In this paper, we propose a method to apply a competitive neural network to land cover mapping of remote sensing data. This neural network is trained by the Learning Vector Quantization (LVQ) method. For the neural network, several weight vectors of neurons in the competitive layer represent a category. We employ LVQ1 and OLVQ1 in learning algorithms of LVQ. OLVQ1 is introduced in order to obtain high speed convergence compared with LVQ1. The three types of pattern distance functions and number of neurons are considered to find an optimal LVQ method. To evaluate the classification accuracy, the LVQ1 and OLVQ1, Maximum Likelihood Method, Back Propagation Method (3 layered neural network) and Nearest Neighbor method are considered. We classify LANDSAT TM data using TR1 or TR2 type training data where TR1 includes the same number of training data for each category, and TR2 is produced by picking up pixels on a grid in a certain processing area. As a result of this experiment, the OLVQ1 using the Mahalanobis' generalized distance and TR2 outperform the other methods with respect to overall accuracy despite a very small number of neurons, for example 24. The LVQ1 and OLVQ1 using TR1 produce better classification results regarding with average accuracy compared to the other methods. This method produces excellent classification images which are more realistic and noiseless compared with the conventional methods.
Once grassland productivity decreases, grass-renovation, accompanied by plowing and seeding, needs to be conducted to improve its productivity. Years after grass-renovation is one index indicating grassland productivity. In this study, the monitoring of grass-renovation status using satellite data is discussed. By extracting 1-year grassland and renovated grassland from each satellite image, we compiled a Grass-renovation years map (1985-1994) . Renovation years from satellite data agreed with that from ground survey with high accuracy, showing that satellite remote sensing is very useful for survey of grass-renovation status.
Satellite remote sensing provides wide-ranging and temporal information on grasslands, which are indispensable for proper grassland management. In this study the satellite monitoring of annual changes of grasslands is discussed. By applying a new method which utilizes a Grassage Standard Map to a study area in Konsen Plain, Hokkaido, Japan, increases in reflectance according to the aging of grasslands were identified from the data observed in May. The increases were considered to be a result of dead leaf accumulation on the ground surface of grasslands. In the June data, sudden decreases of reflectance in the near infrared and middle-infrared bands were identified within the 4 years period after the grass-renovation. The decreases may be a result of changes in grass biomass and grass species composition. The results demonstrate the significance of the new method for grassland monitoring.
A new approach of unmixing with the subspace method is proposed and an experiment using hyperspectral image was conducted. In stead of using conventional statistical unmixing procedures which incorporate all channel data to perform unmixing, the proposed approach assigns subspace for each unmixing class. In this method, unmixing is calculated by the projection of observed pixel vector on the class subspaces. This method is more stable than conventional methods against noises and works effectively as a feature extraction and data reduction procedure at the same time. Owing to these advantages, this approach is suitable for the unmixing of hyper spectral image which has high correlation between channels. The performance of this method is tested by an experimented using a hyper spectral airborne casi image acquired over a wetland area. Unmixing of 7 wetland vegetation classes were calculated using least square, quadratic programming, orthogonal subspace projection and subspace method. Finally, the results of unmixing experiment were compared and evaluated for the use of wetland vegetation monitoring.
In this paper, a PC based SAR processor to produce SAR images from space borne SAR row data is introduced. It is quite practical doing this work, because process flexibility like changing reference function or multi-look number for a specific image analyses is important while data centers can not respond to such requirement from various data users. The PC based SAR processor have the flexibility to answer these requirements. Program and generated file size along with processing speed is described to show the practicality of the developed programs. Some application examples like interferometry or differential interferometry using the program are also introduced.