Wu Defeng, Chen Tianfei, Li Aiguo
School of Marine Engineering, Jimei University, Xiamen 361021, China.
Fujian Provincial Key Laboratory of Naval Architecture and Ocean Engineering, Xiamen 361021, China.
Sensors (Basel). 2016 Aug 30;16(9):1388. doi: 10.3390/s16091388.
A robot-based three-dimensional (3D) measurement system is presented. In the presented system, a structured light vision sensor is mounted on the arm of an industrial robot. Measurement accuracy is one of the most important aspects of any 3D measurement system. To improve the measuring accuracy of the structured light vision sensor, a novel sensor calibration approach is proposed to improve the calibration accuracy. The approach is based on a number of fixed concentric circles manufactured in a calibration target. The concentric circle is employed to determine the real projected centres of the circles. Then, a calibration point generation procedure is used with the help of the calibrated robot. When enough calibration points are ready, the radial alignment constraint (RAC) method is adopted to calibrate the camera model. A multilayer perceptron neural network (MLPNN) is then employed to identify the calibration residuals after the application of the RAC method. Therefore, the hybrid pinhole model and the MLPNN are used to represent the real camera model. Using a standard ball to validate the effectiveness of the presented technique, the experimental results demonstrate that the proposed novel calibration approach can achieve a highly accurate model of the structured light vision sensor.
本文提出了一种基于机器人的三维（3D）测量系统。在所提出的系统中，一个结构光视觉传感器安装在工业机器人的手臂上。测量精度是任何3D测量系统最重要的方面之一。为了提高结构光视觉传感器的测量精度，提出了一种新颖的传感器校准方法来提高校准精度。该方法基于在校准靶标上制造的多个固定同心圆。利用同心圆来确定圆的实际投影中心。然后，在校准机器人的帮助下使用校准点生成程序。当准备好足够的校准点后，采用径向对齐约束（RAC）方法来校准相机模型。然后使用多层感知器神经网络（MLPNN）来识别应用RAC方法后的校准残差。因此，混合针孔模型和MLPNN被用来表示实际相机模型。使用标准球来验证所提技术的有效性，实验结果表明所提出的新颖校准方法能够实现结构光视觉传感器的高精度模型。
