Classification of Volume Data Based on Support Vector Machine
Hong Liang · Jisuanji fangzhen · 2009
In the volume data derived from the industrial CT images, the accuracy of classification and the results of volume rendering are degraded due to the similarity of the gray of low density materials comparing with the background and artifact.Consequently a new classification algorithm based on support vector machine was proposed for volume rendering.Firstly, the decision function was achieved by training the specific samples which contained four features:gray and gradient of voxel, entropy and moment of local histogram.Secondly the volume data was marked by this decision function.Then the gray of the voxel with the same marker was transformed to appointed intervals.Finally the opacity of the voxel was assigned by the transfer function and the classification of volume data was completed.Experimental results indicate that the classification of volume data is obtained with the better accuracy by the proposed algorithm.The industrial components are shown explicitly in the volume rendering results and successfully disassembled in the computer simulated.