Intelligent Damage Detection Method For Glass Curtain Wall Panel Based On UAV Image Acquisition

Glass curtain wall has rapidly developed into the main external maintenance form of modern high-rise buildings with its light weight and good decorative effect.However,curtain wall glass belongs to brittle material,which is prone to brittle failure,leading to safety accidents.With the aging of early glass curtain wall becoming more and more serious,curtain wall safety has become a problem that has to be paid attention to.In recent years,serious accidents of curtain wall glass falling off from high altitude have become more and more frequent.Three curtain wall panel falling off events occurred in Shanghai in one month in 2016.Curtain wall safety management has become an urgent problem in modern urban safety management.For a long time,the curtain wall panel damage detection mainly relies on manual inspection through hanging basket,which is inefficient,high cost.And complex high-rise building environment,which greatly interferes with the results of manual inspection,and it is difficult to ensure the safety of inspectors.Therefore,it is necessary to explore an intelligent and automated method for curtain wall panel damage detection to reduce the fall accident of curtain wall panel caused by failure to detect the damage of curtain wall panel in time.For this reason,a new damage detection method for curtain wall panel is proposed based on aerial photo UAV and image processing technology in this paper.Firstly,based on the characteristics of glass curtain wall panels,the image of glass curtain wall is captured by a combination of Four-rotor aerial photography UAV face up and slant photography.Then,image processing technology is used to extract the curtain wall panel image,enhance,extract,measure and identify the curtain wall panel damage.The following work has been accomplished:1)This paper summarizes and analyses the principle of three types of damage of curtain wall glass,including self-explosion,thermal cracking and external impact,as well as the characteristics of damage shape,location and direction.A method of acquisition of curtain wall images with elevation as the main part and slope as the auxiliary part is proposed by studying the reflection,gray distribution and gray characteristics of damaged parts in different glass curtain wall images,which effectively solves the interference problem of high reflectivity glass curtain wall reflection on panel damage detection.2)This paper determines the processing idea of firstly extracting the image of curtain wall panel,and then detecting the damage of curtain wall panel based on the characteristics and damage characteristics of curtain wall panel.(1)Firstly,the illumination of curtain wall image is balanced by improved Mask homogenization method,and then the perspective distortion of curtain wall panel image is extracted by combining Hough linear transformation and perspective transformation.And the detection effect of curtain wall panel boundary is improved by improving Hough straight line algorithm.(2)This paper proposes ZHJC curtain wall panel damage image enhancement method through a large number of tests and algorithm improvements on the basis of image extraction of curtain wall panel.This method can reduce the image data of curtain wall panel and enhance the contrast of curtain wall panel damage on the basis of retaining the damage information of curtain wall panel,which can provide a clear image of curtain wall panel for the extraction and identification of curtain wall panel damage.3)In the aspect of curtain wall panel damage edge extraction,this paper compares and analyses the experimental results of curtain wall panel by using classical edge detection algorithm,and finds that Canny operator has the best effect in the extraction of curtain wall panel damage.Aiming at the problem that Canny operator has not strong continuity in the extraction of curtain wall panel damage edge,this paper proposes Kmeans-Canny curtain wall panel extraction method of clustering the curtain wall panel first,and then edge detection.Then,after filling the damage edge of the curtain wall panel,the area of the curtain wall panel damage is measured based on the object-image relationship.Finally,support vector machine(SVM)is used to filter the pseudo-damage of curtain wall panel by using three indicators: height-width ratio of connected area,target pixel ratio and circumference.The results show that the damage of curtain wall panel can be correctly identified as the damage of curtain wall panel on the basis of correct extraction.4)In order to better realize the extraction and identification of curtain wall panel damage,based on the above image acquisition methods and processing algorithms,a glass curtain wall panel damage detection system based on UAV and image processing technology is constructed.Aiming at the problems encountered in the testing process,the effects of UAV selection,illumination environment and curtain wall surrounding environment on the method are discussed and analyzed.The countermeasures and suggestions are put forward,and the future improvement direction is defined.