Demonstration And Verification For Automatic Alignment Of ICF Diagnostics In Laboratory

With the ever-changing world situation and increasingly prominent energy issues,laser inertial confinement fusion（ICF）is of great significance in the field of controllable thermo nuclear fusion research.In the ICF experiment,multiple high-power lasers are incident on a tiny holhraum located in the center of the vacuum target chamber,so it can drive the holhraum to achieve high-gain fusion.In order to obtain the physical state information of the center of the holhraum during the gain process,a variety of diagnostic systems are used for measurement and diagnosis,and the alignment accuracy of the diagnostic system is particularly important.The current diagnostic holhraum in our country adopts the method of binocular alignment and needs to use manual interpretation to locate the holhraum,resulting in unsatisfactory recognition accuracy.On the basis of investigating relevant domestic and foreign diagnostic alignment methods,this thesis proposed a holhraum recognition and segmentation algorithm based on Mask R-CNN,and developed a laboratory offline diagnostic alignment verification and demonstration system combined with the diagnosis and alignment process of the real experimental environment.The main work of the thesis includes:（1）The motion model and binocular alignment method of the real diagnostic platform are analyzed,and a simplified strategy for the diagnosis platform based on the electronically controlled displacement platform and the alignment section was proposed.Based on the image acquisition card and the motion control card,an interface software integrating motion control and image acquisition was written,and the construction of the laboratory alignment platform was completed.（2）Indepth study of traditional image recognition algorithms,this thesis used edge contour detection and Hough transform-based circumference detection for ball holhraum,calculated the coordinates of the center point and point out that under different environments and holhraum types,traditional image recognition algorithms have limitations.（3）Aiming at the problems of manual interpretation and traditional image recognition,a segmentation algorithm based on Mask R-CNN is proposed,which effectively overcomeed the recognition problems caused by different target types in different scenes,and improved the Mask R-CNN algorithm,which further improved the recognition accuracy of the model.（4）Aiming at the problem of data labeling difficulty,the original target image data set was produced by means of data enhancement,and an automatic generation method of holhraum image annotation data based on the HOOPS simulation platform and the laboratory holhraum setting platform was proposed,which realized the arbitrary expansion of the data of the simulated holhraum image and the real holhraum image under different attitudes.And the Mask R-CNN algorithm model was improved,the model comparison test was carried out to further improve the recognition accuracy of the model.（5）Aiming at the problem of the accuracy verification of the aiming system,this thesis proposed an accuracy verification method for the laboratory holhraum platform based on the binocular alignment section.The alignment system is evaluated with the in-plane accuracy and radial accuracy,and the calculated aligning accuracy meets the experimental expectations.The experimental results and accuracy verification show that the holhraum image data generation method and the improved Mask R-CNN algorithm proposed in this thesis can further improve the model recognition accuracy.The developed laboratory diagnostic alignment platform has passed the accuracy verification algorithm,and the alignment accuracy in the plane is 8.49μm,and the radial alignment accuracy is 126.63μm,which realized the high-precision alignment in the ICF diagnostic alignment laboratory scene.