边缘提取在涂胶缺陷检测中的方法研究

船舶发动机盖外板与内板之间常涂覆一层防震密封胶,但具体涂胶形态难以观测,常用机器视觉手段获取产品涂胶图片。针对涂胶区域分割和检测精度低,效率慢的问题,提出一种新的边缘提取算法获取涂胶区域。为了避免传统Canny边缘检测方法易受噪声干扰的缺点,采用具有保边特性的加权最小二乘法滤波器对图像滤波去燥,增加45°和135°两个方向计算图像梯度,利用改进后的Canny算法获取边缘,对获取到的边缘点用区域生长方式将涂胶区域提取出来,最后将提取到的涂胶区域与标准涂胶区域进行配准和减影,得到减影结果即为涂胶缺陷。试验仿真表明：本算法能够很好的提取到胶水区域,并且获取胶水缺陷。

Research on Edge extraction in Defect Detection ofGlue-coated Pictures

A layer of shockproof sealant is often coated between the outer plate and inner plate of Marine engine cover, but the specific form of sealant is difficult to observe, and pictures of product sealant are often obtained by means of machine vision. Aiming at the problems of low segmentation and detection accuracy and low efficiency, a new edge extraction algorithm was proposed to obtain the glue-coated region. In order to avoid the traditional Canny edge detection methods are susceptible to noise interference of faults, the cover edge characteristics of filter weighted least-square method of image filter to dry, 45 ° and 135 ° two directions based on gradient image, using the improved Canny algorithm for edge, the edge points to get to the region growing way to pick up the glue area, in the end will be extracted to glue area and standard coating area for registration and subtraction angiography, subtraction angiography results is the coating defects. Experimental simulation shows that this algorithm can extract the glue area and get the glue defect well.