Research On Pedestrian Visual Recognition In Driverless Scene Of Forklift

Forklift is an important logistics handling equipment.With the increasing demand for intelligence in the warehousing and logistics industry,logistics equipment is also gradually developing towards intelligence and unmanned.Visual recognition is one of the most critical technologies for realizing the unmanned driving of forklift.It is of great significance to ensure the safety of pedestrians in the working environment of handling equipment.Therefore,the visual recognition of forklift driverless scene is studied by using deep learning technology.The research content mainly includes the following three parts:Firstly,in order to accurately find the central coordinates of pedestrians and their areas from pedestrian images,a two-stage progressive pedestrian area center calculation method is proposed.In the first stage of the algorithm,the Ghost Net improved YOLOv4 algorithm is used to detect pedestrian targets in the image.At the same time,in order to reduce the influence of bad environment on detection accuracy,image enhancement is carried out in the image preprocessing stage to improve the adaptability of the algorithm to bad environment.The experimental results show that the YOLOv4 algorithm improved by Ghost Net has a great improvement in real-time performance,and can ensure better detection accuracy.In the second stage of the algorithm,the rectangular pedestrian area obtained in the first stage is cut,and then the center of the pedestrian area is solved by U-NET algorithm and thresholding segmentation.The experimental results show that the accuracy of the pedestrian area center calculation algorithm in this paper is much higher than that of other algorithm models.Secondly,in order to calculate the relative position relationship between pedestrians and forklifts and determine the safety of pedestrians.According to the driving characteristics of forklift,the coordinate system in the driving process of forklift is established,and the safety area of the coordinate system is divided according to the safety requirements of forklift.At the same time,the binocular camera is calibrated.According to the installation position and calibration parameters of the camera,the coordinate system conversion formula is obtained,and the conversion from pixel coordinate system to forklift driving coordinate system is realized.Then,the transformation relationship from photographing images,detecting pedestrians,obtaining pedestrian area center coordinates to determining their actual position and safety is established.Finally,in order to further ensure the pedestrian safety in the forklift driverless scene,this paper introduces the Social-STGCNN trajectory prediction algorithm into the forklift driverless vision system,and predicts its future trajectory through the pedestrian historical trajectory obtained from the top view.According to the demand and characteristics of downward trajectory prediction in forklift driverless scene,a new evaluation standard is proposed.The experimental results show that the method used in this paper is better than other trajectory prediction algorithms in various evaluation indexes.