18. SOME RESEARCH AND APPLICATIONS IN THE CSIRO
(AUSTRALIA) EARTH OBSERVATION CENTRE ON SCENE
BRIGHTNESS DUE TO BRDF
163
leaves us with the question “What kernels are needed to deal with the land
surfaces we are interested in?”
3.
VIDEO DATA
Video data are used for low cost environmental monitoring. Such
monitoring requires consistent and standardized information. However, a
great deal of cost has been added to video data to try and achieve this
consistency in the face of the BRDF and atmospheric effects, which
compound calibration and instrument problems. The extensive hotspot and
BRDF effects are a major problem for central perspective sensors like videos
and cameras (even digital ones). Reducing this added cost is an important
commercial objective.
Australian groups who have used empirical methods with video data are
the Perth Minesite Rehabilitation group (Ong et al., 1995) and the Alice
Springs rangelands group (Pickup et al., 1995). The Alice Springs work
involves a variant on the kernel function approach applied to images
averaged along a run. The model is then used to balance the brightness
variations in the individual frames. The Perth Minesite Rehabilitation group
have been working to overcome spatial and angular variation by an
innovative use of a base image with little or no brightness variation—in their
case a Landsat image.
At CSIRO Mathematical and Information Science (CMIS), work is in
progress to statistically analyze video images against known kernels using
robust statistical methods. The objective for the video data producer is (low
cost) seamless mosaics approximating reflectance. However, there are a
number of research questions. For example, “Are there a few universally
representative functions? Do they change with land cover”? Since the
archive of video data may supply their own answer to these questions, it has
been decided to explore them with the range of existing kernels.
4.
CAL/VAL SITES
Calibration sites [and reflectance panels!] need BRDF models if they are
to be successfully used with satellite (or even airborne) data. An Australian
high reflectance calibration and aerosols site at Tinga Tingana in northern
South Australia has been modeled by Denis O’Brien and Ross Mitchell at
(AUSTRALIA) EARTH OBSERVATION CENTRE ON SCENE
BRIGHTNESS DUE TO BRDF
163
leaves us with the question “What kernels are needed to deal with the land
surfaces we are interested in?”
3.
VIDEO DATA
Video data are used for low cost environmental monitoring. Such
monitoring requires consistent and standardized information. However, a
great deal of cost has been added to video data to try and achieve this
consistency in the face of the BRDF and atmospheric effects, which
compound calibration and instrument problems. The extensive hotspot and
BRDF effects are a major problem for central perspective sensors like videos
and cameras (even digital ones). Reducing this added cost is an important
commercial objective.
Australian groups who have used empirical methods with video data are
the Perth Minesite Rehabilitation group (Ong et al., 1995) and the Alice
Springs rangelands group (Pickup et al., 1995). The Alice Springs work
involves a variant on the kernel function approach applied to images
averaged along a run. The model is then used to balance the brightness
variations in the individual frames. The Perth Minesite Rehabilitation group
have been working to overcome spatial and angular variation by an
innovative use of a base image with little or no brightness variation—in their
case a Landsat image.
At CSIRO Mathematical and Information Science (CMIS), work is in
progress to statistically analyze video images against known kernels using
robust statistical methods. The objective for the video data producer is (low
cost) seamless mosaics approximating reflectance. However, there are a
number of research questions. For example, “Are there a few universally
representative functions? Do they change with land cover”? Since the
archive of video data may supply their own answer to these questions, it has
been decided to explore them with the range of existing kernels.
4.
CAL/VAL SITES
Calibration sites [and reflectance panels!] need BRDF models if they are
to be successfully used with satellite (or even airborne) data. An Australian
high reflectance calibration and aerosols site at Tinga Tingana in northern
South Australia has been modeled by Denis O’Brien and Ross Mitchell at