Research On Automatic Detection And Recognition Algorithm For Road Crack In Complex Scene

With the great attention from society, the automatic pavement crackdetection technology based on image processing on expressway has beendeveloping rapidly in recent years. The crack detection work will encounterbad road condition frequently in actual application, like broken road sign, oilstain, shadow of obstacle, high frequency noise and other factors. In theseconditions, the normal crack extraction algorithm can not meet the demandsof detection. Many important locations are missing because the useful data isnot available. It limits the practical application of automatic crack detection.Aimed at solving these problems, this paper mainly describes the three keytechnologies, such as image enhancement, crack information extraction andcalculation of crack parameters.Firstly, the grey level correction based on homomorphic logarithmmodule and adaptive direction filter are used for the image enhancementprocessing. The algorithm of grey level correction based on homomorphiclogarithm module can solve the nonuniform illumination problem, andintensify the image detail. Through calculating the direction of grey level incurrent area, the adaptive direction filter algorithm using corresponding filtertemplate to restrain background noise and protect the crack information. Theprocessing of these two steps has created a stable environment for thefollowing crack extraction algorithm.Secondly, the crack extraction algorithm which has merged multiplefeatures intelligently is proposed. The algorithm based on high-frequencyfeature is used to distinguish the healthy image from crack image. Then, basiccrack information is extract according to the method based on non-negativefeature and contrast feature. Circular projection based on linearity feature isapplied to enhance the crack region of fracture, combined with connectivity region denoise, the false crack information is eliminated.Finally, morphology conversion and chain code tracing algorithm areused to calculate crack parameters. The skeleton information of crack isobtained by morphology dilation and morphology thinning. Then we calculatethe length, length-width ratio, area, number and other parameters. Classifyingthe crack and judging the damage level.A series experiments have been performed to test the efficiency andadaptability of proposed algorithm. In the process of experimental, everysteps of algorithm have been verified by the experiments in the first place.Next, the extraction experiments under variety complex scenes have beentaken place. And the proposed crack extraction algorithm was compared withadvanced algorithms in extraction results and testing time. It is proven that theproposed algorithm can obtain useful crack information in all kinds ofcomplex scenes accurately. It has overcome the disadvantages of presentextraction algorithm, and has a well stability and sensitivity. The testing timeis reasonable and satisfied with the demand of detection cycle.