AirSpex 2007

7. Conclusion

 

AirSpex 2007 did not achieve all of its predefined objectives. Due to a scarcity of available flight time, bad weather conditions and technical difficulties with the drones, we only obtained one useful data set for analysis, during the helicopter flight over Longyeardalen, 03.05.2007. We measured the ground reflectance using a spectrograph, while simultaneously capturing digital images with an SLR and recording a movie using a web camera. We performed a supervised classification on this data set.

 

For the first classification attempt, we used four classes, and the result was considered fairly successful. Snow, water, buildings and roads were well separated in the classified image, and the resemblance of the simultaneously captured aerial photos of Longyearbyen was clear. The day we obtained the data, it was overcast. Compared to data obtained during the flight campaign in 2005, when they had sunny conditions, our data appeared to be more easily classified. The overcast conditions provide a much more consistent illumination of the target scenery on the ground. Thus, our classification does not need to consider the problematic effects caused by direct sunlight, like shadows due to objects on the ground and scattered clouds.

 

The next classification was performed using five separate classes for the buildings, specifying the colors of the roofs. This was done in order to see how the effect would be on the resulting image, expecting it to be difficult to perfectly separate the colors of all the roofs. We also compared the result obtained using Bayes classifier, to the result obtained with the Minimum Distance classifier. As expected, Bayes classification provided the best result. The colors of the roofs were fairly well separated, considering the close resemblance of some of the classes (e.g. “Dark Grey” and “Black”).

 

We were not able to acquire any more useful data, meaning that we never sent an image from the airplane using the GPSCam Iridium System, and we did not obtain any data suitable for analyzation from the drone ”Otto”.

 

The helicopter is likely to be more ”wobbly” during flight than the airplane. The effect of the wobbling is easily observed in all of our images. Had the gyro correction software worked, it might have improved the image quality and revealed whether the helicopter is just as useful as a means of transportation for performing airborne spectral imaging of the ground. The gyro correction software might be well worth testing in future campaigns where helicopters are used. As helicopters in general are capable of flying at a lower speed and at a lower altitude than airplanes, the successful usage of helicopters may broaden the range of use of airborne spectral imaging, as helicopters are more flexible to weather conditions.

FLIGHTS

DATA ANALYSIS