基于小波矩特征和神经网络的灰度图像识别算法研究

Study on Image Recognition Based on the Wavelet Moment and Neural Network

小波矩结合了矩特征和小波特征,既反映了图像的全局性信息,又反映了图像的局域性信息,并且具有旋转、平移和缩放不变性。文章研究了基于小波矩的灰度图像特征提取算法,并将其与BP神经网络组合,利用神经网络的强学习能力和容错性,形成一个灰度图像识别系统。仿真实验表明,在图像失真较小的情况下识别率可达到100%,较之未提取特征的神经网络识别方法而言,网络收敛速度与识别精度都有较大的提高。

The wavelet moment which combines the moment features with the wavelet features, reflects both the global and local information simultaneously and is invariant to translation, scaling and rotation. In this paper, a methord of the feature extraction on gray-image based on wavelet moment is studied, which is combined with the BP artificial neural network, in virtue of the strong ability in study and tolerance of error of the network, a recognition system on gray-image is build. It is proved that the rate of recognition can reach 100%, both the speed of convergence and the method of image recognition based on neural network precision of recognation enhanced compared with the which does not have feature extraction.