Vision:: Tasks

Vision:
Incoming images are first converted into HSV domain, and hue plane is
extracted for further processing as most of the tasks to be done are
based on colour. After doubling intensity of each pixel (for contrast),
canny edge detector is used to get Edge image for each frame, with
threshold set every time using Otsus method of minimum class variance
within foreground and background values for maximum reproducibility.
Tasks:
1) Aligning with plank and following:
The main stages here are segmentation and extracting line.
Segmentation is done by getting the minimum of the pixels in the view
and based on the histogram distribution around the minimum, the
window for thresholding the image with this minimum value is chosen. In
the binary output image, the external contours are extracted. Out of all
the contours observed, the one with maximum area is selected. The
output contours likeliness of being in rectangular shape is found by
extracting its minimum area bounding rectangle and matching with its
convex hull (for reducing noise in thresholding). Based on matchs
likeliness decision is taken whether the contour detecting is that of plank
or not.
A line is fit within the contour area using linear regression.
Another approach is using canny edge detection, but is prone to noise.
Water pixels intensity is found from the maximum of histogram and that
of plank as minimum one. Further processing it with Hough lines would
give us two lines.
Thus, output would be angle made by the line with horizontal axis and
perpendicular distance from it to the centre, if a plank is detected.
2) Detecting and tracking buoys:
Here segmentation is done, either based on colour or canny edge
method.
In colour thresholding method, the intensity of the buoys is extracted
from histogram for every frame, and used for getting binary image to
track with the help of the contour moments detected.
Either detecting maximum from Yellow, Green, Red planes of the image
for segmentation. As nothing else other than these can have max values.
Area of these segmented areas and also likeliness with its minimum
enclosing circle is checked for conforming.
In canny edge method, Hough circles are extracted from edge image.
Thus, output would be approximate centre pixel coordinates of circles, if
detected.
The estimate of the distance between AUV and buoy is computed in two
ways. Either with one camera, by estimating the area covered by the
buoy in the view or by getting the exact distance using stereo vision.
3) Detecting its orientation and position w.r.t Love lane:
Tracking these is possible in two ways. Colour thresholding and Edge
detection.
[In colour thresholding method, pipe lines are segmented out, dilated,
contour is found. Threshold point is made accurate by experimentation.
Convex hull of contour found is computed to neglect the disturbances,
angle made at the pipes joint (computed from contours vertices) is used
for positioning AUV to cross it without touching it.
Hough lines is used for line extraction from edge image and used to
compute angle between them.]
The position (perpendicular distance) of vertical and horizontal pipes from
the centre of the image is calculated and used to navigate AUV such that
it crosses it without hitting the vertical one and above the horizontal one.
Thus, output would be the perpendicular distances of the pipes from the
centre.
4) Detecting bins for dropping markers:
Blue colour and rectangular shape of bins is used for confirming that AUV
has bins beneath it. Exact position of cross, lines are detected in the
cropped image (Blue inner surface).
5) Detecting Cupid:
Red and Green planes are searched for max intensity area. If the area
segmented in both planes are rectangular (contour having four vertices),
also if adjacent, the centres of rectangles as the targets for torpedo
shooting. [Also derivative of Heart shaped contour if has two
discontinuities]
6) For floating up:
A circle detected in the feed of camera placed to view bottom surface of
pool.

Vision:: Tasks

Uploaded by

Copyright:

Available Formats

Vision:: Tasks

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Vision:: Tasks

Uploaded by

Copyright:

Available Formats

Vision:

You might also like