Recognition Of AGV Vehicle Auxiliary Features Based On Monocular Vision In Complex Environment

With the arrival of industry 4.0,various industries are constantly developing in the direction of intelligence.As one of the key technologies in industrial production,industrial robots have become an important part of enterprises' industrial optimization and upgrading.As a kind of robot,automatic guided vehicle has become more and more widely used with the continuous development of various guidance technologies in recent years.Among them,visual guidance has become a research hotspot in automatic guidance technologies due to its huge application potential.In this paper,an auxiliary identification system to monocular vision AGV car is proposed according to the actual application of AGV car driving requirements in indoor scenes.The system consists of scene recognition,assisted driving in the scene,and feature detection.In view of the complexity of the indoor environment,this paper divides the indoor environment into three main scenes,including stairs,corridors and turning platforms.In view of the scene recognition,a staircase scene recognition algorithm based on Gabor filter and a corridor scene recognition algorithm based on feature line detection are proposed.Aiming at the assisted driving in scenes,in the scene of stairs and corridors,an auxiliary driving scheme is proposed that obtains the boundary line of the scene area and calculates the centerline driving line,and calculates the car driving state according to the relative position of the center driving line in the image.In this paper,edge detection combined with Hough transform is used to extract the baseboard features of the corridor and the side boundary characteristics of the stairs,and corner detection combined with line fitting is used to extract the side boundary characteristics of the stairs.In the scene of turning platform,an assistant driving scheme based on the key time points of turning is proposed.Aiming at the feature detection during driving,the two-dimensional codes arranged on the wall and the potential pedestrians are regarded as features.The twodimensional code features are recognized to obtain the car turning information.HOG feature extraction and SVM classification learning method are used to train and get a pedestrian classifier which realizes the detection of pedestrian features in the scene.