基于多方向角Gabor与RBF神经网络结合的目标识别算法

Target Recognition Algorithm Based on Multi-direction Angle Gabor and RBF Neural Network

针对导弹末端制导过程中CCD摄像头抓捕目标的背景、光照及相对旋转角度变化和噪声等因素造成目标识别率降低问题,在结合多方向角Gabor滤波器和RBF神经网络的基础上,提出了一种基于特征匹配的目标识别算法。该算法采用Gabor滤波器对待匹配图像进行预处理,通过将多个不同方向角Gabor滤波器的结果进行叠加和归一化处理的方法,突出了目标轮廓特征,然后提取纹理图像的4类Haar-like特征,再利用训练完成的RBF网络模型进行识别。实验结果表明,算法在保证实时性的基础上提高了目标识别率。

Aiming at the problem that the factors of target＇s background,illumination,and the relative rotation angle changes and noise result in the reduction of target recognition rate in the missile terminal guidance process of CCD capture,in combination with the multi-direction Gabor filter and RBF neural network,a target recognition algorithm based on feature matching is proposed.The algorithm uses Gabor filter to deal with the matching image,superimposed and normalized processing through the results of the multiple direction Gabor filter method,highlighting the target contour features,and then extract the image texture of 4types of Haar-like features,and use the trained RBF network model for recognition.Experimental results show that the algorithm can improve the recognition rate of target based on the guarantee of real time.