- What did you personally accomplish this week on the project? Give files or photos that demonstrate your progress. Prove to the reader that you put sufficient effort into the project over the course of the week (12+ hours).
I started research on image pre-processing and segmentation using OpenCV libraries. In order to convert an image or a photo taken from a camera into a binarized image that could then be segmented and recognized, various pre-processing steps needs to be taken. For example, given an imported image of a braille document,
various OpenCV functions that would i) convert the image into gray scales, ii) reduce the overall noise of an image, iii) enhance the edge contrast, iv) then binarize the current image using similar approach as below, giving images like .
Further on, by i) finding the connected components and extracting the mean height and width using cv’s cv2.connectedComponentWithStats function and np.mean() function. ii) Find empty rows, defined as having less than mean_h/2 pixels. iii) Group and assign each labels to rows, and then find the mean row value associated with each label.After successive steps, the image will then be segmented like this:
Next steps from here is to adopt Hough transform to identify and recognize each of the segmented braille alphabets. And the work will follow in the following week.
- Is your progress on schedule or behind? If you are behind, what actions will be taken to catch up to the project schedule?
Things that were due this week were i) research on image pre-processing and segmentation and ii) initiation on the hardware design. Everything planned is completed on track and as scheduled.
- What deliverables do you hope to complete in the next week?
By the end of next week I plan on accomplishing the following: 1) collaborative initial hardware design, in a form of camera mounted glasses in combination with a wearable vest that would hold the jetson xavier. 2) Tradeoffs for various image pre-processing and segmentation methods and start testing out the effectivenesses of each preprocessing algorithms to decide whether the code needs to be written from scratch to match our metrics requirements for recognition and translation. 3) Initiate camera integration