Pyramid Matching Based Automatic Screen Defect Detection for Smart Meter

自动检测具有成本低、效率高的优势，逐步成为智能电表字符缺陷检测领域的主要趋势。然而，目前的自动检测方法依赖于字符分割的准确性，极易受屏幕倾斜、背景噪声或亮度不均的影响。为此，本文提出基于高斯金字塔匹配的自动检测方法。该方法无需字符分割过程，而是首先对模板图像中每个字符，线下提取其多尺度金字塔模板，然后对待检测的屏幕图像，利用快速模板匹配搜索其中的字符，最后依据字符个数判断屏幕是否残缺。研究结果表明：该方法能够避免图像质量的影响，快速、准确地检测屏幕缺陷。 Due to the advantages of low cost and high efficiency, automatic detection, instead of manual de-tection, will become the major trend for smart meter character defect detection. However, current automatic detection methods heavily rely on the accuracy of character segmentation. Moreover, they are easily influenced by the image quality, e.g., inclination, noise, and non-uniform intensity of the screen. To this end, this paper proposes a Gaussian pyramid matching based on automatic method without character segmentation. This method first proposes the multi-scale pyramid temples of the standard image. Then, the fast template matching is used to search the characters in the image. Finally, according to the number of characters detected, we can determine whether the meter has defection. Experiments on real meters show that the proposed method is not sensitive to the image quality, and can achieve fast and accurate detection of the screen defects.