A cultivated crop type monitoring system based on progressive spatio-temporal spectral representation learning
By using a progressive spatiotemporal spectral representation learning system, the spectral, temporal, and spatial characteristics of cultivated crops are decoupled in stages, solving the problem that spectral and phenological changes are difficult to capture in existing technologies, and achieving more refined and stable identification of cultivated crop types.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SUZHOU IND PARK SURVEYING MAPPING & GEOINFORMATION CO LTD
- Filing Date
- 2026-03-27
- Publication Date
- 2026-07-17
AI Technical Summary
Existing deep learning models struggle to effectively capture spectral and phenological variations in arable land crops, leading to decreased classification accuracy. Furthermore, conventional convolutional networks struggle to simultaneously capture local textures and global structures, impacting the accuracy and reliability of arable land crop type identification.
A progressive spatiotemporal spectral representation learning system is adopted to decouple the spectral, temporal, and spatial features of multi-temporal multispectral remote sensing images in stages. Feature extraction is performed through a spectral sensing module, a temporal sensing module, and a spatial modeling module. Techniques such as NDVI, CNN encoding, Transformer blocks, and positional encoding are used to improve the quality of feature extraction.
It improved the classification accuracy of crop types in cultivated land and the quality of field-level identification, significantly enhancing the accuracy and reliability of crop type identification in cultivated land.
Smart Images

Figure CN122416243A_ABST