线激光条纹特征的健壮性快速提取

The Fast Robust Extraction Method of Line Laser Stripe Features

为了提高传统的线激光条纹特征提取算法的精度、健壮性和计算效率,使其能够满足更大范围的工业应用,本文在平顶激光模型的基础上提出了线激光条纹中心的健壮性快速提取算法。首先将线激光条纹建模为平顶激光来提高其对各种场景激光条纹的检测鲁棒性;同时,在遍历图像查找激光条纹中心点的过程中加入一种"加窗预判加速机制",将每个激光条纹中心的检测范围限制在与前一个目标点相连的梯形域中,极大地提高了条纹中心点的检测速度。仿真实验表明:当图像中亮度噪声水平小于6时,该算法的条纹中心提取误差小于0. 07pixel,并且检测单张130万像素的线激光条纹图像耗时小于5ms。对各种零件的实际测量实验表明:该算法可以用于光滑、粗糙、深色、浅色、普通金属材质等各种物体的线激光条纹特征的提取,具有很强的健壮性。

In order to improve the accuracy,robustness and computational efficiency of the traditional line laser stripe feature extraction algorithm,this paper proposes a fast robust extraction algorithm of the line laser stripe center based on the flat-top laser model,which can meet a wider range of industrial applications. Firstly,the line laser stripe is modeled as a flat-top laser to improve its robustness to the detection of various laser stripe. At the same time,a "windowing pre-acquisition acceleration mechanism"is added in the process of traversing the image to find the center point of the laser stripe,limiting the detection range of each laser stripe center to the trapezoidal domain connected to the previous target point and greatly improving the detection speed of the stripe center point. Simulations show that when the brightness noise level is less than 6,the stripe center extraction error of the algorithm is less than 0. 07 pixel,and the detection of the line laser stripe image whose size is 1. 3 million pixel takes less than 5 ms. The actual measurement experiments on various parts show that the algorithm can be used to measure the line laser stripe features of various objects such as smooth,rough,dark,light color,and common metal materials,and has high robustness.