A Design Of Concrete Pavement Disease Detection And Estimation System Based On Image Processing

In order to improve the acquisition efficiency and detection precision of concrete pavement disease information, this paper realizes A Concrete Pavement Disease Detection and Estimation System Based on Image Processing, which can rapidly acquiesce concrete pavement disease information and estimate maintenance period and fees.At first, confirm the requirements by investigating pavement maintenance and categories of concrete pavement disease, then design the whole scheme. The system applies a scheme with terminals of handheld, client and server. Handheld terminal collects pavement disease image information at whole angle, which is sent to client terminal and sever terminal thought wired or wireless. Image information is classified and analyzed at Client terminal automatically to obain disease parameter, meanwhile, client terminal predicts maintenance period and fees. Sever terminal saves all of information and result.Handheld terminal is based on Android smart device equipped with touch display screen, camera, communication module, etc. It installs a concrete pavement disease image and information acquisition software designed on Java and Android Develop Tools. Client terminal hardware is applied on personal computer with Windows OS. A pavement disease image and information processing software is developed thought Python, PyQt and OpenCV. Besides, set up a FTP file server and a Sqlite database on server terminal to receive and save information came from handheld terminal.For purpose of boosting auto-processing speed of concrete pavement crack image, this paper also has studied the method of concrete pavement crack auto-detection and processing. Because of inevitable distortion on pavement disease image captured by handheld terminal, camera calibration is accomplished at first to obtain pterior parameters and exterior ones. Secondly, preprocessing is implemented, which includes gray image, gray-scale stretching and median filter. Thirdly, choose some sorts of edge detection algorithms to segment crack and background. After that, due to the different count of background connection domain between linear crack image and reticular one, mark background connection domain algorithm is used to classify cracks automatically. Linear crack can be subdivided to vertical, horizontal and slant crack thought projection method. The length of Linear crack can be obtain by thinning to single pixel line, then the mean width of linear crack is calculated by its percent of plate pixel count. Finally, finding reticular crack minimum enclosing rectangle, the damaged area is calculated. Through experiment, the result shows that the algorithm improve the accuracy rate of classification of cracks and disease parameters effectively.