Embedded Files
video_open2.avivideo_open3.aviComparison between Fourier-based tracker and mean-shift-based tracker for gas leak detection in open air simulation scenario
Comparison between Fourier-based tracker and mean-shift-based tracker for gas leak detection in open air simulation scenario
Above video clips show the convergence capability of the Fourier-based tracker (FT) and the mean-shift-based tracker to the source of the gas leak. Trackers try to go to the maximum of the distribution, which corresponds to the source of gas leak. The green dots correspond to the location of ordinary nodes which provide uncalibrated measurements. Super-nodes (trackers) have calibrated sensors and they can access the data generated by uncalibrated sensors.
Mean-shift tracker, which is the magenta-colored supernode meanders too much as shown in the video clip on the right. On the other hand, FT tracker (red-colored supernode) rapidly converges to the source of the gas leak.
We also developed a modified mean-shift tracker, which is the yellow-colored supernode on the video clip (left). It converges to the source of the gas leak faster than ordinary mean-shift tracker but its performance is not as good as the Fourier-tracker.
video_open_gaussian.aviThis video also shows the performance of both trackers in the presence of excessive noise in the measurements obtained by the ordinary nodes (green dots). As we can see, the FT tracker (red-colored) performs significantly better than the mean-shift tracker. This is due to the fact that imposing bandwidth constraints in the frequency domain suppresses the high-frequency component in the noisy measurements and, therefore, mitigates any meandering behavior, unlike in the case of the mean-shift tracker, which can be very sensitive to the noise present in the samples used to perform updates.
Report abuse