Research On Automatic Recognition Of Image Line Features And Its Application In Road Crack Detection

Due to traffic pressure,weather and human factors,the occurrence of road disasters has become more frequent,which has seriously affected the service life and driving safety of roads.Especially on highways,the adverse effects caused by road disasters will be more pronounced.Cracks,as one of the typical road disasters,if can be discovered and properly reported and handled intime,it can not only greatly reduce the risk of driving,but also greatly save the cost of road maintenance.Therefore,the automatic crack detection method is extremely urgent and important.The existing methods and systems have not yet achieved full-automatic detection of road cracks.And at present,scholars have less research on the detection of cracks based on the influence of complex noise.In the actual situation,however,the crack detection needs to take into account many factors such as shadows,stains,light and uneven gray scale,etc.Therefore,a method that can be applied to the actual situation and better realize the automatic crack detection becomes extremely important.In this paper,the image line feature recognition method combined with common crack detection methods are proposed and improved.Three kinds of crack detection algorithms for different road conditions are proposed:For the road surface image with less road noise and better lighting conditions,this paper proposes an automatic detection method based on adaptive moving average threshold.This method proposes a moving average adaptive threshold segmentation method for coarse image segmentation;Then,setting the rule of the connected domain of the linear structure to remove the tiny patches,shadows,and water-stained noise which are operated from the rough segmentation.Finally,detecting the crack pattern automaticallybased on Hough transform.This method is based on the detection of part gray differences in the spatial domain,and is mainly applied to the detection of differential cracks which have fine and small gray scales.Since this method is a spatial domain algorithm,it is faster and is suitable for crack detection in the case of less road noise.In view of the complex noise on road,this paper proposes a crack detection method based on phase consistency and morphology.The method firstly performs rough segmentation on the image based on the principle of phase congruency,then connects the cracks based on the crack connection rules,and finally removes the noise from the connected crack images with morphological knowledge.This method is less affected by nonlinear noise,which is suitable for road images with relatively complex noise and less influence of uneven grayscale,and can be used to extract linear cracks.Aimed at the problem that the image of pavement cracks is seriously affected by the unevenness of grayscale,this paper proposes a crack detection method based on Markov Random Field.The method firstly proposes a gray correction method based on Gaussian fitting to correct the grayscale of pavement image and reduce the uneven gray level interference;Then Markov random field is used to segment the image and extract cracks.Finally,cracks are connected and the noise is removed by means of fracture connection and morphology.Finally,the fracture connection and morphology method are used to connect the cracks with breakpoints and remove the noise.The noise extracted by this method is relatively complete,is less affected by unevengrayscale,and has better anti-noise ability,and is suitable for road images with uneven gray scale and noise.This article combines these three methods with the actual situation of crack detection in the Institute of Transplantation to improve problems such as high misidentification rate of road crack detection,difficult detection of small cracks,and incomplete extraction of cracksaiming at three different road conditions.All three methods can extract cracks from the road surface image.The extraction effect is better,and it has certain practical value.Finally,through analyzing and contrasting the three methods proposed in this paper,reflecting the characteristics of the popular crack detection algorithms in recent years.