Team Meeting

Full team minutes can be found here hosted on Google Drive (DIT Sign in required) Team Minutes

Summary

Here, we discussed the final set of research for our own different elements for the project. The main aim of this meeting was to finalise our implementation of the project, in the direction of completion.

Discussed

Robert:

Had some steler success with obtaining the width and the length of the vehicles in an image. Robert made use of the Canny Edge Detection function:

cv2.Canny(image, threshold1, threshold2[, edges[, apertureSize[, L2gradient]]]) → edges

Using these edges found, he looked at retrieving the contours of the edges:

cv2.findContours(image, mode, method[, contours[, hierarchy[, offset]]]) → contours, hierarchy

And finally, found the extreme top & bottom corners of the contours to get an approximation of the box size around the edge detected vehicle. No further work is required here, code is ready for main branch deployment. See his work here.

Diarmuid:

Had some success with recognising windows in an images. Diarmuid made use of the Canny Edge detection and findContour functions, mentioned earlier. Furthermore, he looked isolating rectangles found by these contours:

cv2.boundingRect(points) → retval

Diarmuid was able to locate the windows with about 30-40% accuracy. Further development will come. See his work here.

Ciarán:

Had success in replacing skimage's Sobel edge detection, with openCV's sobel edge detection

cv2.Sobel(src, ddepth, dx, dy[, dst[, ksize[, scale[, delta[, borderType]]]]]) → dst

The resulting image is not the same standard of edge detection as what was obtained by skimage's results. It was decided that Sobel must unfortunately be dropped as our edge detection method, in favour for Canny Edge detection, which is proving to be on par with Sobel in it's results as part of Robert's and Diarmuid's work. Further discussion can be found here.