Research On Image Mosaic And Crack Extraction Method And Their Applications In Bridge Inspection Based On Multi-legged Robot

At present, with the rapid development of bridge construction, bridge health monitoring and maintenance have become very important. Bridge Crack is an important indicator of the health monitoring. Nowadays, the bridge crack detection relies on manual method which has low detection efficiency and security risk. Therefore, this paper researches on the method of identifying crack using image processing technique.Due to the complex environment and light condition under bridges, uneven illumination and noise affect image capture seriously. To solve these problems, the paper designs a bridge crack detection scheme, the main contents are as follows:(1) The crack image mosaic method: In order to ensure the accuracy of crack image, some long cracks can’t be taken with one image, this paper studies on the image mosaic. For the requirements of real time and high precision, this paper presents an image mosaic method based on improved FAST algorithm. The method can be divided into four steps: The first step is to improve the FAST algorithm to extract the scale invariant feature points; The second step is to improve the accuracy of feature matching through combining the distance ratio method with the cosine similarity method; The third step is to do an exact match using RANSAC algorithm; The last step is using the weighted average method to fuse image. Experimental results shows that the crack image mosaic method which has high matching accuracy and robustness can stitch the crack images quickly.(2) The Bridge Crack Extraction technology: Based on the characteristics of crack image, this paper presents an optimal crack extraction method from three sections: image preprocessing, image segmentation and calculation of crack geometry. The first section presents an adaptive multi-structure median filtering algorithm to preserve image details while denoising; Then, as crack detection requires real-time processing and high quality of segmented image, this paper presents an algorithm of two-dimensional cross entropy using bee colony optimization and crack restoration method based on morphological, Experimental results shows that the algorithm has short processing time, stable performance, robustness and good segmentation; In the last section, a method of extracting skeleton is presented to calculate single crack length. And according to the type of crack, this paper put forward a new method to calculate the max width of cracks. Experimental results show these methods have high precision and meet the requirement of bridge health monitoring.(3) Bridge crack detection system based on multi-legged robot: This system consists of image acquisition and image processing. Image processing module is mainly a crack detection software based on MATLAB. The software integrates many crack detection algorithms, so crack detection became more intuitive and practical.