The paper describes a method using digital image processing in the detection of vaguely defined visual defects on objects symmetric with respect to a rotation axis. Automotive wheels and hubcaps, fans, turbines, symmetrical ceramic goods, merchandise, etc., are examples of such objects. The method uses the object’s surface symmetry to identify areas that do not meet the requirement for the symmetry. The method is based on the brightness comparison of areas of the object’s surface under test corresponding to each other with respect to the object’s rotational symmetry. The area containing a defect is located through the difference between its brightness and average brightness of the all symmetric areas. The reliability of the method requires opaque and not too broken surfaces with solitary defects that do not overlap when the object is rotated. The method is advantageous for larger defects. Minimum defect size is limited by segmentation of the object and its production tolerances. Uniform illumination is another prerequisite for the reliable detection of the defects. This work focuses on testing the method and determination of the optimum brightness difference characterizing the defect. Next, limitations of the method are analyzed, especially the relationship between the uncertainty of the object shape, the camera resolution, and the minimum size of the detected defect.

中文翻译：

---

检测旋转对称物体上的视觉缺陷

该论文描述了一种在检测相对于旋转轴对称的物体上模糊定义的视觉缺陷时使用数字图像处理的方法。汽车车轮和轮毂罩，风扇，涡轮机，对称陶瓷商品，商品等就是此类对象的示例。该方法使用对象的表面对称性来识别不满足对称性要求的区域。该方法基于相对于物体的旋转对称性彼此相对应的被测物体表面区域的亮度比较。包含缺陷的区域通过其亮度和所有对称区域的平均亮度之间的差来定位。该方法的可靠性要求具有不透明且不太破损的表面，这些表面具有孤立的缺陷，这些缺陷在旋转对象时不会重叠。该方法对于较大的缺陷是有利的。最小缺陷尺寸受对象分割及其生产公差的限制。均匀照明是可靠检测缺陷的另一个先决条件。这项工作着重于测试方法和确定表征缺陷的最佳亮度差。接下来，分析该方法的局限性，特别是对象形状的不确定性，相机分辨率和检测到的缺陷的最小尺寸之间的关系。这项工作着重于测试方法和确定表征缺陷的最佳亮度差。接下来，分析该方法的局限性，特别是对象形状的不确定性，相机分辨率和检测到的缺陷的最小尺寸之间的关系。这项工作着重于测试方法和确定表征缺陷的最佳亮度差。接下来，分析该方法的局限性，特别是对象形状的不确定性，相机分辨率和检测到的缺陷的最小尺寸之间的关系。