图像目标的矩特征量的照度不变性

在运动目标识别中,需要提取与目标远近、方位,照度、投影等变化不敏感的特征量。不变矩组对于目标平移、旋转和尺度变化不敏感,因此可以用不变矩组外加一些辅助特征来进行运动目标识别.本文研究了数字图像目标的不变矩组特征量经适当修改后,可以使其对目标平移、旋转和环境中的光照强度不敏感.这是对矩不变理论在模式识别应用中的一个重要补充,文章最后给出了修改前后两组公式的适用范围和实验结果。

The Light Illumination Invariance in the Moment Features of Image Objects

In moving object recognition the features extracted are those insensitive to variations in the distance, position, illumination and projection of the moving ob-ject. The set of moments is insensitive or invariant to translation, rotation and scale change. So some supplementary features can be added to the set of moments to recognize moving object. Proves by theoretical deduction and experiment that the moment feature can be made robust and insensitive to translation, rotation and the illumination of environment with proper modifications. This is an important addition to the moment invariance theory applied to pattern recognition. In the last portion of the paper the proper range and experimental results are given to one set of formulas before modification and the other to one after modification respectively.