Research Of Monocular Visual Pose Measurement Technology Under Complex Light Environment

Monocular visual pose measurement technology plays an important role in the fields of robotics,industrial manufacturing,transportation,aerospace,life and entertainment,etc.due to its high precision,high speed,simple structure,low cost and strong anti-interference ability.However,complex light conditions,such as uneven light,low light,shadow,local reflection,large variation of light color and temperature,restrict the effective implementation of the vision pose measurement technology in some application scenarios(tunnel,ball mill,spacecraft,rain and snow weather,etc).Based on this background,this paper presents a complete set of pose measurement technology based on monocular vision,which can effectively measure the object pose under complex light conditions.Firstly,the cooperative target is designed.The simulation experiments are carried out.Optimal target parameters(number of feature points and whether feature points are coplanar or not)are determined through a large number of experimental data.A reasonable target size is determined by calculation of view field.A color based feature point coding scheme is proposed to effectively match feature points.Finally a cooperative target is designed which is easy to make,recognize and decode.Then,for complex light environment,image enhancement,smoothing filtering and color correction algorithms are studied.Gauss homomorphic filtering can uniform image light.Median filtering can effectively remove image noise under low light.An improved specular reflection method is proposed,which can effectively correct image color under conditions of more or less color and different color temperature light.Next,recognize target feature points.Combining Gauss homomorphic filtering with Otsu algorithm,the target region can be effectively segmented in the environment of uneven light,low light,light shadow and local reflection.Recognize the target boundary line by Hough linear transformation to determine target area.Recognize target feature points based on round-degree.Improve the expression method of connected domains to solve the abnormal calculation of round-degree.Define ellipse-degree and recognize target feature points based on ellipse-degree.Aiming at the problem that the target strong shadow is difficult to be solved by Otsu algorithm combined with Gauss homomorphic filtering,a target shadow removal algorithm based on ellipse-degree is proposed.A color distance is defined in RGB color space,and the feature points are decoded by color recognition.Finally,calculate the position and posture.Calibrate the camera parameters.According to the recognized pixel coordinates of the target feature points,EPnP algorithm is used to calculate the position and pose.Simulation experiments and real experiments are carried out to verify the method.The experimental results show that the method has high accuracy and can meet the actual needs.