Vegetation Resource Classification on Deep Neural Network and Composite Remote Sensing Image Data

The classification of vegetation resources based on remote sensing images can provide important data support for forest resources survey, grassland monitoring, and wetland vegetation change monitoring. Remote sensing images can quickly and widely obtain the distribution of vegetation resources and its change information, combined with classification algorithms can achieve the extraction of vegetation type and coverage, which provides an effective means for the relevant departments to carry out the survey and monitoring of vegetation resources, and greatly improves the work efficiency. Deep learning methods show high performance as big data mining tools. However, when dealing with remote sensing image tasks, such as the vegetation resource classification problem, it exhibits low accuracy and efficiency. Therefore, a vegetation resource classification method based on deep neural networks and composite remote sensing image data is proposed. Firstly, composite Remote Sensing (RS) image data based on Geographic Information System (GIS) is used to extract relevant attribute data and spatial data of vegetation resources to form a preliminary input image. Following the extraction of several local characteristics from the initial input picture, the deep neural network employed for judgment is fed with these features. The assigned picture labels are used to categorize each local feature. The overall image is adjudicated based on a simple voting method. Finally, classification experiments are conducted using WorldView2 high resolution satellite remote sensing image data, which shows that the proposed method outperforms other classification methods and has better classification accuracy and classification efficiency. © 2024, Taiwan Ubiquitous Information CO LTD. All rights reserved.