Research On High Accuracy Volume Measurement Method Of Industrial CT Sequence Image

As a most advanced nondestructive detection technology, CT technology is able toclearly show the interior structure detail of the detected object, and quantifiedly applythe two and three-dimensional size information of details. Medical CT is able tomeasure the3-D size of lesion. But in the actual treatment, it does not require too highmeasurement accuracy. In corresponding, with the detected objects structuresophisticated and various, and moreover, the inequality of inside slice SpatialResolution and inter slice Spatial Resolution, the existing industrial CT sequence imagebased3d volume measurement method needs to be improved in accuracy, speed, andautomation degree. As this background, based on the existing image measurementtheory, and with CT sequence image the experiment subject, two ICT sequence imagebased volume measurement methods are proposed.Firstly, a volume measurement method called Frustum Volume Accumulation isproposed and implemented, which is based on Zernike moment sub-pixel edge detectionand polynomial fitting. After preprocessing the image, the sub-pixel-level contour ofeach slice is detected through the Zernike moment, and then tracked by EightNeighborhood Tracking Method. Then, by using the polynomial fitting method, thecontour is fitted, and the fitted contour is sampled by the same angle interval. After that,each edge point at the same phase angle of each slice is matched to calculate the contourpoint coordinates of the middle slice through inter slice fitting&interpolation algorithm,and then apex of work-piece is calculated through extrapolation method. Finally, as eachtwo adjacent slice contour the top and bottom of a frustum, accumulating the volume ofeach intermediate frustum to be the total volume of the workpiece.Secondly，a volume measurement algorithm based on the integral of the sub-pixelcontour rotation is proposed and implemented. The preprocessed images is detected byZernike moment method to extract sub-pixel level contour, and then tracked by EightNeighborhood Tracking Method to get the complete closed edge contour. By using thepolynomial fitting method the contour is fitted, and then the fitted contour is sampled bythe same angle interval. Then, according to the degree of overlap of projection contoursof each slice to determine the topology of each slice, then find a best rotation axis. Eachedge point at the same phase angle of each slice is matched to calculate the contourpoint coordinates of the middle slice through inter slice fitting&interpolation algorithm, and then apex of work-piece is calculated through extrapolation method. Finally,rotating each fitting inter slice contour one phase angle interval to calculate the integralas a rotation volume, accumulating the volume of each rotation to be the total volume ofthe workpiece.Equidistant CT scanning experiments were done towards three typical shapedstandard steel work pieces and an actual workpiece with cavity. By using the proposedmethods, volume measurements experiments and analysis were carried out towards thescanned sequence images. The result indicates that the measurement accuracy of bothproposed volume measurement algorithms have improved by2-3order of magnitude.Each measurement error is below1%, which can meet the high precision requirementsof engineering applications.