Pose Estimation Of The Loading Load Under Unmanned Aerial Vehicle Based On Machine Vision

In the field of computer vision,the pose estimation of moving objects is an advanced research field which is popular among experts and scholars.It involves many fields such as target detection and tracking,digital image processing,optimal estimation and so on.In this paper,based on the background of the lift load of the unmanned aerial vehicle(uav),the pose estimation system is constructed from the detection and tracking of the moving load,and the feasibility of the designed system is analyzed.In order to obtain the pose information of the lifting load,two approaches are proposed: one is the target detection algorithm based on Faster r-cnn.The method is used to detect the lifting load and judge the loading posture.The specific steps are to manually calibrate the small sample target intercepted in advance,and then to realize the detection of the target using the Faster R-CNN algorithm as the framework and the ResNet-50 convolutional neural network as the main body.The other is the target tracking algorithm based on mean-shift.The method is used to track the lifting load and judge the loading posture.The specific step is to describe the target feature of the original image frame,and then achieve mean-shift iteration until convergence to the target position.Finally,the method of pose estimation based on imaging model is used to estimate the spatial pose of lifting load.By using the target information obtained by the two methods to verify the target pose estimation model,the space map conforming to the spherical characteristics is finally obtained.Experimental results show that the system can detect and track the target effectively.The pose estimation of the target is also more feasible than that of the fixed point(camera).