NUS ME5405 Module - Machine Vision
Images 1 and 2 are 64x64, 32 level images. These images are shown as coded arrays that contains an alphanumeric character for each pixel. The range of these characters is 0-9 and A-V, which corresponds to 32 levels of gray.
Image 1:
For Image1 (chromo.txt), the main function file is Project_1.m.
Perform the following tasks:
- Display the original image on screen.
function [histogram_num, small_gray_img, gray_img] = original_image(filepath)- Threshold the image and convert it into binary image.
function [small_gray_img, gray_img] = binary_image(histogram_num)- Determine an one-pixel thin image of the objects.
function [no_pad, no_pad_enlarged] = onepixel_image(binary)- Determine the outline(s).
First method
function [outline_img, outline_enlarged] = outline_image(binary)Second method
function [height, width,output] = canny_edge_detection(input_image)Label the different objects. Discuss and compare 4-connectivity and 8-connectivity in your report.
Image 2:
For Image1 (charact.txt), the main function file is Project_2.m.
Perform the following tasks:
- Display the original image on screen.
function [histogram_num, small_gray_img, gray_img] = original_image(filepath)- Create a binary image using thresholding.
function [small_gray_img, gray_img] = binary_image(histogram_num)- Determine a one-pixel thin image of the characters.
function [no_pad, no_pad_enlarged] = onepixel_image(binary)- Determine the outline(s) of characters of the image.
First method
function [outline_img, outline_enlarged] = outline_image(binary)Second method
function [height, width,output] = canny_edge_detection(input_image)Label the different objects. Discuss and compare 4-connectivity and 8-connectivity in your report.
Segment the image to separate and label the different characters.
Arrange the characters in one line with the sequence: AB123C
Rotate the output image from Step 6 about its center by 30 degrees.
Using the training dataset (p_dataset_26.zip), train the (conventional) unsupervised classification method of your choice (Here k-nearest neighbors (kNN) is used) to recognize the 6 characters (“1”, “2”, “3”, “A”, “B”, or “C”). You should use 75% of the dataset to train your classifier, and the remaining 25% for validation (testing). Then, test your trained classifier on each characters in image 1, reporting the final classification results. Do not use the characters in image 2 as training data for your classifier.
Throughout step 8 (training of the classifier), also experiment with pre-processing of the data (e.g., padding/resizing input images) as well as with hyperparameter tuning.
You can just directly run the project_1.m and project_2.m to see the results, which is very simple.