As the problems caused by the global ecological environment become more and more serious, environmental protection has become an important issue in countries all over the world. In order to prevent natural disasters and monitor natural conditions, including monitoring areas affected by deforestation and floods, satellite image segmentation and automatic monitoring of forests and water bodies are becoming more and more important, which leads to the research work of satellite image segmentation in this paper. This paper is a continuation of the previous work of Fw-U-Net, and the model architecture proposed in the previous work still has many shortcomings, such as the design of the basic network architecture and the setting of parameters are not optimal, resulting in the performance cannot be further improved, which means that there are more directions that can be more improved. In the work of this paper, we make more in-depth improvements on Fw-U-Net. Pay attention to the realization of a lot of details and the exploration of the essence. It has been verified that the segmentation verification scores of our upgraded model in forest cover area and water area performance test set are 87.48% (84.51%) and 90.7% (85.83%) respectively, compared with previous work. Increased by 3.11% (0.86%) and 2.71% (0.41%), respectively, and the loss value was reduced to a certain extent. Transfer learning has higher accuracy and reference value for forest and water satellite image segmentation.
As global environmental problems are becoming more and more serious, satellite image segmentation and automatic monitoring of forests and water bodies are becoming more and more important in order to prevent natural disasters and monitor natural conditions, including monitoring deforestation and areas affected by floods. In this paper, the traditional U-Net model is improved and a FW-U-Net model suitable for forest water segmentation is proposed. After verification, the segmentation verification scores of the forest coverage area and water area performance test set are 84.37% (83.65%) and 87.92% (85.42%) respectively, which has high accuracy and reference value for forest and water satellite image segmentation.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.