Comparison of Effectiveness of Dual Tree Complex Wavelet Transform and Anisotropic Diffusion in MCA for Concrete Crack Detection

Abstract

Detection of micro cracks present on the concrete surface of a bridge deck is an important task to maintain the quality, reliability and the health of the bridge. In manual inspection humans inspect the sites, prepare the sketches of cracks manually and note the conditions of irregularities. But this process is very long and does not give the guarantee for accuracy. To overcome the demerits of the manual inspection, in our work micro crack detection in parts of concrete bridge is done by using a drone with the high-resolution proximity camera. However, due to micro vibrations present on the bridge it is very difficult to take the fine pictures of some parts of the bridge and due to intensity and contrast instability, the mm width measurement of concrete cracks is also very difficult. It requires the study of texture features of the images such as variance of gray scale distribution, anisotropic structure, directionality and orientations of local textures etc. To study the texture features of the images of bridge deck we have used Morphological Component Analysis (MCA) based on sparse coding. Every image is decomposed into its coarse and fine components. Dual Tree Complex Wavelet transform (DTCWT) and Anisotropic Diffusion (AD) have been used in MCA to find the coarse components of the images acquired by camera. The coarse component is subtracted from the original image and the subtracted component is treated by the Sobel edge detector to demonstrate the cracks in a fine way. Results are compared to determine the best dictionary to find coarse component and the cracks present in the image. The compared results illustrate the quality of selection of dictionary to detect the cracks.