金属球面缺陷的图像检测方法

为解决测量速度和精度之间的矛盾,提出了一种用图像识别技术探测缺陷的大致位置、用激光位移传感器精确测量缺陷三维形貌的测量方法.设计了扫描测量球形表面的图像和数据采集装置;在分析图像特点的基础上,采用低通滤波算法降低了图像背景噪声;根据当前待处理图像上极大极小值的包络线差值信息,初步确定了缺陷的大致位置和区域;用Snake 模型迭代逼近到缺陷的边界;讨论了缺陷三维形貌数据的处理和表征方法.该方法在复杂背景图像分割的自适应能力方面比传统方法更具优势,测量系统能满足自动、可靠地检测金属球面缺陷的需要.

An Image Measurement for Defect on Metallic Spherical Surface

In order to resolve conflict of measuring speed and precision, a novel measurement was proposed to detect approximate defect positions by image recognition and to measure accurately the three-dimension profile by laser displacement sensor. The image capture and data acquisition equipment for scanning measure was designed for spherical surface. The image background noises were reduced by low-pass filtering algorithm based on analyzing image features.According to the difference between high and low envelope curve constructed by maxima and minima on the current processing image, the approximate defect positions and region were determined. The defect edge was approached iteratively based on Snake model. The data processing and defect representations of the three-dimension profile were discussed. This method outperforms traditional image segment methods at serf-adaptation to complicated. background images. Moreover, the measuring system can measure automatically and reliably to detect surface defects.