基于不变矩特征的图像识别

Image Searching and Recognition Based on Moment Invariant

针对目标二维形状发生平移、旋转和尺度变换的情况,采用不变矩作为图像识别的特征量。图像灰度变化时,对应的直方图形状基本保持不变。将直方图的统计特征与不变矩相结合,提出了不变矩理论的图像识别方法。通过实验验证,这种方法对图像的亮度、对比度变化不敏感,可以有效地降低光照的影响,能够更准确地对图像进行识别。

Aim at the target which has been translated, rotated and scaled, the paper gives a method which uses moment invariants as the matching feature of the image searching and recognition. When the gray level of the imge changes, the shape of histogram will not change. With the statistical features of the image histogram and the invariant moments, a matching algorithm based on moment invariants is given. This method is not sensitive to luminance and contrast, and could reduce the influence of illumination. It can recognize the image more accurately.