Automatic Detection of Environmental Change in Transmission Channel Based on Satellite Remote Sensing and Deep Learning

The detection of environmental change in transmission channel based on high-resolution satellite remote sensing is the emerging trend in the operation and maintenance of power grid. To improve the precision, a change detection method based on the deep learning was put forward in this paper. Firstly, by the improved change vector analysis algorithm and grey level co-occurrence matrix considering the neighborhood information, the spectral and texture changes of images at different times were calculated. Moreover, the most probably changed and unchanged zone samples were extracted through setting the adaptive sampling interval. Additionally, the modified deep convolutional neural network model Inception-v3 was constructed and trained. The in-depth features in changed and unchanged zone were extracted, and the changed zone was distinguished effectively. The results showed that the proposed method could effectively extract the environmental change in transmission channel, and the accuracy reached over 92.5%, which was obviously superior to the contrast methods.