基于目标纹理的机器视觉转角检测方法

目标姿态角度检测系统是确保智能制造设备自动化、智能化的核心部件之一，在智能制造设备中得到了广泛的应用。提出了一种基于目标纹理的机器视觉转角检测算法，该算法通过椭圆聚类提取图像中的感兴趣区域，采用Hough梯度变换寻找感兴趣区域中包含特定纹理特征的区域，结合先验信息对特征区域进行标记、匹配，进而计算出待测物体的滚转角度，算法的运行耗时小于0.5s。采用该算法实现了浮子组件在浮液中的径向滚转角检测，并模拟实际检测环境测试了该算法的精度。测试结果表明，绝对精度小于0.3°，重复性精度小于0.08°，满足实际生产现场对检测精度的要求，解决了人工检测精度低、耗费人力的问题，提升了浮子转角检测的精度和智能程度。

Machine Vision Angle Detection Method Based on Target Texture

In intelligent manufacturing equipments, the target attitude angle detection system has been widely used, which is one of the core parts to ensure the automation and intelligence of intelligent manufacturing equipments. A machine vision angle detection algorithm based on target texture is proposed in this paper. The algorithm extracts the region of interest in the image by elliptic clustering, uses Hough gradient transform to find regions with specific texture features in regions of interest, and combines the prior information to mark and match the feature regions, then calculates the rotation angle of the object to be tested. The running time of the algorithm is less than 0.5s. The algorithm is used to detect the radial rotation angle of the floater in the floating fluid, the accuracy of the algorithm is tested by simulating the actual detection environment. The test results show that the absolute accuracy of the algorithm is less than 0.3°, and the repeatability accuracy is less than 0.08°. It meets the requirement of the actual production site for the detection accuracy, solves the problem of low manual detection accuracy and labor consumption, and improves the accuracy and intelligence of the floater rotation angle detection.