The document contains simple steps to implement machine vision applications in matlab.
Following are the applications covered in the document,
1. Counting connected objects using watershed algorithm (number of fruits)
2.Liquid level estimation in beverage bottles
3.Segmenting nuts/bolts and counting
4.Pencil length identification
5.Rice grain Inspection
6.Blister Inspection
7.Nut sorting
1. Sriramemarose.blogspot.in
COUNTING NUMBER OF FRUITS USING WATERSHEDING
Problem statement:
Fruits distributed closely will be considered as a single blob in normal thresholding,
therefore counting is impossible with thresholding
Sample image: Input image
Object boundaries and regional mamima superimposed on orginal image after watershed
Output image with counted fruits
2. Sriramemarose.blogspot.in
Steps used:
Filter the image to eliminate noise
Create an edge emphasizing filter kernel(say ‘a’) after converting the image to grayscale
Create a transpose of the filter kernel(say ‘b’)
Obtain two images with one filtered with a and other filtered with b
Calculate the gradient magnitude of the two images
Perform morphological operations and reconstruct the image on the original image
Convert the resultant to binary image and estimate the distance transform
Perform watershedding and segment the watershed boundary lines
Obtain the regional minima of the gradient magnitude by morphological reconstruction of
the boundary lines and regional maxima of the original image
Find the number of fruits from the boundaries of the new image
Other examples:
Counting number of cells in medical imaging
Connected objects segmentation
3. Sriramemarose.blogspot.in
LIQUID LEVEL IN BEVERAGE BOTTLES
Problem statement:
Overfill and Underfill identification
Quantity estimation
Sample image:
Processed image:
Steps involved:
Perform color segmentation based on sample`s threshold
Smoothen the segmented image with suitable filter
Apply morphological operators to remove remaining components other than sample
Calculate the pixels contributing to the sample
Calibrate the pixels in terms quantity(volume)
Label the calibrated quantity value to its corresponding sample
Applications:
Pharmaceutical Industries
Beverage Industries
Batch processing
4. Sriramemarose.blogspot.in
Nuts and Bolts
Problem statement:
Distinguish between nut and bolt
Count number of nuts and bolts
Sample image Processed image
Steps involved:
Adjust the contrast after converting to grayscale image
Obtain the binary image with suitable threshold level
Filter the noises with suitable filters
Apply morphological operators to enhance the features
Detect the nuts using hough circle transform with appropriate sensing radius and
sensitivity
Subtract the detected nuts from the image, which leaves only with the bolts
Detect the number of bolts using binary labeling
Applications:
Automotive Industries
Manufacturing Industries
Industrial Automation
5. Sriramemarose.blogspot.in
PENCIL LENGTH IDENTIFICATION
Problem statement:
To identify objects (pencil) length to ensure manufacturing defects
Sample image:
Test image Pencil length Pencil and lead length
Steps involved:
Obtain a Boolean image with suitable threshold value
Apply filters to remove noises
Perform morphological operation to enhance the detection, without altering the object
dimension
Segmented the object from background and label the object blob
Find the region properties of the object blob
Measure the pixels and calibrate in real world units
Applications:
Manufacturing industries
Factory Automation
Quality control
6. Sriramemarose.blogspot.in
RICE GRAIN INSPECTION
Problem statement:
To identify broken grains
To segment good quality grains
Sample image:
Input image
Steps involved:
Eliminate the uneven illumination using morphological tophat operation
Adjust the image contrast
Obtain the binary image with suitable threshold value
Find the connected components in the image to locate each grain, use filter if needed
Find the region properties of the grains
Traverse through every connected component (pixel index list) and check its
corresponding properties
If a grain does not satisfy the standard quality (based on its property value), subtract that
particular component(grain) from the pixel index list
Applications:
Food processing Industries
Quality control
7. Sriramemarose.blogspot.in
BLISTER INSPECTION
Problem statement: To identify the missing in the tablet strips( Blisters)
Sample images:
Good sample Processed image
Sample with defect Processed image
Steps involved:
Convert to grayscale image and adjust the contrast
Obtain the binary image with suitable threshold value
Eliminate the noise with appropriate filters
Perform morphological operations to segment tablet and tablet strip
Apply hough transform to find the tablets
Based on the detection, mark the blister as defected or good.
Applications:
Pharmaceutical Industries
Manufacturing industries
8. Sriramemarose.blogspot.in
NUTS SORTING
Problem statement:
To measure the diameter of the nuts
To sort them based on their size
Sample image:
Processed image:
Nut with minimum diameter Detected nuts
Steps involved:
Convert to grayscale image and adjust the contrast
Obtain the binary image with suitable threshold value
Eliminate the noise with appropriate filters
Perform morphological operations to enhance the features
Use hough circle transform to detect the nuts since it has circular feature
Detect the required nuts radius using mathematical operators
Segment the detected nuts
Applications:
Manufacturing Industries
Industrial Automation
Quality control