Abstract
This study aimed to develop a wide-area mineral identification system for placer deposits using a multispectral camera mounted on a drone. The system is expected to provide rapid and accurate identification of the mineral sands found in the placer deposits, used in various products essential to our daily lives. Unlike other types of mineral deposits, beach placer deposits are constantly changing due to the water flow in oceans, making it difficult to manage production volume and quality. Moreover, exploration projects located near coastlines or urban areas are environmentally conscious. In recent years, remote sensing using a multispectral cameras mounted on drone has become more prevalent and already being utilized in a wide range of fields.
In this study, the system was attempted to be constructed to identify target minerals with spectral bands, using AI analysis and a drone equipped with a multispectral camera. The beach placer deposit used for the experiment was located on the Kunisaki Peninsula in Oita Prefecture, Japan. This beach placer deposit is mainly composed of magnetite, so we attempted to verify the ability of the hyperspectral camera and multispectral camera mounted on the drone to identify magnetite.
Firstly, verification was conducted under ideal conditions using halogen lights to determine whether the identification was possible. Samples of magnetite and the beach placer deposit were photographed using multispectral camera, and identification was performed using deep learning. The results confirmed that identification was possible. Next, an attempt was made to determine whether the identification of magnetite was possible in the actual field by analyzing the sample collected from the site and multispectral photographs taken by the drone.
As a result, the developed system has the potential to be an efficient tool for mineral exploration and resource evaluation in placer deposits. It was confirmed that magnetite can be effectively identified and discriminated from beach placer deposits using this method, and the effectiveness of the multispectral camera mounted on the drone was demonstrated.