Haruyuki Seki

Characterization of Rice Paddies by a UAV-Mounted Miniature Hyperspectral Sensor System

A low-cost, small, lightweight hyperspectral sensor system that can be loaded onto small unmanned autonomous vehicle (UAV) platforms has been developed for the acquisition of aerial hyperspectral data. Safe and easy observation is possible under unstable illumination conditions by using lightweight and autonomous cruising. The hyperspectral sensor system, equipped with a 256-band hyperspectral sensor covering a spectral range from 340-763 nm, a GPS and a data logger, is 400 g in total weight. The acquisition period for each sampling, 768 bytes, is 100 ms. The aerial hyperspectral data of rice paddies are collected under cloudy weather. The flight altitude from the ground is 10 m, and the cruising speed is 2 m/s. The high-accuracy estimation of the chlorophyll densities is confirmed, even under unstable illumination conditions, by frequent monitoring of the illumination level and the chlorophyll indices, based on the red-edge (RE) and near infrared (NIR) spectral ranges.

Development of a Low-Cost, Lightweight Hyperspectral Imaging System Based on a Polygon Mirror and Compact Spectrometers

Low-altitude hyperspectral observation systems using aerial observation with unmanned aerial vehicles (UAVs) have advantages over satellite systems with respect to frequency, accuracy, and spatial resolution. Although low-cost lightweight UAVs have become available in recent years, the current price ranges of lightweight pushbroom and snapshot hyperspectral sensors remain high. For sustainable operation of UAV-mounted hyperspectral sensing, the challenge in production has been shifted from the size and weight to the cost of the lightweight hyperspectral sensors. In this paper, we develop a low-cost, lightweight whiskbroom hyperspectral imaging system. The gross weight of the sensor is 1200 g. The spectral range of the 256-band spectrometer extends from 340 to 750 nm with a 14-nm spectral resolution. The viewing angle across the flight direction is controlled by the rotation of an eight-sided polygon mirror. When the exposure time, flight altitude, flight speed, and focal length of the optical lens are 3.2 ms, 10 m, 10 m/s, and 8 mm, respectively, then the estimated values of the swath and the area coverage per second are 13.4 m and 134.1

Development of a Low-Cost Hyperspectral Whiskbroom Imager Using an Optical Fiber Bundle, a Swing Mirror, and Compact Spectrometers

\hbox{m}^2

Development of hyperspectral imaging system using optical fiber bundle and swing mirror

/s, respectively. The spatial resolution is 0.97 (m) (flight direction)

Development of uav-mounted miniaturure hyperspectral sensor system for agricultural monitoring

\times

Coastal observation using new hyperspectral imager for UAVs

0.46 (m) (scanning direction). In preliminary close-range measurements with a 3.2-ms exposure time per area and a 1224-ms rotation period of the polygon mirror, the reflected light from 12 areas of a printed checkered color pattern in the moving direction were measured. We found that the calculated reflectance based on the measurements is spatially consistent and spectrally accurate.

Measurement of a coastal area by a hyperspectral imager using an optical fiber bundle, a swing mirror and compact spectrometers

Both inductive and deductive approaches based on hyperspectral remote sensing data require abundant observation opportunities under various conditions due to the high dimensionality of the hyperspectral data. With the recent advent of low-cost lightweight unmanned aerial vehicles (UAVs), UAVs for low-altitude aerial observation are becoming commodities rather than special equipment. Therefore, the appearance of low-cost hyperspectral imagers is anticipated for aerial hyperspectral sensing via UAVs. In this paper, we describe the development of a low-cost hyperspectral imager based on a whiskbroom scanning mechanism. The main components of the developed system include an optical fiber bundle, a swing mirror, and compact spectrometers. An image formed by an objective lens is quantized into a set of pixels by a two-dimensional array of quartz fiber-optic cables at one end of an optical fiber bundle. The quantized image travels to the other end of the bundle, inside of which a swing mirror is used for cross-track scanning. The light in each pixel of the quantized image is then measured using a compact spectrometer. Calculated reflectances in close-range measurements of color checkered patterns were spatially and spectrally accurate. In an aerial measurement of a coastal area from a 20-m altitude via a lightweight UAV, a hyperspectral image with a 0.5-m spatial resolution and an 8-m swath was acquired. Based on pattern matching using cross correlation, classification of three classes of marine macrophyte beds, agar, coralline, and sand realized overall accuracies of 0.755 (diffuse dominant illumination) and 0.719 (direct sunlight dominant illumination).

Development of lightweight hyperspectral imaging system for UAV observation

With the recent rise of unmanned aerial vehicles (UAVs), a low-altitude hyperspectral remote sensing is a requisite tool for precise environmental monitoring. However, the low-altitude hyperspectral remote sensing cannot be expanded into a ubiquitous, sustainable monitoring tool because of the high cost of hyperspectral imaging sensors. In this study, we developed a low-cost hyperspectral imaging system for UAVs that is composed of low-cost spectrometers, an optical fiber bundle and a swing mirror. Compared with the system that was developed in our previous work, the spatial resolution of the newly developed sensor was improved by increasing the data acquisition frequency of the spectrometer and the spatial fiber density of the optical fiber bundle.

DEVELOPMENT OF A PORTABLE HUMAN SKIN DETECTOR BASED ON ACTIVE INFRARED ILLUMINATION

Lightweight unmanned autonomous vehicles (UAVs) are becoming practical platforms for aerial data acquisition. We developed a low-cost, small, lightweight hyperspectral sensor system that can be loaded onto small UAV platforms. The hyperspectral sensor system of the total weight 400 g is equipped with a 256-band hyperspectral sensor covering a spectral range from 340-763 nm, a GPS and a data logger. The acquisition period for each sampling is 200 ms. The aerial hyperspectral data of rice paddies acquired from an altitude of 10 m above ground level are collected under cloudy weather. The high-accuracy estimation of the chlorophyll densities is confirmed by the chlorophyll indices based on the red-edge (RE) and near infrared (NIR) spectral ranges, even under unstable illumination conditions.

Human Detection Based on Active Infrared Illumination

Observation of coastal area is important for preserving Earths ecosystem. In this paper, we introduce a customizable, low-cost whiskbroom hyperspectral imager (bands: 288, spectral response range: 340–820, spectral resolution: 15 nm) for UAVs. A newly developed high-performance UAV is capable of landing on the water, so that underwater measurement without sea surface reflection is possible. We investigate the characteristics of remotely sensed hyperspectral images of the east coast of the Izu Oshima, Japan. In an aerial measurement of a coastal area from a 20-m altitude, hyperspectral images with a 0.5-m spatial resolution and an 8-m swath were acquired. Gelidium and sand were classified by pattern matching using cross correlation.