The Pavement Crack Detection Algorithm Research Based On OpenCV

In recent years,China's economy has developed steadily,and highway mileage has been increasingly more and more.Pavement crack detection has gradually attracted the attention of people.Artificial pavement crack detection has high risk coefficient,low efficiency and high cost of detection,this way cannot meet the actual demand of social development,people are looking for more economic and efficient way of pavement crack detection.At present,computer vision technology is developing rapidly.Combining computer vision technology with efficient image acquisition system and GPS positioning system can realize quick and accurate pavement crack detection.In this paper we according to the actual demand of pavement crack detection,study the solution both in algorithm and theory,using C++programming language developed a pavement crack detection platform based on OpenCV and Qt.Image processing algorithms are implemented by OpenCV,UI design and database functions are implemented by Qt.The main tasks of this paper are as follows:Firstly,preprocess the pavement image.The three channels color image is transformed into grayscale image,and the image contrast is enhanced by gamma correction.Through median filtering and bilateral filtering,the noise is reduced and the preprocessing is completed.Secondly,In view of most current detection algorithm extract the cracks directly after preprocessing.This method usually needs to know a priori knowledge of whether exist cracks,not meet the requirements of intelligent detection.In this paper,through construct machine learning model detect and locate cracks.Realize the function that not need artificial observation and let computer automatically obtain the knowledge of whether there are cracks,meet the requirement of intelligent detection.We compared the indexes of the classification models constructed by SIFT,SURF and ORB feature extraction algorithms to determine their applicable conditions.Thirdly,if there are cracks in image extract crack region,identify crack type and calculate the corresponding parameters.Using local adaptive threshold segmentation algorithm for image segmentation,then extract the connected domains of segmented image,calculate circularity and area of connected domain,and wipe off the connected domain whichcircularity relatively large and area relatively small,now we get the crack target.For the image which crack extracted,we identify the crack type using improved projection algorithm,calculate the distribution area of network cracks,and calculate length and maximum width of linear cracks.Finally,we using C++ programming language developed a pavement crack detection platform based on OpenCV and Qt.Our system provides a friendly user interface,implemented pavement crack detection and extraction,identification and parameter measurement functions,we can step processing,we also can process using one key,we can handle multiple images at the same time,and also designed the database module to implement the function of manage image information.