A supplemental material for the paper
A statistical model for fast obstacle image-map estimation from unmanned surface vehicles
Matej Kristan, Vildana Sulic Kenk, Stanislav Kovacic and Janez Pers
(Submitted)


2014


From the authors: This page contains some supplementary video examples to demonstrate the performance of the obstacle-map estimation model proposed in the paper entitled "A statistical model for fast obstacle image-map estimation from unmanned surface vehicles". The following videos have been rescaled and encoded under high compression to enable a fast download and are not equal to the input data used in the experiments. The videos are meant to give an illustrative addition to the results presented in the paper and to give a better insight in the performance of our method.

 

Video examples:

The USV used in the experiments

Here is a video of the USV that we used in our experiments.

Examples of selected methods on two videos from the MODD dataset

Below we show object-map estimation results in a demanding environment (only two videos from the Modd dataset). We show the results for our method (SSM), the grab-cut variant (GC), the superpixel-based segmentation (SPX),Felzenswalb-Huttenlocher graph-based segmentation at full-sized image (FZH_full) and reduced-size image (FZH). The images contain overlaid images, ground truth segmentation and output of the algorithm on each frame: the image contents has been compressed into the blue channel, manual water annotations into the green channel, and algorithm-generated water segmentation into the red channel. Therefore, the cyan region shows the area, which has been annotated as water, but has not been segmented as such by the algorithm (bad). The magenta region shows the area, which has not been annotated as water, but has been segmented as such by the algorithm (bad). The yellow area shows the area which has been annotated as water and has been segmented as such by the algorithm (good), and blue region shows the area which has not been annotated as water and has not been segmented as such (good). Finally, the darker band under the annotated edge of the water in all colors shows the "ignore" region, where evaluation of small obstacle detection does not take place.

method
Color video
SSM (our)
CG
SPX
FZH
FZH_full
Video: "Marine"
method
Color video
SSM (our)
GC
SPX
FZH
FZH_full
Video "bag drop"

Additional video examples of SSM performance

Below we show additional videos that further illustrate performance of SSM. The videos show two images: The left image diplays the original image in gray and overlays in blue the estimated water region. It also shows the spatial parts of the Gaussian components in our semantic model (blue component denotes water, green is the middle component and red is the upper component). The right-hand image shows the color image and overlays the estimated edge-of-water (green) along with the in-water object detections (yellow rectangles).


Video2.

Video1 .

Video3.

Video 4.

Video 5.

Video 6.

Video 7.

Video 8.

Video 9.

Video 10.

Video 11.

Video 12.