Blob Analysis

The total number of pixels constituting a single connected component. Area is the most fundamental scalar property, directly correlating to the physical size of a defect or feature.

• Zero-order moment: Computed as the sum of all white pixels in the blob.
• Calibration: Converted to real-world units (mm²) using the camera's spatial calibration factor.
• Application: Used to reject parts where a scratch or void exceeds a maximum allowable area threshold.

The geometric center of mass of the blob, calculated as the average (x, y) coordinate of all pixels in the region. The centroid provides precise spatial localization.

• First-order moments: The x-coordinate is M10/Area; the y-coordinate is M01/Area.
• Sub-pixel accuracy: Computed as floating-point values, enabling precision beyond the discrete pixel grid.
• Application: Guides robotic pick-and-place systems to the exact center of a part or defect for further action.

The length of the boundary contour enclosing the blob. Perimeter quantifies the complexity of the shape's outline and is sensitive to edge roughness.

• Chain code tracing: Calculated by traversing the boundary pixels and summing the distances between consecutive points.
• Compactness ratio: Often combined with area to derive form factors like circularity.
• Application: Distinguishes a smooth, circular void from an irregular, jagged crack, even if both have identical areas.

The minimal axis-aligned rectangle that fully encloses the blob. The bounding box provides the object's extreme dimensions.

• Parameters: Defined by (x_min, y_min) as the top-left corner and (x_max, y_max) as the bottom-right corner.
• Derived metrics: Width = x_max - x_min; Height = y_max - y_min.
• Application: Used to measure the maximum length and width of a defect, which is critical for standards that specify linear tolerances rather than area.

The convex hull is the smallest convex polygon that contains all points of the blob. Solidity is the ratio of the blob's area to its convex hull area.

• Solidity formula: Area / Convex_Hull_Area. A value of 1.0 indicates a perfectly convex shape.
• Concavity detection: A low solidity score reveals a shape with deep indentations or a highly branched structure.
• Application: Differentiates a solid particle of contamination (high solidity) from an agglomerate of smaller particles (low solidity).

The angle of the principal axis of the blob relative to the horizontal, along with the lengths of the major and minor axes of the best-fit ellipse.

• Second-order moments: The orientation is derived from the central moments, finding the axis of minimum inertia.
• Elongation: The ratio of major axis length to minor axis length quantifies how stretched a blob is.
• Application: Determines the directional alignment of a fiber or the dominant angle of a scratch, which can indicate a systematic process error.

The critical preprocessing step that converts a grayscale image into a binary black-and-white image based on a threshold value. Without proper binarization, blob analysis cannot function, as the algorithm requires a clear separation between foreground objects (defects) and background. Common techniques include:

• Global thresholding: A single intensity cutoff for the entire image
• Adaptive thresholding: Local thresholds that adjust for uneven illumination
• Otsu's method: Automatic threshold selection by minimizing intra-class variance

The core algorithm underlying blob analysis that scans a binary image and assigns a unique identifier to each connected group of foreground pixels. This process transforms a raw pixel map into discrete, analyzable objects. Two primary approaches exist:

• Two-pass algorithm: Scans the image twice, first assigning provisional labels and then resolving equivalences
• One-pass algorithm: Uses a union-find data structure to track connected regions in a single traversal
• Connectivity rules: 4-connected (edges only) or 8-connected (edges and corners) define what constitutes a connected region

Pre- and post-processing operations that refine binary images before or after blob analysis. These operations use a structuring element to probe and modify the shape of connected regions:

• Erosion: Shrinks foreground regions to eliminate small noise pixels
• Dilation: Expands foreground regions to close small gaps within a blob
• Opening: Erosion followed by dilation, removes small noise while preserving overall shape
• Closing: Dilation followed by erosion, fills small holes within blobs These operations are essential for cleaning up thresholding artifacts.

The quantitative measurement phase where geometric and statistical properties are computed for each labeled blob. Standard features include:

• Area: Total pixel count, directly correlating to defect size
• Perimeter: Length of the blob boundary, useful for shape complexity
• Centroid: Center of mass coordinates for precise defect localization
• Bounding box: Minimum enclosing rectangle for region of interest cropping
• Circularity: Ratio of area to perimeter squared, distinguishing round defects from irregular ones
• Convexity: Ratio of blob area to its convex hull area, identifying concave defect shapes
• Orientation: Angle of the major axis, relevant for directional scratches or cracks

The hardware and software system that serves as the deployment environment for blob analysis in manufacturing. AOI systems integrate:

• Illumination: Controlled lighting (LED arrays, structured light) to maximize defect contrast
• Imaging: High-resolution cameras or line scan sensors capturing the device under test
• Processing pipeline: Binarization → blob analysis → feature extraction → classification
• Pass/fail gating: Decision logic that compares extracted blob features against tolerance thresholds Blob analysis is often the first analytical stage in an AOI pipeline, identifying candidate defect regions for downstream deep learning classification.

A statistical validation methodology used to certify the measurement precision of a blob analysis system. GR&R quantifies two sources of variation:

• Repeatability: Variation when the same operator measures the same part multiple times (equipment variation)
• Reproducibility: Variation when different operators measure the same part (appraiser variation) For a blob analysis system measuring defect dimensions, a GR&R study ensures that area and perimeter measurements are consistent and trustworthy, typically requiring a total GR&R value below 10% for the system to be considered acceptable for production metrology.