Research On Road Crack Visual Detection Method Based On UAV Cruise

In recent years,with the development and progress of unmanned aerial vehicle(UAV)technology,UAV has been widely used in such fields as electric power inspection,pesticide spraying,logistics distribution and urban illegal building monitoring.The use of UAV has increased the level of automation in these fields and reduced the labor intensity of workers in these fields.In terms of road maintenance and maintenance,researchers are just beginning to study the use of UAV.At present,road disease detection still mainly relies on road detection vehicle.The automation level,energy consumption and labor intensity of workers of road detection vehicles need to be further improved.The combination of UAV technology and pavement disease detection technology can make full use of UAV ’s advantages of high degree of automation,low energy consumption and significantly reduced labor intensity.Therefore,the application of UAV in road detection research will have great significance and far-reaching impact on highway maintenance and monitoring.In this paper,UAV is used as a platform to study the detection of expressway road crack disease.Combined with the research background,the algorithm of pavement crack detection is further explored on the basis of the existing technology,and the characteristics of the UAV image are analyzed and considered.Aiming at the aerial image of UAV,a set of road pavement crack disease recognition system with three algorithms is proposed,which realizes the process from the original image to the accurate recognition of cracks.The main work of this paper is as follows:(1)This paper proposes a pavement surface image recognition algorithm based on UAV detection platform,and designs a pavement surface recognition algorithm process of UAV by combining the existing digital image processing technology with the characteristics of UAV aerial images.The process first uses the Hough linear transformation to detect the road boundary line,then traverses the entire image in the pixel coordinate system,removes complex backgrounds,and implements road surface segmentation.Finally,aiming at the shortcomings of the inaccurate recognition of Hough’s linear transformation method in complex scenes such as road bending and bifurcation of aerial images,the Lab color model is used to analyze and identify the road surface image,thereby improving the application range and level of the UAV road surface recognition algorithm.(2)This paper proposes a road crack discrimination algorithm based on neural network,which is mainly used to judge whether the road surface identified in the previous step crack diseases,so as to select the images containing cracks from the massive image data.First of all,establish a database contains no cracks and crack of 10000 images,used for neural network training when the training set,and then design a neural network structure model according to actual demand,finally choosing appropriate training function,using the error back propagation algorithm for neural network learning,so as to complete the whole neural network fracture judge training mission.(3)This paper proposes a pavement crack recognition algorithm based on uav detection platform,considering that crack diseases are less in the image and are not easy to be detected and recognized.On the basis of the above operation results,splitting the image matrix into subblocks for local detection is more conducive to improving the detection accuracy.When detecting a sub-block image,first,I used the maximum inter-class variance method to obtain a binary map of the image of the cracked sub-blocks,and then used the connected domain labeling method to filter the binary map through the constraints of area,aspect ratio and other constraints to achieve the sub-block crack extraction.Finally,the sub-block images are physically stitched,and then the morphological closed operation is used to connect the cracks and fractures to detect and identify the cracks in the aerial image.