论文部分内容阅读
针对惯性约束聚变(ICF)光学元件损伤问题,提出了一种基于随机抽样一致性(RANSAC)及最小二乘支持向量机(LSSVM)回归的高精度检测方法。建立了损伤区域总灰度与实际尺寸的回归模型,通过该回归模型对待检测损伤区域的尺寸进行预测,得到损伤区域的高精度尺寸。为剔除回归模型建立过程中离群样本点的影响,采用RANSAC方法对训练样本进行优化选择。针对抽样组中样本数对检测精度及检测效率的影响进行了相关实验,确定了抽样组中样本数的合适区间。RANSAC-LSSVM方法可通过改变误差评价函数得到不同评价体系下的最优回归模型。实验证明,在传统像素级检测方法的基础上,该方法将损伤尺寸检测的平均相对误差降低了近90%。
Aiming at the damage of inertial confinement fusion (ICF) optics, a high-precision detection method based on RANSAC and LSSVM regression is proposed. A regression model of the total gray area and the actual size of the damaged area is established. The size of the damaged area to be detected is predicted by the regression model, and the high-precision size of the damaged area is obtained. In order to eliminate the influence of the outlier sample points during the regression model establishment, the RANSAC method was used to optimize the training samples. Aiming at the influence of the number of samples in the sample group on the detection accuracy and detection efficiency, experiments were carried out to determine the appropriate number of samples in the sample group. The RANSAC-LSSVM method can obtain the optimal regression model under different evaluation systems by changing the error evaluation function. Experiments show that this method reduces the average relative error of damage detection by nearly 90% based on the traditional pixel-level detection method.