Efficient Target Selection in Similarity Preserve Hash for Distributed Geographical Origin Identification System of Vegetables

Abstract

Recent years, especially in Japan, camouflaging geographical origin of agricultural products is a big problem. Therefore, we introduced a distributed system to identify their geographical origin using their differences of trace elements, or very small quantities of elements. Vegetables grown in farms absorb metals form the soil. Since compositions of trace metal elements differ from geographical places, this can be utilized to identify geographical origin of vegetables. In proposing system, trace element compositions of vegetables are measured when they are shipped from a farm, and the data is stored in databases which are located in farming districts. When a doubtful vegetable is found in food distribution channel, its trace element compositions are measured and compared by calculating correlation coefficients to ones accumulated in databases. This system can be used to verify geographical origin data by food traceability system. Because correlation coefficients are not known when they are once calculated, so correlation coefficients between all accumulated data in databases and doubtful vegetable. This means that proposing system is not scalable when the number of accumulated data is increased. Therefore, we introduced a method to reduce the number of target to calculate correlation coefficients using similarity preserve hash (SPH) which gives similar output for similar input. This could reduce time to calculation itself, however, computation time including picking out target data for calculation of correlation coefficients from database. Therefore, we introduce a method to accelerate picking up data form database by grouping value of SPH