Using Kalman Filtering and Segmentation Techniques toCapture and Detect Cracks in Pavement

摘要

For this study we used a CCD video camera torncapture the pavement image information via the computer.rnDuring investigation processing, the CCD video camerarncaptured 10～30 images per second. If the vehicle velocity isrntoo fast, the collected images will be duplicated and if thernvelocity is too slow there will be a gapped between images.rnTherefore, in order to control the efficiency of the imagerngrabber we should add accessory tools such as the DifferentialrnGlobal Positioning System (DGPS) and odometer.rnFurthermore, Kalman Filtering can also solve these problems.rnAfter the CCD video camera captured the pavement images,rnwe used the Least-Squares method to eliminate images ofrngradation which have non-uniform surfaces due to thernillumination at night. The Fuzzy Entropy method calculatesrnimages of threshold segments and creates binary images.rnFinally, the Object Labeling algorithm finds objects that arerncracks or noises from the binary image based on volumernpixels of the object. We used these algorithms and testedrnthem, also providing some discussion and suggestions.