基于模式识别技术的高光谱图像分类研究

Research on hyperspectral image classification based on pattern recognition technology

针对当前高光谱图像分类方法图像特征向量应用环节设定较为落后,导致分类精度差,无法获取完整图像分类结果的问题,提出基于模式识别技术的高光谱图像分类方法。采用加权平均法,完成高光谱图像预处理,使用多尺寸局部二值法,提取高光谱图像特征向量,确定不同类型光谱信息的联合分布密度,结合模式识别技术,完成高光谱图像分类。通过总体图像分类精度、平均图像分类精度以及kappa系数对此方法的应用效果进行评估,结果表明,本方法精度较高均在96%以上,且kappa系数超过0.9,可缓解当前方法在应用过程中出现的问题。

Aiming at the problem that the application link of image feature vector in the current hyperspectral image classification method is relatively backward,resulting in poor accuracy of classification and unable to obtain complete image classification results,a hyperspectral image classification method based on pattern recognition technology is proposed.The weighted average method is used to complete the preprocessing of hyperspectral images.Multi size local binary method is used to extract the feature vector of hyperspectral image.Determine the joint distribution density of different types of spectral information,and complete hyperspectral image classification combined with pattern recognition technology.The application effect of this method is evaluated through the overall image classification accuracy,average image classification accuracy and kappa coefficient.The result shoe that the accuracy of this method is higher than 96%,and the kappa coefficient is more than 0.9,which can alleviate the problems in the application of the current method.