The study on multi sensors observing for ground truth using vehicle
3. Methods
3.1 Multi sensors system
We put on the arm on the car-top and the tip of the arm has multi sensor system . The height of the sensors from ground approximately 2m. This system includes Digital Still Camera (DSC,KONIKA Corp.), CCD Camera (SONY CCD-MC-1) with Digital Video Recorder (SONY DCR-TRV9), GPS Camera (KONICA Corp.), 3D Scanner (PULSTEC INDUSTRIAL CO., LTD TDS3100, 680nm laser scan), thermometer (HORIBA,Ltd., IT-540S), GPS (NovAtel Inc., Single-GPS) and spectometer (Soma Optics Inc). Spectrometer can measures in visible/near infrared between 350nm and 1050nm (measure with 512 ch. mode). This study is shown only spectral reflectance from 520nm to920nm, because the data beyond 520nm -920nm involve noise of this instrument. Thermometer can measure between 8mm and 16mm(1ch).
Fig. 2 Mobil Survey System
| Sensor | FOV |
| Spectrometer | Approx. 20° |
| DSC | 48.8° X 37.0° |
| CCD Camera | Approx. 65° |
| Thermometer | Approx. 24° |
| 3D Scanner | 30° X 28° |
Table. 1 FOV of reach sensors
Sensors FOV (Fields of View ) of each instrument is shown in Table.1. The directions of view for each sensor are adjusted to parallel. The information of spectrometer, GPS and thermometer are saved as ASCII file and 3D scanner data (distance and laser intensity) and images obtained by DSC are saved as BINARY file on PC. CCD images were recorded with DVC. This system can monitor spectral information are ground condition in the car. DSC and GPS camera is reconstructed to be able to release remote shutter.
3.2 Data analysis
In this study, we analyze spectral information, GPS positioning data, temperature data and vegetation coverage ratio derived from DSC and 3D scanner data. We calculated spectral reflectance from 520nm to 920nm and convolved broad band reflectance in red (580nm-680nm) and near infrared (725nm-920nm) region, and we compared with the spectral property of three types of Mongolian grassland.
This system record only GPS positioning data (x,y,z) but also UTC, standard deviation (x,y,z), GDOP (accuracy), direction and number of satellite in one ASCII file with every spectral information. We confirmed distribution of measurement point and accuracy each measurement points.
Calculation of vegetation coverage ratio is derived from DSC images using RGB pattern method. When we use only this method, it is very difficult to decide threshold value because it is needed trial and error. In this study, at first, we calculated VCR using both of distance data (grass height) and laser intensity data obtained with 3Dscanner, and then we compared with the VCR derived from DSC images and VCR derived from 3D scanner data. We could decide the threshold value.
4. Results and conclusion
We could analyze 100 measurement points in site 1 and site 2 respectively, and 57 measurement points in Site 3. And we could obtain the 100 points data in about 1.5 hours except Site 3. Fig. 3 shows the distribution of measurement points. In case of Site 1,system measured in approx. 2.5km X 2.5km, in case of site 2 and site 3, measured area is approx.1km X 1km area. Standard deviation of Longitude direction is 20m-30m, and latitude direction is 5m-18m. This indicates that these data are comparable with satellite scale. Therefore it is clear that this system measured more various kinds of data in enough wide area in the short time than previous system.
Fig.3 Distribution of measurement points
Fig. 4 depicts distribution of ground temperature measured with thermometer at each measurement points. It is inclined that Site 1, dense vegetation coverage area is lowest temperature but Site 2, sparse vegetation coverage is highest temperature. It is not able to decide to show the effect of vegetation condition because the condition of sunlight is also different in each observation date. It is needed data obtained by fixed point by fixed-point observation in each measurement site.
