Abstract:The paper puts forward data frame of the remote sensing knowledge base. Oriented-object method is adopted to organize data of the knowledge base. The typical land surface objects, which are classified into vegetation, soil, water, snow, minerals, man-made object, are the center to be used to link measured data at field, prior spatial data, expert system data, images and models.The knowledge base consists of five aspects: 1)The database for measured spectrum and relevant environmental factors.The database consists of 6 kinds of data.The first is the measured data, which are made up of canopy spectrum, soil physical-chemical data (N, P, K, humidity etc), plant bio-chemical data (chlorophyll, nitrogen etc), vegetation structure (LAI, LAD etc) and biomass data, component (leaf, stalk etc) spectrum data, and data about microclimate and weather. The second is the instrument information. The third is data of campaign site condition. The fourth is the documents about observation methods and measured data processing methods. The fifth is the documents about surveyors. And the sixth is the metadata.2)The spatial database for landform,land-use and land-cover etc.3)The expert system for vegetation seasonal changes and spatial patterns.4)The remote sensing image base, which are acquired by field, aerial and space approaches such as multi-angle, multi-temporal, multi-spatial resolution, multi-spectral resolution, multi-platform etc. And 5) the remote sensing physical model base. The knowledge base can not only retrieve spectrum and environmental factors of typical land surface objects, but also simulate the spectrum of remote sensing pixel when the measured spectrum is unavailable. The model system consists of the analytical models and computer simulations for remote sensing mechanism. The spatial scale of the models is from the material (such as leaf, soil) to end-member (such as canopy) to remote sensing pixel (such as a pixel consisted of wheat, road and lake).The user enters object/place, time, observing direction, bands and spatial resolution,and the knowledge base provides the reference spectrum and relevant description. The spectrum of vegetation depends on the seasons, and it is impossible to measure the spectrum of vegetation at all times. So the vegetation parameters have to be extended temporally when the measured data is unavailable.Based on the knowledge of the vegetation expert system, the vegetation structure parameters and spatial pattern can be computed by the vegetation shaping models. The land surface parameters have to be scaled up/down when the pixel size is different to the size of view field of the measured data. The describing parameters of DTM and land-cover and soil should be based on the pixel size. Selecting remote sensing physical models for simulation depends on three factors:1)spectral band: visible light near infrared, thermal infrared, and microwave;2)spatial scale on which the models are built;and 3)land surface type, grass or forest. In other words, radiation transfer model, geometric optics model or compositive scene model is more suitable to this type. The agriculture remote sensing applications based on the spectral knowledge base also is discussed. The approach is a feasible quantitative remote sensing way in agriculture.

Spectral Derivative Features For Supervised Classification Of Remote Sensing Data: An Experimental Evaluation

Spectral Derivative Features for Classification of Hyperspectral Remote Sensing Images: Experimental Evaluation

On the Sampling Strategy for Evaluation of Spectral-Spatial Methods in Hyperspectral Image Classification

On the Sampling Strategies for Evaluation of Joint Spectral-Spatial Information Based Classifiers

An Improved Spectral Reflectance And Derivative Feature Fusion For Hyperspectral Image Classification

Scalable Semi-Supervised Classification Of Hyperspectral Remote Sensing Data With Spectral And Spatial Information

Deep Learning Spatial-Spectral Classification of Remote Sensing Images by Applying Morphology-Based Differential Extinction Profile (DEP)

Deep Learning Classification by ResNet-18 Based on the Real Spectral Dataset from Multispectral Remote Sensing Images

Image classification of hyperspectral remote sensing using semi-supervised learning algorithm

Merging Spectral and Textural Information for Classifying Remote Sensing Images

Influence of Different Optimizers on Classification Results in Deep Learning

Classification of Hyperspectral Imagery Based on Spectral Gradient, SVM and Spatial Random Forest

REMOTE SENSING IMAGE CLASSIFICATION WITH THE SEN12MS DATASET

Recent Advances in Domain Adaptation for the Classification of Remote Sensing Data

Semi-Supervised Subcategory Centroid Alignment-Based Scene Classification for High-Resolution Remote Sensing Images

An automatically recursive feature elimination method based on threshold decision in random forest classification

Comparison and Evaluation of Classification Features in Land Cover Based on Remote Sensing

Extraction of Hyper-Spectral Remote Sensing Alteration Information Based on SAM and SVM

Supervised Classification Performance of Multispectral Images

Adoption of Machine Learning in Intelligent Terrain Classification of Hyperspectral Remote Sensing Images

Data Frame and Spectral Simulation for Remote Sensing Spectral Data Base