Key Technology Research And System Implementation Of Bridge Crack Detection

With the development of China's transportation industry,the number of bridges has grown explosively,and the demand for disease detection of bridge facilities has increased.As an early manifestation of various serious bridge diseases,cracks have been the focus of conservation departments.In order to ensure traffic safety,the maintenance department needs to periodically perform crack detection on the bridge facilities through manual inspection.However,manual detection methods have significant defects such as high human and material costs,high subjectivity,and low reliability.The research and implementation of automated crack detection methods based on image processing have theoretical research value and practical application value.This paper studies and implements the key technologies involved in image crack detection.The main innovative work includes the following aspects:First,because the image collection of cracks is mostly performed outdoors,due to environmental factors in the shooting environment,there are often low signal-to-noise ratios,blurring,and other degraded phenomena,resulting in a decrease in the detection rate.In this paper,the dark channel defogging algorithm is applied to image preprocessing to improve the image quality.Experiments show that after the image enhancement preprocessing,the recall rate of the detection algorithm in this paper can be increased by 13%.In view of the processing speed of the dark channel defogging algorithm,the GPU parallel acceleration method of this algorithm is deeply studied.For a large number of operations and a timeconsuming minimum filter and a mean filter,this paper proposes a block-parallel computing method that makes full use of the GPU's multi-level storage structure and greatly improves the processing speed.Experimental results show that the GPU parallel algorithm proposed in this paper can achieve a 30-times faster speedup compared with the CPU serial algorithm,and the color image processing time for 1920*1080 resolution can be shortened to 26 milliseconds.The processing speed is better than that of the existing algorithms.Secondly,aiming at the long time-consuming problem of the tensor voting module in the existing crack detection algorithms,this paper proposes an angular quantized tensor voting algorithm.By calculating the discrete angle tensor field and generating the look-up table,the algorithm computation amount is reduced.,processing speed increased by more than 3 times.After quantification,the precision,recall,and F value of the detection algorithm did not change significantly,indicating that the quantization algorithm proposed in this paper can improve the operating efficiency of the algorithm without changing the detection accuracy.Third,this paper proposes a crack image detection method based on multi-scale detection results fusion.By fusing the detection results of the measured images at different scales,accurate detection of cracks with different widths can be achieved.The experimental results show that compared with the existing methods,the detection rate of this paper has a certain improvement.