Data Collection | Results
As of this writing the approach yields detection rates of 99.7% up to 3.5 miles and 96.1% up to 4.5 miles with a false positive rate of 0.02 per frame ( using an IPX-4MP5 and Zeiss 85 mm lens thereby yielding 0.005 false positives per megapixel per frame ). Furthermore almost all of the false positives that occur are due to objects that may be relevant to collision avoidance: birds, radio towers.
This graph shows the ROC curve of performance for the 4 stages of the vision algorithm.It can be seen that the Stage 4 ( MorphiseSVM ) eliminates most of the false positives and achieves detection rates of upto 99.7% upto 3.5 miles. The FAA regulation specifies 3 miles as the radius of detection for a human equivalent Sense and Avoid system.
This graph shows the performance of the algorithm as a function of range for 3 different false positive rates per frame. It can be seen that even with 0.022 false positives per frame there is almost 96 % detection upto 4.5 miles.
This video shows an example of the Piper Archer being tracked out to a range of 4.5 miles and also a couple of birds being tracked. The inset in the video shows the plane and the birds along with dust on the lens which are shown in blue. The dead pixels and dust on the lens are detected automatically by the algorithms and are eliminated from the tracking stage.
Collision Avoidance Results:
This graph shows the plot of minimum detection distance required for 3 different ownship speeds as a function of the speed of the intruder aircraft.Under the best scenario of intruder speed of 60 knots and intruder speed of 100 knots the intruder needs to be detected only about 700 meters away in order to guarantee a collision free path. For the worst case scenario the minimum detection distance is about 2.1 Km.
This graph shows the minimum detection distance as a function of different lateral acceleration bounds of the intruder aircraft. It can be seen that there is no effect of the lateral acceleration values on the minimum detection distance for the range of values commonly found in UAVs. However for higher values of lateral acceleration there is a direct effect on the minimum detection distance which increases.