Construction and System Analysis of the Seed Image Database

• Main Authors: Fu-Jin Wei, 魏甫錦
• Other Authors: Kae-Kang Hwu
• Format: Others
• Language: zh-TW
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/90272729932200448856

碩士 === 國立臺灣大學 === 農藝學研究所 === 89 === The identification of seed is among the main tasks of taxonomy study, ecological research as well as seed testing. The common problems encountered in the identification of seed were the training of the personnel, the shortage of reference and atlas of the seeds, and the lack of consistency of the terminology for describing the characteristics of the seed. The information technology has been proved to be an excellent tool for handling repetitive tasks and processing huge information. The application of information technology in the taxonomy can be traced back forty years ago. Several real world applications have been published and put into work. The primary objective of this research is to incorporate the latest development of the information technology including computer image, database system and front-end user interface into a integrated seed image database system to assist the identification of the seed. The file format developed by DELTA (DEscription Language TAxonomy) recognized by the Taxonomy Database Working Group (TDWG) as one of the standard taxonomic data exchange format was chosen as the file format in this research. The full package DELTA system developed by Dallwitz (2000) was selected as the primary editing and database management tool for the construction of the seed image database system. Seed samples of sixty-eight species received from the International Seed Testing Association (ISTA) were used to construct the database. The seed samples provide an ample coverage of the plant kingdom with emphasis on the grasses, 25 species from poaceae were included in the sample. The seed images were taken at least from three different angles of view for each species in the sample for the construction of the image database. The image of the specific morphological characteristics in related reference was taken accordingly. To better assist user in using the system, a diagram of each characteristic was created and shown during the operation of the program in the appropriate time. After the database been constructed, the best separating power of the keys calculated with the INTKEY (the indexing and querying program of the DLETA system) was found to be in the order of length/width, shape and surface characters. The seed coat epidermal derivatives (hair, awn or wing) exhibit similar power as the surface characters. Fifty-three users, who do not have any experience in the identification of seed, evaluated the constructed database along with the query system. The result found on average after three queries, an answer will be reached. The selected key and their order been selected indicated: 1. There was no difference in the number of the each key been queried. 2. the more common characters were more likely to be queried. The result suggested the description of each key needed to be revised to give user better direction. In the assistance of the visual confirmation of the queried results, three dimension images with virtual reality display were built for part of the sample, it appeared to be a good mean in assisting the user making final decision. It should be included in the system.