The Extraction And Dynamic Tracking Research Of Spot Images By Femtosecond Laser Ablation

In the process of processing single crystal silicon by the femtosecond laser,there is a phenomenon of derived plasma luminescence,which formed the series spots.The laser spot image contains effective information about processing parameters,such as processing power and speed of femtosecond laser.Through analysis of the geometric features of the spot,it can be used in the closed-loop feedback system of micromaching.Because the spot centroid can be used to locate the spot location accurately,the serial spot centroids were used as the sampling data,which were tracked by Kalman filter,the movement law of the spot was grasped,and then the ablation law was explored.Firstly,the series spot images is collected and their ablated core area is extracted by the way of bit layering,so that the peripheral halo is eliminated effectively.The spot image is enhanced by Gauss guided filtering,so that its interpretation is improved;secondly,the two domain product gray centroid method is introduced to extract the spot centroid,the extraction error caused by the influence of skewed spot is reduced by 1.0042px;thirdly,in order to compensate for the noise influence and optimize Kalman filter equation,the spot centroid are tracked,the tracking accuracy is improved,and the error is reduced to 0.3490 px;fourthly,according to the characteristics of skewed spot,the spot diameter is calculated by the distance proportion sum,and by tracking the change speed of the spot,it can be known that the actual ablation speed is the sum of the processing speed and the average change speed of the spot.Based on the two influencing factors,the spot overlap rate is redefined;finally,by analyzing the characteristics of dynamic spot,it can be seen that the spot exhibits two different motion states.Based on the dynamic analysis results,an optimized Cascade SVM classifier is proposed to classify the spot images under 10 mw,20mw,and 50 mw,and separate the spots with different motion states.The corresponding classification accuracy is increased to 100%,100%,98.5%,the number of spots under different ablation states is grasped,the ablation law can be better controlled,which helps to improve the processing efficiency and accuracy.Figure 39;Table 3;Reference 62...