Taiwan Provincial Food Department utilizes aerial photo interpretation for rice crop inventory each year to calculate the areas. If an automated classification method can be developed, the amount of time, manpower, and resources needed in the current work can be reduced. Meanwhile, the errors caused by human subjective interpretation can be avoided. This research uses artificial neural network, which simulates human neuron and fault-tolerance for classification. In this study, error back-propagation (BP) and learning vector quantization (LVQ) neural network algorithms are selected. Meanwhile, two data coding techniques are applied for data representation to input network model. The data used in the experiment are multi-temporal SPOT images and multi-temporal NDVI images of Changhua area. All the classification results are com- pared with those produced by Gaussian maximum likelihood algorithm. Finally, the contribution of texture images for classification are studied. In general, the experiments reveal that neural network approaches are better than maximum likelihood classification. Especially BP, and LVQ is thesecond­best.

In this research, the Hoyenshan Nature Reserve at Sanyi was selected as the study site. This research utilized aerial photos obtained in four different years to investigate the soil erosion in the study area. Specific procedures of the research include: (1) to obtain coordinates of ground control points applying GPS surveying; (2) to collect data of topographic features using digital photo-grammetry; (3) to establish DTMs applying automatic matching techniques. Finally, by analyzing those DTMs, we quantify the soil erosion and establish a model for evaluating the geomorphological change within the study area. When applying automatic matching technique in collecting DTM points the quality and percentage of overlap of the aerial photos usually affect the results. To obtain better results, manual digitization of break lines is required for those areas that have poor results in previous runs, then the interpolation is performed to generate regular DTM points. Automatic matching is based on the gray values of digital images therefore, digital surface model (DSM) is obtained through the process. The results indicate that DSM can be used to derive drainage system and watersheds, the geomorphological change can then be e valuated by com­paring the drainage systems and watersheds of different years. To estimate the volume of soil loss, however, the areas covered by forest should be excluded so that the computed volume is not affected by the tree growth. The geomorphological change of the study area is affected by many factors. More-over, the fre­quency, intensity, and locations are varying. Therefore, it is very difficult to model the geomor­phological change. However, based on the volume and rate of soil loss, it is estimated that the unique geomorphologic features of the study area should still remain for a long period of time.

For high-precision GPS satellite positioning, Carrier phase observables must be processed. There are two algorithms to. follows. One is to determine the correct phase ambiguity param­eters, the other uses the trigonometric functions to eliminate the integer ambiguity parameters. The purpose of this paper is to establish a set of model, namely cosine function models of carrier phases. This paper begins with cosine function models of carrier phases, and compares the differences between the method and phase ambiguity resolution. A detailed discussion concerning the characteristics of cosine function models of phases is offered, it can be applied to satellite leveling with proof of fulfilling the accuracy requirement of second-order class II leveling. Be- sides, this model helps us to estimate the path delays of tropospheric as well as ionospheric alto­gether. Kinematic positioning can be applied with this model with such an achievement equiva­lent to that of ambiguity resolution. The last part of this paper contributes to designing the cosine function models of phases with muti-refference station.

Technique of thermal infrared imagery was applied to monitor seasonal changes in canopy temperature of pangolagrass (Digitaria decumbens Stent.) grown in the experimental field at Hsinchu branch station of Taiwan Livestock Research Institute. Fluctuation of canopy radiometric temperatures in response to ambient air temper-ature and soil surface temperature remotely measured from a portable thermal inf-rared imaging radiometer was also studied. Results indicated that seasonal changes in temperature of pangolagrass vegetative can­opy were influenced by both growth and environmental effects. Generally can­opy temperature was lower than air and soil temperatures, except in the period imm ediately after cutting where soil background effect was strong and canopy was not full covered. Temperature distribution of can­opy depended on the uni­formity of vegetative cover as well as soil temperature, air temperature, and climatic condition. The strong reflected solar radiation over soil plus the emitted longwave radiation from soil resulted in an over-estimated soil radio-metric temperature against soil dynamic temperature near noon. However, no close rel­ationship was found between growth characters (plant fresh weight and dry we­ight) and canopy temperature, yet plant weight and plant height was closely related. Re-sults suggest that the ability of temperature adjustment enables plants of pangolagrass tolerating to less favorable environmental conditions.