共 21 条
Femtosecond laser ablation power level identification based on the ablated spot image
被引:3
作者:
Wang, Fu-bin
[1
,2
]
Tu, Paul
[2
]
Wu, Chen
[3
]
Liu, Yang
[1
]
Feng, Ding
[4
]
机构:
[1] North China Univ Sci & Technol, Sch Elect Engn, Tangshan 063009, Hebei, Peoples R China
[2] Univ Calgary, Dept Mech & Mfg Engn, Calgary, AB T2N 1N4, Canada
[3] Univ Sci & Technol Beijing, Sch Automat Engn, Beijing 100083, Peoples R China
[4] Yangtze Univ, Sch Mech Engn, Jingzhou 434023, Hubei, Peoples R China
关键词:
Femtosecond laser;
Silicon wafer ablated mark;
Image feature;
Ablation power;
SVM recognition;
MATERIAL REMOVAL;
ALGORITHM;
D O I:
10.1007/s00170-017-1007-5
中图分类号:
TP [自动化技术、计算机技术];
学科分类号:
0812 ;
摘要:
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.
引用
收藏
页码:2605 / 2612
页数:8
相关论文