Research On Bridge Crack Detection Algorithm Based On Geodesic Distance Transform

With the rapid economic development of our country,the transportation infrastructure has been increasingly perfected,and the number and total mileage of bridges are continuously increasing.Cracks are an early manifestation of various bridge diseases and are important indicators for measuring the safety of bridges.Therefore,the effective detection of bridge cracks is of great significance to ensure the safety of the bridge.The bridge crack detection technology based on digital image processing can automatically obtain bridge crack information and overcome the shortcomings of low accuracy,low efficiency and high cost and being dangerous in manual detection methods,which has important practical application value.How to detect real cracks from a bridge crack image with noise interference is the key and difficult point of using digital image processing technology to detect cracks in bridges,which has attracted more and more attention of scholars.Based on the analysis of the characteristics of bridge cracks and background noise,the thesis proposes a bridge crack detection algorithm based on geodesic distance transform which includes the steps of seed extraction,crack enhancement and crack extraction.The time complexity of algorithm is lower and the accuracy and the recall rates of crack detection is higher.The main work of the thesis includes:(1)Aiming at the problem that there are many false seed points in seed candidate points extracted by existing sliding window algorithm and the problem of high miss rate of seed points,a seed point extraction algorithm based on symbiotic edges is proposed which restricts the selection of pseudo seed points from the aspects of gradient size,direction,distance of symbiotic points and gray information.Experiments show that compared with the sliding window algorithm,the seed recall rate of the proposed algorithm is improved by 4.7% and precision rate is increased by 50.0%.(2)Aiming at the problem of disturbance from pseudo seed point to seed point generating crack profile in seed point binary image,the thesis combines the position information of seed point and false seed point and gray change information and proposes a bridge crack enhancement algorithm based on geodesic distance transform which realize the enhancement of crack in combination with graph theory after using the geodesic distance transform in order to quickly implement the algorithm to determine the adjacency relationship of the seed candidate points and geodesic distance.Experiments show that the recall rate of reinforced cracks has reached 59.04%,and the precision rate has reached 72.67%.Compared with before enhancement,the cracks have been significantly improved,and the detection of crack profiles has basically been achieved.(3)Aiming at the difficulty to distinguishing between crack target point segments and a few non-crack target point segments in the enhanced crack binary image by length and morphology,the thesis use the minimum spanning tree to connect all the target points in the binary image into a minimum spanning tree and then realize the crack extraction by cutting the tree and extracting longest path.Experiments show that compared with the existing two crack detection algorithms,the crack detection algorithm presented in the thesis has higher accuracy,recall rate and processing speed,and the extracted cracks are more complete and natural.