论文部分内容阅读
针对常用仿射不变矩对轮廓边缘信息敏感的特点,提出了一种基于关键点的局部仿射不变矩,以分割出来的物体灰度图像为基础,首先计算出物体的质心,然后以质心为扩展点向周围引伸出多条射线,寻找每条射线方向上的最近灰度极值点,将所有灰度极值点当作关键点集合,并按照文中提出的计算仿射不变矩的方法提取出多阶不变矩,将其当作神经网络的输入向量,放入已经训练好的神经网络来达到识别物体的目的,将此方法应用在了飞机图像的识别上,实验结果证明此方法在物体轮廓分割不完整和有噪点污染的情况下能保持很好的稳健性,简单有效并具有较广的应用范围.
Aimed at the characteristics of the commonly used affine moment invariants that are sensitive to contour edge information, a local affine invariant moment based on the key points is proposed. Based on the grayscale images of the segmented objects, the centroid of the object is calculated firstly, The centroid extends multiple rays around the extension point to find the nearest grayscale point in each ray direction, treats all the grayscale extreme points as the set of key points, and calculates affine moment invariants Method to extract the multi-moment invariant, which is regarded as the input vector of the neural network and put into the trained neural network to achieve the purpose of identifying objects. The method is applied to the recognition of the aircraft images. The experimental results show that This method can maintain good robustness in the case of incomplete object segmentation and noise pollution, simple and effective and has a wide range of applications.