The Study Of Crack Detection Based On Computer Vision

Crack detection based on computer vision is always the hot topics and difficult problems in the field of engineering quality evaluation.In the past for a long time,many researches have been done to try to solve this problem.It's complexed to detect cracks in different environments,so many challenges will be faced in gap detection.Especially in the handling of the noise and the precise detection of the tiny fissures which need to be further improved.The object of this paper is to complete the measurement of the gap parameters to provide data support for the detailed analysis and evaluation of the quality of the engineering.In view of the structural features of the gap,a measurement and recognition method based on machine vision is adopted.The gap parameters are measured by using the image of the gap as the samples.After processing a series of image processing for CCD photo,including noise reduction,edge location and binaryzation,the edges are positioned to single pixel,and the initial measurement of the gap is completed.According to the characteristics of gap image noise,after image processing in accordance with the experimental steps,method of non-local mean filter for noise reduction using the improved,better eliminate the effect of noise,has a significant effect to retain fine structure,details and textures,to achieve a better edge location.Aiming at the limitation of existing gap detection,a gap connection method based on gravitational model in physics is adopted in view of the phenomenon of virtual breaking in the actual gap recognition,which fully takes into account the directivity of the slot figure itself.The algorithm takes full account of the extension and local transitivity of gaps,and searches for the starting point of gaps,so that the determination of whether the gap is connected is limited to the local category of the direction where the starting point extends.The robustness of gap junctions is enhanced under the condition of noise and multi spots.Considering the gray interval,a method of multiple combination of gap extension and combination is developed.The experimental results show its superiority.The classical sub-pixel based edge detection algorithms are analyzed.Trigonometric fitting method and curve fitting method are used to carry out experiments.The advantages and disadvantages of the two algorithms in recognition accuracy and noise immunity are compared.And according to the actual gap difference between the actual images,the point spread function and point spread function of the theory innovation of the discussion for acquisition in different experiment data to analyze the causes and effects of point spread,in full consideration of the point spread function,measurement of the compensation gap and achieve better results.At the same time,the gray compensation formula of the actual CCD image is discussed,and the subpixel detection based on gray level is improved,and good results have been achieved.Finally,the performance of the algorithm is analyzed,and the recall and accuracy of the experimental results are tested,which proves the superiority and feasibility of the proposed algorithm.In this paper,the recognition and measurement of cracks is realized by machine vision.At the same time,the accuracy and accuracy of measurement based on sub-pixel edge detection and recognition algorithm are improved.