基于边缘检测及RBF神经网络的遥感图像帧特征动态识别技术

Dynamic recognition technology of remote sensing image frame features based on edge detection and RBF neural network

摘要:

为解决分辨率超限问题，实现对遥感图像帧特征对象的精准识别，提出基于边缘检测及RBF神经网络的遥感图像帧特征动态识别技术。求解微分算子与OTSU阈值，并以此为基础，确定边缘节点追踪参数的取值范围，实现对遥感图像边缘检测。根据RBF神经网络机制的构建标准，推导神经性激活函数，完成RBF神经网络识别模型的设计。在所选遥感图像中，实施帧特征分割处理，再联合动态合并条件，计算超像素指标与并行识别参量，完成基于边缘检测及RBF神经网络的遥感图像帧特征动态识别方法的设计。实验结果表明，在边缘检测与RBF神经网络模型的作用下，主机元件在长、宽、高三个方向上对于遥感图像帧特征对象的识别精度都达到了100%，分辨率超限问题得到较好解决，符合精准识别遥感图像特征的实际应用需求。

Abstract:

In order to solve the problem of resolution overrun and realize accurate recognition of remote sensing image frame feature objects, a dynamic recognition technology of remote sensing image frame feature based on edge detection and RBF neural network is proposed. Solve the differential operator and OTSU threshold, and determine the value range of the tracking parameters of the edge node based on this, so as to realize the edge detection of the remote sensing image. According to the construction standard of RBF neural network mechanism, the neural activation function is deduced and the RBF neural network recognition model is designed. In the selected remote sensing image, the frame feature segmentation processing is implemented, and then combined with dynamic merging conditions, the super-pixel index and parallel recognition parameters are calculated, and the design of dynamic recognition method of remote sensing image frame feature based on edge detection and RBF neural network is completed. The experimental results show that under the action of edge detection and RBF neural network model, the recognition accuracy of the host component for the remote sensing image frame feature object in the three directions of length, width and height has reached 100%, and the problem of resolution overrun has been well solved, which meets the practical application requirements of accurate recognition of remote sensing image features.