ROTATION OF SPECTRAL CHANNEL AXES. APPLICATION TO A RICE GROWING AREA OF WEST AFRICA
In the framework of a research on rice production forecast in West Africa through remote sensing methods, simulated SPOT data have been acquired by two air flights, at an interval of 20 days, shortly before harvest, near Mopti (Mali). This zone, where rice growing is entirely dependent on Niger flooding conditions is characterized by great spatial variability: fields are small and closely linked to geomorphological conditions. A detailed and precise ground truth map could be obtained through ground observations and simultaneous I.R. colour air photography. A spectral data analysis, based on a two-step axis rotation in the feature space, is presented. First, the radiance of every pixel is considered to be the vectorial sum of three components : soil, water, vegetation. Three derived channels are thus obtained, of which the first evidences dry soil details, the second maps wetness and the third is a vegetation index similar to the (3-2)/(3+2) biomass index. Then, in order to distinguish the vegetable species rice, Bourgon, Nenuphar and Vetiver grass one studies, in detail, the distribution of pixels in the channel space, and especially their vectorial directions by means of a stereographic plan. It is shown that by a judicial choice of the axes based on that study, it is possible to obtain, by rotation, images in which the vegetable species are well distinguished. The classification results obtained by the above method indicate a good concordance with the information area, especially concerning the rice. Besides the study of th vectorial directions facilitates the comparison of the facts on both occasions (dates) and their comparison with the radiometric measurements taken on the ground.
Bibliographic Reference: 18EME SYMPOSIUM INTERNATIONAL SUR L'OBSERVATION DE LA TERRE, PARIS (FRANCE), OCT. 1-5, 1984 WRITE TO CEC LUXEMBOURG, DG XIII/A2, POB 1907 MENTIONING PAPER E 31729 ORA
Availability: Can be ordered online
Record Number: 1989123102100 / Last updated on: 1987-01-01
Available languages: en