基于神经网络的图像识别方法研究

图像识别涉及大量的信息运算,要求处理速度快、识别精度高,神经网络的实时性和容错性要符合图像识别的要求。利用改进的BP神经网络算法对旋转畸变图像进行了定位和识别,改进算法将附加动量项与自适应学习速率相结合,有效地抑制了网络陷入局部极小点,提高了网络的训练速度。结果表明,基于神经网络的图像识别方法是有效的、可行的。

Research on images identification method based on neural network

There are a mass of information computations on images identification and it needs high speed and precision.The real-time and robustness of neural network accord the demands of the images identification.The orientation and identification method of rotation distorted images is described by used of improved BP algorithm,and the algorithm that connects additional momentum with self-adaptive learning rate can prevent network to stack into the minimal value locally and improve the convergent velocity.The experimental results show the images identification method based on neural network is effective and feasible.