靶准直传感器封装误差检测模型与修正方法

利用三个面阵电荷耦合元件(charge-coupled device,CCD)构成的靶准直传感器实现柱状腔靶高精度姿态检测时,靶准直传感器中CCD封装误差等因素的存在会给柱状靶姿态检测及定位带来不确定影响，针对此，提出了靶准直传感器的标定模型与实验方法。将直径为0.8mm的微型球靶作为检测参考目标,利用微型球靶在靶准直传感器中不同位置时CCD上成像的图像点来构建位置矩阵,利用位置矩阵,通过归一化线性优化方法求解表示CCD位姿关系的单应性矩阵,通过单应性矩阵修正因靶准直传感器封装误差给柱状靶姿态检测带来的影响。考虑到利用图像检测方法提取微型球靶中心点位置时误差的存在,在构建的位置矩阵中加入均值μ=0、方差σ=0.5的高斯噪声,模拟检测点位置误差对单应性矩阵求解的影响，分析结果显示位置误差对单应性矩阵中各个参数影响较小，从而证明了该优化方法求解单应性矩阵的鲁棒性。为验证靶准直传感器对腔靶姿态检测的准确性,实验以0.02°的步距角绕X轴旋转柱状靶,利用靶准直传感器检测柱状靶姿态变化并与实际的姿态变化进行对比分析,实验结果显示最大误差小于0.009°,证明了修正方法的可行性。

Detection Model and Rectified Method for Packaging Errors of TAS

Considering packaging errors of TAS composing of three CCD sensors and other factors to bring an uncertain impact of orientation detection and positioning accuracy for micro cylinder target, this paper presented a calibration model and experimental method. Using a spherical target with a diameter of 0.8mm as a detection reference target, position matrix was composed by placing the spherical target to different positions in the TAS. A normalized optimized linear method was proposed to resolve the homography matrix using position matrix which expressed the orientation relation for two CCD sensors, and the homography matrix can eliminate the packaging errors to reduce the effect of orientation detection for spherical target. Considering the existence of the position detection errors using image analysis to extract the center of spherical target,the position matrix was constructed by adding the Gaussian noise with the mean of  0 and variance of 0.5 pixels to analyze the detection errors  how to affect the resolution of homograph matrix. The simulation results prove that Gaussian noise has little effect on the homography parameters, thus the optimization method is verified to be robustness. To testify the orientation detection accuracy of cylindrical target using TAS, a high-precision experiment was set up. The orientation change of cylindrical target was measured rotating around the X-axis comparing with the change of actual orientation. The experimental results show that the maximum error is less than 0.009 degrees and the modified method is verified to be feasible.