Research On Path Recognition Strategy Of Crane Climbing Robot

In our country,many large cranes have been out of service or in the late stage of service.There are huge safety risks when they are used.Inspection and analysis,safety assessment,life prediction and maintenance of in-service cranes can better prevent safety accidents.Crane climbing robots can replace human climbing and assist staff to carry out the above tasks through various equipment carried by themselves.Using image recognition technology to analyze and identify path is an important part of the daily work of climbing robots and it is also a prerequisite for its autonomous navigation and various tasks.This article focuses on the path identification strategy of crane climbing robot,explores the application prospect of image recognition technology in the field of construction machinery.The specific work content is as follows:(1)Establishing path edge recognition strategy and path images pre-processing.By analyzing the working environment of the crane climbing robot and metal structure path,an edge recognition strategy based on the crane metal structure path is designed.By digitizing the path image,the path image mathematical model is established.During image preprocessing,using the improved "over-color operator" to grayscale the path image.The noise in the gray image is removed by median filtering and the small dark areas in the gray image are eliminated by the gray closure operation.The gray image that is advantageous for path edge recognition is finally obtained.(2)Path image analysis.Using the improved Line Segment Detector(LSD)algorithm to detect line segments in the path grayscale image,obtain the line segment detection image.In order to reduce the number of interference straight line segments detected by the LSD algorithm,three constraints were added to the LSD algorithm: by using the optimal classification line of the support vector machine to determine the gradient threshold,add gradient threshold constraints;by constraining the principal direction angle of the smallest circumscribed rectangle of the line support region,add the line segment tilt angle constraint;by constraining the endpoint position of the line segment,add the line segment plane position constraint.By adding the above three constraints,the number of interference straight line segments is significantly reduced and the LSD algorithm is improved.(3)Path edge recognition.Feature extraction is used for all detected line segments and build the clustering data set.Based on the characteristics of dynamic change in the clustering data set,the path edge line segments with similar features are grouped into one class by adopting the Affinity Propagation(AP)clustering algorithm.The AP clustering is improved and the clustering accuracy of the path edge line segments is improved by adding the discriminant model based on prior information.Through the endpoints of the path edge line segments,the final path edge line is obtained by fitting.(4)Obtaining the images of crane metal structure path through field shooting,carrying out the experimental verification and the algorithm simulation of path edge recognition strategy through matlab language and Matlab R2017 b platform.By listing a large number of test examples,the reliability of the experimental and simulation results is improved and the effectiveness of the path edge recognition strategy in this paper is finally verified.