Research On Key Technologies Of Image Detection For Bridge Deformation

High precision detection of the huge construction deformation using image processing technology is one of the most active research areas in detection and measurement. Because of the civil engineering such as bridge are large scene, the acquiring of the total bridge often meets the contradiction of high resolution and small region of detection. For the purpose of improve the image resolution of the bridge, we choose sequential imageas the acquiring methodto deal with this problem. Based on the sequential image, this paper aimed at studying of the image segmentation, image mosaic, and the extraction of bridge deformation, In short, the main contribution of this paper may be generalized as:· Based on the image segmentation model of ACM (Active contour model), this paper focused on the influence of initial contour choosing, and proposed an region and edge overlap ratio based initial contour choosing algorithm. This method may improve the stability and compatibility of ACM based segmentation algorithms.· When the input image is large, the iteration of ACM based segmentation models are time and memory consuming. To keep the segmentation processing in control, this paper adjusted the segmentation procedure into two step segmentation, the rough and accurate segmentation. The first step aimed at the segmentation of room outed rough image and acquire the key content of bridge regions from each sequential image. Then, the key region is divided into some small images.The combined small image segmentation is just the result of one sequential image. The proposed segmentation method improved the iterative efficient and may be used in parallel processing.· In the study of image mosaicking processing, this paper aimed at the distribution, number, and extraction procedure of character points of the adjacent sequential images. Considering of the adaptive of image contents, this paper proposed a characteristic value based and fixed number extraction method. According to the experiments of different images, it may keep the compliance and uniformity of character points.· To improve the mosaicking accurate and reduce the wrong matching probability, this paper improved a two-step mosaicking algorithm. In the first step, we aimed at finding out the overlapped region of each pair of sequential images and add the judgement standard of matching result in the RANSAC procedure. After the overlapped regions have been determined, the character points re-extracted accurate matching would be done on the overlapped region. The two-step mosaicking procedure may improve the accurate of detection.· To realize the extraction of bridge deformation, this paper gives a comparative study of different extraction ways. In this paper, we extract the deformation of each sequential imageand merge the total data according to the overlapped region. The simulating experiment shows that this extraction method may keep the detection accurately.The image detection of bridge deformation faced many problems, in this paper, we have been given and tentatively study of the total procedure, such as, bridge deformation image acquiring, the segmentation and mosaickingof sequential images, and the extraction of deformation data.