Casting Defect Image Simulation Based On A Multiple Templates Model

With the growing quality demand of the casting production and the development computerscience, the casting defect auto-detection technique has started to draw the attention of thedigital image processing researchers. The development of auto-detection algorithm has beenlimited by the lack of test samples. Our casting defect image simulation research is to settlethe limitations and provide test samples for the auto-detection algorithm. Dendritic shrinkageis one of the most sophisticated casting defects. The simulation of dendritic shrinkage will behelpful to the development of other casting defect simulation algorithm.First, we study the formation process of the casting shrinkage in metal solidification theoryand point out that the shrinkage formation is a time-span process. Then, we analyzed thehistogram, gradient distribution, and the geometric characteristics of the defect radiographs.With the analysis above, we present a triple layers spatio-temporal model for dendriticshrinkage simulation and used skeleton as the representation of the defect geometricinformation.Second, this paper's simulation algorithm consists three parts: defect geometric templatesgeneration, defect gray values generation and defect-background fusion. In templatesgeneration, we present3skeleton points weight value assignment methods, and compare theiradvantages and disadvantages. Then we described a under-controlled anti-geometric diffusionalgorithm to simulate the shrinkage formation process.Third, we analyzed the defects' histograms and their gray values spatial distribution andpresent2defect gray value generation algorithm: random gray value generation method andgray-value generation method.Finally, we summarized our simulation algorithm, gave our simulation results andevaluated our simulation results from both subjective view and objective view. Then wepointed out that our algorithm is lack of randomness and gave our forecast of the simulationdevelopment.