论文部分内容阅读
为提高阴影检测精度,提出一种新的遥感影像阴影检测方法—将径向基函数神经网络构建的混合模型(称作SMM-RBFNN)应用于遥感影像阴影检测。灰度共生矩阵中的能量、熵、对比度和逆差矩4种统计特征量作为混合模型的输入特征矢量,采用类“期望-最大化”算法(类EM)进行参数估计,训练检测器实现阴影检测。对多幅带有浓厚阴影的遥感影像进行实验,结果表明所提出的方法明显优于传统的高斯背景法和直方图阈值法,能够较好地解决强反射性地物漏检和水体错检问题,能够克服基于阈值思想的检测法需要反复实验选取阈值的缺点。
In order to improve the shading detection accuracy, a new shadow detection method for remote sensing images is proposed. The hybrid model constructed by radial basis function neural network (called SMM-RBFNN) is applied to the shadow detection of remote sensing images. The four statistical features of energy, entropy, contrast and inverse moment in the gray level co-occurrence matrix are regarded as the input eigenvectors of the hybrid model, and the parameters are estimated using the class “expectation-maximization” algorithm (EM) to train the detector Shadow detection. Experiments on several remote sensing images with strong shadows show that the proposed method is obviously superior to the traditional Gaussian background and histogram threshold methods and can well solve the problem of missing objects and water bodies misjudgment , Can overcome the shortcomings that the detection method based on the threshold idea requires repeated experiments to select the threshold value.