Research On The Image Detection System And Recognition Of Pavement Dilapidation

Pavement dilapidation data is the most important data of pavement maintenance management. At present, this data is acquired by manual detection. But manual detection is inefficient and time-consuming, at one time there are many unsafe factors on fieldwork. And so, there is the important meaning on developing research on Pavement dilapidation auto-detection and recognition.The system of pavement dilapidation, the worldwide state-of-the-arts and trends of pavement dilapidation image processing and the classified methods of image processing are introduced by this thesis. The main work concluded hereinafter: firstly, the laboratorial system of pavement dilapidation image processing is designed. The designed work concludes that confirming the framework of vision detection subsystem and installation parameters, how selected the camera and object lens. Secondly, the pretreatment of pavement dilapidation image is achieved. On pretreatment the image noise is removed by filter technology, a bilinear interpolation based on mean square fit is erected and the image distortions are corrected. Thirdly, pavement dilapidation image is divided into subblock image, and the marginal information is extracted from subblock image by the region detection algorithm. Based on above works, the length and width of pavement dilapidation image are acquired by computational technique of long diameter and short diameter, and the defect areas are worked out at according to defect bounding pixel amount. On defect recognition aspect, tri levels BPNN (Back Propagation Neural Network) classifier is designed in this thesis. The eigenvector of BPNN is combined with geometrical shape features and textural features. The designed BPNN is trained by models, the result shows that convergence rate of this BPNN is faster.The pavement dilapidation images, such as transverse crack image, longitudinal crack image and block crack image and so on, are processed, measured and recognized by laboratorial detection system described by this thesis. The results of experiment show that this system can auto-detect and accurate-recognize many pavement dilapidations such as, transverse crack, longitudinal crack and block crack, but the veracity of check crack detection and recognition is further improved.The research of this thesis is completed on laboratory. Some helps must be gived for the auto-detect and recognition of pavement dilapidation by these exploratory works.