMOS camera module GC2035 with CSI interface was adopted to complete the design of image acquisition module.EMMC is used to store and read system images.CH340 E and RTL8211 E were used to design the serial port to USB circuit and Ethernet communication circuit to realize the design of the communication module.A separate HD monitor was attached via HDMI as display module.The embedded Linux operating system was built by compiling U-Boot and kernel and making root file system.On this basis,the system software platform was built by porting the TensorFlow deep learning framework.By analyzing and comparing the convolutional neural network algorithm RCNN,Fast-RCNN,Faster-RCNN based on candidate regions and the convolutional neural network algorithm YOLO and SSD based on image regression method,the SSD tracking algorithm of truck's loading port was proposed.The SSD neural network was trained based on the transfer learning,and the deep learning model was established to realize the tracking of truck's loading port.The feasibility and practicability of the proposed system were verified by the tracking experiment.The experimental result shows that the design of visual tracking system of truck's loading port based on artificial intelligence has good robustness and practicality.It can output the coordinate position of the truck's loading port accurately in both ordinary and complex environments,and realize the stable tracking of the truck's port,which can fully meet the technical requirements of the detection and tracking of the truck's loading port during the process of loading,and provide the basis for the subsequent motion control.