Research On Detection The Center Of Rotation In Industrial Computed Tomography System

The center of rotation（COR） of the industrial computed tomography systemdeviating from the reconstruction COR makes the edge of reconstruction image blur、spatial resolution and the precision of size-measurement decline.Therefore,it’sparticularly important to position the COR accurately.This dissertation introduces and analyzes principle and characteristics of the twoCOR positioning methods in detail, which are based on projection. Actual CT projectiondata experiments show that these kinds of methods are simple and efficient. However,they are weak in positioning accuracy and anti-interference. The opposite-angle methodis easily affected by air noise and random noise as well as the influence of volume，soits precision is low. The method based on the mass center of projection is easily affectedby air noise distribution difference of the projection’s both sides，so its accuracydeclines. Based on the analysis above, I proposed a method based on edge extraction,and it’s accuracy is higher than other methods, reaches sub-pixel, while it can’t meet thehigh requirement for the high accuracy CT system still.Aiming at the problem of low position accuracy of methods based on projection,this dissertation analyzes its reasons, researches the relationship between the center ofrotation offset and the edge response function of a CT image. Researching onexperimental data from actual CT images, we find that COR offset and the extremevalue of point spread function form gaussian distribution within a certain scope. In thatcase, this dissertation proposes a method to position the COR by fitting the height of thepoint spread function which has different COR offset, referred as PSF method. A lot ofpractical experiment show that, under the condition that image pixel size closes tospatial resolution limit, this method can position the COR offset accurately reaching10-2pixel.Compared with other methods based on projection, the positioning accuracyof this method improves an order of magnitude. After the center of rotation correctionby PSF method, the spatial resolution of reconstructed CT images absolutely improveswell.Finally it discusses the suitability of the method, also analyzes whether the threefactors which are image SNR, image ring artifacts and crosstalk artifacts caused bydetector can affect the positioning accuracy of my method or not. A lot of experimentson actual CT images show that they slightly influence its accuracy. But the positioning precision of the method is still much higher than other methods when the image SNR islow or there are ring artifacts or crosstalk artifacts. So the method proposed in thisdissertation can accurately position the image reconstruction of rotating center and hasstronger adaptability.