It is an inevitable trend to detect and recognize the states of different component on distribution cabinet panel more effectively and accurately by using inspection robots instead of manpower. Aiming at the problems of multiple recognition targets and large size difference in image recognition of distribution cabinet panel, an improved Faster R-CNN multi-target detection method is designed. Automatically detect and recognize of different targets on required cabinet panel is achieved by this method. In this paper, Resnet50 is used instead of VGG16 as the feature extraction network in Faster R-CNN, Adam Optimizer is used instead of Momentum, and anchor box size is changed to adapt target size difference on cabinet panel. Experiments show that the improved model has higher accuracy and less computational cost on the required distribution cabinet panel dataset.
Due to the limitation of sensor application in special environment such as relatively closed and magnetic interference environment, the positioning and the heading angle errors of wall climbing robots accumulate with time. This paper proposes a difference projection localization method based on an external RGB-D camera and a robot-carried inertial measurement unit (IMU). We differential the depth image to obtain the distance change due to the robot occupancy. Then, the 3D point cloud information is converted into 2D image information by projecting the above distances along the normal vector of the robot chassis, which greatly speeds up the computational speed. The position of the robot is calculated by studying the statistical characteristics of the projection. Two EKFs are designed to estimate the attitude, taking the gravity vector and the normal vector of the robot chassis as observation. The experimental results show that the localization error of the wall climbing robot is within 0.017m, and the heading angle error of the attitude estimation is within 3.1∘. The obtained results prove its applicability in self-localization of the wall climbing robot.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.