This paper investigates the conditions for ridge lines and valley lines to be detected as edges in the LANDSAT-MSS imagery with respect to their directions, their side slope angles, the sun azimuth and evelation. The test site is specified around Mt. Goyo in Iwate Prefecture, Japan which has a typical ups and downs topography. First, by using the LANDSAT data of three seasons and 159 standard samples of ridge lines and valley lines, the detection rates by five kinds of edge detection operators were evaluated. The results via the average-difference operator (AV-DIF) applied to the band-7 images demonstrated the highest marks. Then by introducing a model of illuminated side slopes by the sunlight, an index of BD is defined as the difference of cosy values of the side slopes where γ means the angle between the normal vector of a slope and the vector to the sun. The critical values for the ridge lines to be detected by AV-DIF from the band-7 images were found to be around 0.3. The same approach was also applied to the pure cosy images, which were synthesized from the digital elevation model and the sun position at the image collection. Then almost equal results to these of the real imageries were observed. As BD depends on the sun position, the detection efficiency for each sample varies temporally. In the test area, the BD values of the ridge lines and the valley lines of which directions are near to the southeast stay in low levels for all seasons even though they have rather steep side slopes.
Experiments have been conducted to determine optimum threshold for parametric Maximum Likelihood Classifier (MLC). Optimum threshold indicates different results for Thematic Mapper (TM) and MSS data. This may due to the fact that the TM spatial resolution is 2.7 times finer than MSS, and consequently, TM imagery has more spectral variability for a class. The increase of the spectral heterogeneity in a class and the higher number of channels being used in the classification process may play significant role. For example, the optimum threshold for classifying an agricultural scene using MSS data is about 2.5 standard deviations, while that for TM corresponds to more than four standard deviations. This paper compares the optimum threshold between MSS and TM, and suggests a method of using unassigned boundary pixels to determine the optimum threshold. Further, it describes the relationship of the optimum threshold threshold to the class variance with a full illustration of LANDSAT data. The experimental conclusions suggest to the user some systematic methods for obtaining an optimal classification with Maximum Likelihood Classifier.
Suitability of two band false color images for visual interpretation have already been investigated using MSS data. However, it is necessary to reinvestigate two band false color images, because TM data have severl bands lacked in MSS data. Experiments are made in two steps as follows. Experiment (1) with scene 1 : On the base of the previous result on three band false color images, several kinds of two band false color images for sea or land area are prepared respectively and good band combinations and color sequences are selected from them. Experiment (2) with scene 2 : On the base of the experiment (1) on the land area, two band false color images of all band combinations to one of the good color sequences (21 kinds ) are made, and the best band. combination is selected. Moreover, these are compared with typical three band false color images on every objects. On the sea area, it is found that the best band combination is (2-6) and the best color sequence is (BG-R) collectively. On the land area, in case of band combinations excluding band 6, the best band combination is (3-5) on the large scale objects and (3-4) on the small scale objects. In case of band combinations including band 6, the best band combination is (7-6). And the best color sequence is (R-BG) in all cases. Undoubtedly quantity of information keeped in false color images with two bands is less than three bands, but there are enough cases of two band false color images for visual interpretation. Such cases are especially found in band combinations including bands lacked in MSS data. Because two band false color have advantages in some points such as clearness of images and simplicity in imaging method, it is concluded that two band false color images are very useful for some objects, especially water area.
If we could have access to an optical disk as we have access to a floppy disk or a hard disk, large scale multispectral data analysis will be done on a personal computer system. The authors have studied and developed a personal computer system with optical disk drive system as the external memory which enables us to analyze such huge data, which could be only analyzed by a large computer or special data analysis system so far. Using this system, Landsat TM data for one scene about 300 Moytes can be written upon a side of a five quater inch optical disk. The written data are read out with random access from the optical disk and processed by a personal computer. This point is the most remarkable in this newly developed system. Further, to serve the system to general MSS data analysis, we connect the software with the formerly developed Remote-10 by which we can perform the multi-spectral data categorization very fast. The design concept of the system involves large scale data handling, fast processing, and full color graphics presentation utilizing easily a personal computer. This system may give a suggestion to the future data delivery system of Landsat, Spot, Noaa satellite data, and also to the future data analysis system also. This system may become popular for an educational use at a laboratory level, because of the easy data handling, the easy data accessibility, and the reasonable system cost.
The experimental trial for the direct extraction of the change of building construction using Landsat-MSS data was conducted for the purpose of the evaluation of the impact by high-way construction. The regression analyses which use Landsat-MSS data as independent variables and the transformed numbers of buildings by Logistic Curve as dependent variables were conducted at total 30 cases of the selection of variables. As the results of the analyses, all cases showed relatively low correlations in which the highest value was 0.369. These results suggest that it is rather difficult to extract the change of building construction directly by Landsat-MSS data and that further analyses using Landsat-TM data and the more urbanized test areas are necessary for the completion of the purpose of this study.