Application Research On Weak Edge Detection Of CT Image Based On C-V Model Method

As an advanced nondestructive detection technology, CT technology is widelyused in engineering applications. Industrial CT images can reflect the internal structureand defects of the workpieces. The internal defects in workpieces, such as castingporosity, porosity, and cracks, and material and imaging mechanism of CT system canled to wide gradation in edge areas, which is not exist in ideal step edge. Therelatively smoothly gradation of them is called weak edge. Weak edge detection of CTimages has been one of the difficulties in image segmentation. This dissertationresearches weak edge detection of CT images, which is carefully analyzed in terms ofthe advantages and disadvantages of common image segmentation method. Thisdissertation mainly focuses on C-V model method and its applications in weak edgedetection of CT image. The main work is as follows:First, to solve the over-segmentation when using C-V model method,morphological opening-closing reconstruction filtering is used to erase the details thesizes of which are less than that of the structural elements, and ensuring that the originalcharacteristics in the image are unchanged.Second, this dissertation analyses formation mechanism of volume effect in CTand its impacts on imaging. Firstly, CT images of cylindrical, which are not affected byvolume effects,and simulation pictures are used as experimental pictures to verify thatthe method including morphological opening-closing reconstruction filtering and C-Vmodel method and the follow-up least-squares fitting of edge can achieve sub-pixelaccuracy on strong edge’s segmentation. And then hemisphere and cone CT images withobvious volume effects are used as experiment pictures. And according to thesegmentation results on the volume effect and slice thickness, the relationship betweenthe curve slope of the workpiece axial edge and testing errors is analyzed.Third, defect detection methods and pretreatment method is analyzed. Andaccording to the characteristics of noise and defects in CT images, this dissertation usesmorphological opening-closing reconstruction filtering to remove image noise, whichhas better performance in keeping the basic shape characteristics and information of thedefects in images, and ensuring the edges will not be distorted. Then we reuse theC-V model method to segment CT images, and fill the defect area and calculate the sizeof defect area. The defect area will be fully extracted. Finally we compared the result with three classical methods and the results show the method has better performancein the extraction of weak edges causing by defected.In summary, this dissertation conducts specific application research based onC-V model method, and applies it in weak edge detection of industrial CT images whichis after pretreatment. It analyses the accuracy and application scope of our methodthrough several experiments. The results demonstrate the effectiveness and advantagesof the method on weak edge detection. When the method is applied in defect detectionof industrial CT images, it can extract connected and closed edge, which can providegood conditions for follow-up defect recognition and measurement of industrial CTimages.