Remote Sensing Image Classification Based on AlexNet Network Model

摘要

With the continuous migration and development of remote sensing data application scenarios, the problem of low automation of data and the poor timeliness of information extraction makes it difficult to meet the real-time detection, such as fire monitoring, flying object reconnaissance and so on. With great research, the author select a new method of remote sensing image classification based on AlexNet network model which can shorten training time and improved classification accuracy. The paper used UCM spaceflight remote sensing object detection dataset. Experiments are carried out on the improved network model using the training method of weight transfer. The experimental results show that the improved network model and the training method of weight transfer can improve the classification accuracy by 27.9% without increasing the training time.