一种改进Canny算子芯片图像边缘检测方法

An improved algorithm for chip image edge detection based on Canny

采用机器视角对芯片表面封装缺陷进行检测对提高芯片生产效率意义重大。该文对传统Canny算法进行详细分析后,根据采集到的芯片图像对比度低、容易受光照不均匀和噪声污染等信息影响的特点,对传统Canny算法进行了改进,提出一种新的芯片图像边缘检测方法。通过迭代双阈值最大类间差法取得最优阈值后,再开始边缘检测。在系统自带图像和采集到的缺陷芯片图像上对2种算法进行了对比实验。结果表明,新方法能去掉冗余信息,可准确提取芯片图像的缺陷轮廓,提高后续检测的准确性。

Using the machine perspective to detect the packaging defects on the chip surface is of great significance to improve the chip production.Cause the chip images are easy to be affected by uneven illumination and noise,a new edge detection method is proposed after analysis of the traditional Canny algorithm.By using maximum classes square error algorithm,the optimal threshold is obtained.Compared the two algorithms on the ordinary and the defect chip image,the results show that new algorithm can remove the redundant information and improve the edges accuracy detection rate of chip image.