Femtosecond laser ablation power level identification based on the ablated spot image

When using a femtosecond laser to process the microchannel on silicon wafers in order to determine the machining speed in the X axis and feed rate in the Y axis, a large number of the spot ablation experiments on the surface of silicon wafer sample are needed. To determining the relationship between the size of spot mark and the corresponding power level, this paper presents a method based on image features of spot mark and Support Vector Machine (SVM). Firstly, we propose a preliminary segmentation strategy for spot mark image based on the K-means clustering, the image is clustered into backgrounds, halos, and targets, the image corrosion method is used to remove the interference around the target, and according to the deterministic characteristics of convex hull method to extract the target marks, consequently, the spot target with different ablation power has a uniform segmentation criterion. Secondly, we extract the geometric feature parameters of a spot mark image, such as pixel area, height, width, inner radius, and outer radius, and then construct the geometric parameters sample set of spot mark image. Finally, taking the geometric feature parameters of spot mark image as the input vector and the corresponding laser power level as the output vector, the recognition model of laser power level based on SVM is constructed.