Improvement and Evaluation of SPH on High Dimensional Vector for Geographical Origin Identification System for Vegetables by Trace Elements

Recently, as the result of establishment of brand of lobately of farming district, many geographical origin camouflaging cases have being detected in Japan. Therefore, as co-research project with agricultural chemical area, we are developing a vegetable geographical origin identification system by using differences of vegetables´ trace element compositions. If doubtful vegetable is found, its trace element data is compared to data in database. If it overcomes a certain threshold, the system judges that geographical origin of two is the same. This process is not efficient because it requires calculation against all data in databases. Therefore, we developed a method to select calculation target beforehand by employing SPH (similarity preserve hash) before. In this paper, we will describe its improvement and evaluations.