Hyperspectral data have tens to hundreds bands and finer or better spectral, spatial, and radiometric resolution than multispectral data and are hence widely used in land covers researches in recent years. Reducing the dimensionality and keeping the information of hyperspectral data become an important works in applying such great volume and spectral homogeneously information of remote sensing data. This paper uses the DAIS-3715 data, a product of hyperspectral remote sensing program conducted by Committee of Agriculture of ROC in 1997, to classify 15 types of land-cover of seashore area of southern Taiwan. It aims to study the efficiency of proposed statistical based hybrid criterion, PN, in reducing the data dimensionality or selecting heterogeneously spectral bands for seashore land-cover classification. Results showed that PN criterion is excellent in selecting heterogeneous spectral bands of DAIS-3715 data for classifying the seashore land-cover. PN method can choose effectively only four bands, de­noted as band no. 24, 18, 17, and 5 which correspondingly records the near infrared signals of 0.96, 0.84, 0.82 micrometer and visible signal of 0.59 micrometer, to achieve a good accuracy and which is almost identical to the one of the classification by using all of 24 bands. PN method saved about 7.5 times of the computer time of land uses classification. The accuracy assessments for all of the classification with different band combinations that decided by each band selection criterion are assessed from the same sets of training and assessing data. Overall accuracy (OA) and kappa_hat accuracy ( ) of the classification using only four bands selected by PN criterion for both training and assessing data are 95% and 90% respectively. Although, the accuracy of the PN method based classification is less about 3% than the ones of all 24 bands based classification for the training data while it is more about 3% for the assessing data.

In this study, GIS, digital terrain model (DTM), global positioning system (GPS), and digital surveying using total station theodolite were applied to revise the protection forests. The protection forest management system was developed using Visual Basic and ArcView Avenue programming languages, which can provide useful spatial information to facilitate decision making for protection forest management. The results indicate that: (1) Digital surveying technique has advantages of high efficiency, high accuracy, and automation, as compared to traditional surveying method. The technique was found very effective in identifying the boundary of protection forests, and surveying areas that were not used as their original objectives. (2) The protection forest management system developed in this study can integrate spatial information and attribute data, and provide valuable information for protection forest management, which may let the protection forests function well and meet their designated objectives such as protecting soils, human lives and properties, and maintaining high environmental quality.

Forestland classification is a prerequisite task of ecosystem management. The operational scheme needs to meet the ecological principles and the management requirements of different forestry units. The objectives of this study were to establish a decision support system of multi-scale forestland classification for the Liukuei Experimental Forest, and also to investigate the effect of multi-scale ecosystem units on forestland suitability. The processes included the development of criteria for forestland classification; the establishment of databases and knowledge bases; and the establishment of a decision support system of forestland classification based on the Ecosystem Management Decision Support System (EMDS) developed by US Forest Service. Meanwhile, to investigate the effect of multi-scale ecosystem units on forestland suitability, different watersheds representing different sizes of ecosystems and multi-scale ecosystems (e.g. site, landscape mosaic, and ecoregion) were obtained by the automatic extraction of digital terrain model and multivariate statistical analysis respectively. The result indicates that the developed decision support system can easily analyze and evaluate forestland suitability by adjusting the assessment areas, the variables, and the criteria. It can also adapt to the management requirements of different forestry units, and integrate forestland classification and management requirement together. Therefore, the decision support system of multi-scale forestland classification developed in this study is not only feasible for the forestland planning and ecosystem management of the Liukuei Experimental Forest, but also for the references of island-wide domestic forestland classification and ecosystem management.

Hyperspectral images provide richer and finer spectral information than traditional multispectral images however the volume of generated data is dramatically increased. Data compression will be essential for economical distribution when spaceborn hyperspectral data are regularly available. Contemporary tech- niques of image compression are mainly designed to explore the redundant information in the 2-D image space. It is awkward when they are applied to hyperspectral data. In this study, a 3-D wavelet transform is proposed to explore useful information in the spatial and spectral dimension simultaneously. The procedure proposed for hyperspectral image compression is mainly divided into three steps: 3-D wavelet transform, optimal quantization and Huffman coding. Various combinations of wavelet functions decomposition levels and quantization intervals were experimented with an AVIRIS image. The decompressed images are evalu- ated objectively and subjectively， based on signal-to-noise ratio (SNR) and classification accuracy measures, respectively. The results indicate that 3-D wavelet transform is efficient for hyperspectral image compres-sion and preserves the capability of the classification applications.

Synthetic Aperture Radar ( SAR ) that operates independently of time and weather condition is capable of recording rich texture information and is becoming one of important tools of remote sensing techniques. The current space-borne SAR system can prove only single-band single-polarization image, which provides less band image than space-borne optical images. Then extracting information in the SAR image has to be carried out in a specific way corresponding to the character of the SAR image. The methods of Dyadic Decomposition wavelet or Overcomplete Decomposition wavelet have been applied to the SAR image in order to produce more images with different features and resolutions for automating information extraction by using classification techniques. Image size is reduced to a quarter of an original image while Dyadic Decomposition wavelet transformation has been applied. In order to extract information automati­cally on the basis of the size of the original image, resampling process is used in this research. However, the resampling process does not need while Overcomplete Decomposition wavelet is applied. An air-bone SAR image has been used in this research and shows that the accuracy of image classification is 77.35% when using Dyadic Decomposition wavelet. But the accuracy of image classification can be promoted to 85.7% when Overcomplete Decomposition wavelet is applied. The efficiency of automatically extracting information in a SAR image on the basis of texture features produced using wavelet-based multi-resolution analysis has been shown.