Image Semantic Segmentation Based on High-Resolution Networks for Monitoring Agricultural Vegetation

Abstract

In the article, recognition of state of agricultural vegetation from aerial photographs at various spatial resolutions was considered. Proposed approach is based on a semantic segmentation using convolutional neural networks. Two variants of High-Resolution network architecture (HRNet) are described and used. These neural networks were trained and applied to aerial images of agricultural fields. In our experiments, accuracy of four land classes recognition (soil, healthy vegetation, diseased vegetation and other objects) was about 93-94%.