深度神经网络在自动瞄靶技术中的应用研究

Research on the application of deep neural network in automatic targeting technology

基于Mask R-CNN模型,对惯性约束聚变实验研究中靶图进行关键结构的语义分割,通过计算中心来定位目标靶点位置,通过模拟靶图数据来验证算法的可行性。在算法验证过程中通过程序批量生成不同结构参数、不同角度的模拟靶图作为数据集,然后通过增噪、成像位置变换等操作拓展样本数量。在大样本数量基础上对算法模型进行训练测试,实现语义分割,通过计算注入孔的中心得到目标靶点位置。测试结果显示准确率和召回率均在97%以上,在靶点识别精度上优于10个像素点。

Using the Mask R-CNN to segment the key structure of the CCDs image in inertial confinement fusion(ICF) experimental research,the center of the entrance hole is calculated as the position of the target,and the rationality and feasibility is verified with the simulated image data.Firstly,the program generates batches of simulated images with different structure parameters and different angles as a data set,secondly the operation of noise enhancement and position transformation are made to expand the samples,finally the algorithm model is trained and tested based on the samples,and the center of the entrance hole can be located as the position of the target.The test results show that the accuracy rate and recall rate are above 97%,and the target recognition accuracy is better than 10 pixels.