Research On Key Technology Of Intelligent Glue Spraying Robot For Coal Freight Train Based On Machine Vision

With the increase in the shipping volume of coal freight trains and the decrease in the number of people who choose to engage in the traditional coal industry,the traditional manual operation of coal freight trains spraying glue has gradually been unable to meet the demand.The use of advanced machine vision technology to improve the automation and intelligence level of the glue spraying operation of freight trains has become an urgent issue to be solved in Chinas coal railway freight.According to the working process of coal freight trains and the technical requirements for glue spraying,and after field investigations,the overall technical scheme of intelligent glue spraying robots for coal freight trains based on machine vision was proposed,the main technical indicators of the robot were determined,and the key technical problems faced were summarized.Propose corresponding solutions and optimization strategies.Aiming at the problem of machine vision positioning accuracy,a kinematic model of the main body of the intelligent glue-spraying robot was established,and the accuracy of the model was verified.At the same time,the camera is calibrated and the camera model is obtained.By studying four traditional analytical algorithms for robot Hand-Eye calibration,a Camera-Pose generation scheme for reducing Hand-Eye calibration errors is proposed.At the same time,a camera-oriented Hand-Eye calibration method is proposed.The DLH-GWO-DE optimization algorithm further reduces the Hand-Eye calibration error.For the problem of the glue-spraying robots visual target positioning and detection,a set of image positioning algorithms for glue-spraying seams is designed.The written program can handle various conditions such as day or night,sufficient or severe shedding of the original sealant,and the algorithm is more adaptive.Preliminary experiments are carried out on the effect of the HOG-based SVM algorithm and the Faster R-CNN-based deep learning algorithm on the detection of key structural targets in the spraying operation of coal freight trains.In order to improve the speed of the sliding window detection algorithm,a method is proposed.An improved algorithm based on CPU multi-threaded parallel computing.Finally,for the six key structures that need to be effectively avoided when spraying glue,use the Faster R-CNN-based deep learning algorithm for target detection,and the detection effect is good.In response to the technical requirements of glue spraying,a visual configuration scheme combining the Eye-to-Hand system and the Eye-in-Hand system was proposed,and a virtual prototype based on position-based visual servo control was built.Subsequently,the robot motion control algorithm combined with feedforward control and PD feedback control was used to simulate the trajectory of glue spraying seam,which proved the effectiveness of the system.