| Defect recognition of the DR(Digital, Radiography, DR) image has been an important research direction in the field of image processing.The general algorithm has many steps and cannot recognize the defects fast and in real-time. Base of these problems and the rapidly development of the pattern recognition field, LBP(Local Binary Pattern, LBP) recognition algorithm is applied to the field of recognize defects.LBP algorithm is a recognition algorithm that based on texture, which is originally widely used in the field of face recognition. LBP algorithm has characteristics that the principle is simple and the recognition speed is fast.Above all, it can class quickly these defects according to matching the template image of defects from defect template image database, but the characteristic that LBP algorithm is very sensitive to noise, which limits the application of LBP algorithm. And compared with the face recognition database which is widely used in face recognition, in industrial field, the defect template image database has a small number. These problems hold back the development of LBP. Based on these problems, the main research in this paper is listed as follows:First: introduce the principle and origin of LBP algorithm and expounding the recent development of LBP algorithm.Second: According to the characteristics that traditional LBP algorithm is very sensitive to noise, and the characteristics of DR images, this paper proposes an improved LBP algorithm: WALBP(Weber Adapted Local Binary Pattern). The algorithm combines with other algorithms, considers the more description information of the defect image, which can effectively suppress the effect of noise, improve the recognition rate, and confirm the algorithms has a good recognition result by comparing with other recognition algorithms.Third: There are many template images in industrial field, but there isn’t a good management tool to manage these defect images. In order to achieve convenient sorting defect template image and defect recognition, a small defect template image database is established. Finally, combining the existing improved LBP algorithm and the template images database, defect recognition software is designed.Through the in-depth study of the LBP recognition algorithm and the characteristics of DR image, LBP algorithm is improved and the efficiency of defect recognition is increased. According to the establishment of database and design of software of defect recognition, there is a good integration to the process of recognize defects, which makes great significance to promote the development of LBP in the field of DR image defect recognition. |