To extract information of target features from remote sensing imagery and conduct spatial statistics and analysis has always been a key focus in GIS and RS fields. Traditional methods primarily rely on remote sensing multi-classification technology, using various classification approaches to extract specific ground feature types. However, these conventional classification processes present challenges such as high technical barriers, cumbersome data processing, slow recognition speed, and poor model transferability. The advancement of deep learning algorithms has partially addressed these limitations inherent in traditional remote sensing image analysis methods, enhancing processing efficiency and result accuracy, thereby establishing a robust technical foundation for remote sensing image analysis.
Deep learning technology has evolved from machine learning classification and recognition tasks, rapidly emerging as a research hotspot in geoscience and remote sensing in recent years. With its development, deep learning has been successfully applied to multiple aspects of remote sensing science including image preprocessing, pixel-based classification, object recognition, and scene understanding. It effectively resolves the challenge of abstracting low-level primitive features (typically pixels) into meaningful semantic information. Additionally, deep learning applications in image preprocessing and pixel-level classification (particularly considering the cost requirements for large training datasets) have gained widespread adoption.
SuperMap integrates deep learning technology with GIS capabilities, simplifying complex deep learning workflows and reducing technical barriers. This integration provides users with comprehensive imagery analysis functionalities including Object Detection, Binary Classification, Multiple Classification, Scene Classification, Detect Common Change, and Prompt Segmentation.
Current representative application cases include:
- Aircraft model identification through object detection in remote sensing imagery
- Automated land use data recognition and classification
- Urban expansion prediction using satellite imagery combined with land use data
- Automatic updates of spatial data (e.g., POI) through image analysis
Related Content
Machine Learning Environment Configuration
Imagery Analysis Application Examples