A Visual Guide to Machine Vision Inspections

September 09, 2019

Samsara Machine Vision offers simple yet powerful inspection capabilities for any industry. Whether you’re new to Machine Vision or a seasoned veteran, get started in seconds with over a dozen easy-to-use tools that can be used for hundreds of use cases. The guide below offers a quick look at each inspection and how you can best apply it to your needs.


Angle

The Angle inspection returns the angle between two edges. After using the Edge inspection tool to identify at least two edges of interest in the image, simply use the dropdown options to select the edges you’d like to measure the angle between. Minimum and maximum angle thresholds can then be set so they system looks for your desired angle.

Example use cases: layout inspections, defect detection, package measurement


Barcode

The Barcode inspection tool finds and reads a 1D or 2D barcode within the specified bounding box. You can indicate what the expected barcode should be to validate the barcode found within the bounding box. Our machine vision system supports over two dozen barcode types - a full list can be found here.

Example use cases: label verification, SKU matching & validation


Boolean Logic

Boolean Logic allows you to combine the results from other inspection tools using logical operators (AND, OR, etc.). This enables you to define custom criteria for overall inspection results and works well for use cases involving multiple conditional inspections.

For example, if I’m verifying packaging for pet food, I may use different inspections to check for different packaging elements: a Pattern inspection to check for a picture of a dog or cat, and a Text inspection to check for the words “Dog Food” or “Cat Food.” In this case, I would set up my Boolean inspection with an AND operator: a picture of a dog AND “Dog Food” text. This ensures that if a particular package contains a picture of a dog but is labeled “Cat Food,” that package will fail the inspection.

Example use cases: packaging inspections, label verification, product analysis


Contour

The Contour tool searches for outlines and boundaries within a user-defined bounding box (the larger of the two boxes you see here). Contour allows you to check that part of an object, like a corner, is present in the correct shape. Contour works for both linear and curved shapes.

Example use cases: verifying product dimensions, packaging inspection, defect detection


Copies

The Copies inspection tool will find exact copies of an object that you specify. To use this tool, you’ll need to configure two bounding boxes: the search area bounding box, and the item to search for. Once you’ve identified an object you want to find copies of, you can adjust the match tolerance and angle of rotation to find all the copies within the search box.

Example use cases: product pack verification, label verification, fill level inspection


Distance

The Distance tool takes in two inputs from other inspection tools and measures the distance in pixels between them, with compatible inspection tools being: Pattern, Shape, Copies, Contour, and Edge. Thresholds can be set for minimum and maximum distances, as well as minimum and maximum angles of measurement.

Example use cases: Packaging measurements & inspections, margin inspections, verifying product dimensions


Edge

The Edge tool will search for the edge of an object in the direction of the arrow shown on the bounding box. A threshold filter will automatically be added to help increase the contrast between light and dark areas to make edges easier to identify. You can also adjust the polarity, sharpness, and contrast of the image to ensure the edge of interest is found.

Example use cases: verifying product dimensions, packaging inspection & measurements


Expiration Date

The Expiration Date tool allows you to read printed dates within a bounding box and verify if they fall within specified date ranges. Multiple date formats and timezones are supported, and image filters can be applied to enhance readability.

Example use cases: SKU verification, track & trace, label verification


Label

The Label tool verifies if labels are correct by checking the unique features of a captured image against those of a reference image. Features are defined as unique attributes that our image processing algorithm identifies within an image — think shapes, corners, edges, and points of contrast — and they serve as core comparison criteria for this tool. In order to pass the inspection, the number of features found in the captured image that match the reference image must exceed a user defined threshold.

Example use cases: label verification, SKU validation, track & trace


Pattern

This tool will search the entire image for a specific pattern that the user identifies in the master image. A pattern is a symbol, icon, or unique graphic that exists in the image. Simply drag your bounding box around the pattern of interest to inspect it.

Note that the pattern must be a closed object, meaning that we should be able to see the entire object in frame. If only a partial piece of the object is in frame, you should use the Contour inspection tool instead.

Example use cases: Validating that labels match your SKU, defect inspection


Presence

The Presence tool verifies if specific colors appear in a defined area. After adjusting your bounding box, the eyedropper tool can be used to select colors within the image you want to test, and the matching colors will become highlighted green on your master image. The threshold set for the inspection will be used to determine the amount of variance you’re willing to allow for an item to pass.

The Presence tool is highly flexible and can be used in creative ways to verify your products are properly assembled and packaged. For example, a translucent bottle of shampoo will reflect light differently depending on whether liquid is present within the bottle. You can thus set your Presence bounding box near the top of the bottle to verify the fill level depending on the brightness of the bottle.

In the VS2 Color cameras, additional options are available to build inspections based on the presence of RBG (Red, Blue, Green) or HSV (Hue, Saturation, Value) values within the image.

Example use cases: defect inspection, color matching, fill level inspection, label verification, assembly verification


Shape

The Shapes tool can be used to find and count the number of blobs, circles or rectangles in a defined search area. Each option has specific parameters that are described below:

  • Blob (Binary Large Object) - Blobs are objects of any shape or size. This option will search for any dark objects against a bright background and can be filtered based on size, circularity, convexity, or elongation ratio.
  • Circle - This option will find only circles within the bounding box. You’ll have the ability to filter based on the radius of the circles. This option will filter out shapes that are not close to exact circles. For round objects that are not circles, use the blob option.
  • Rectangle - This option will find rectangles with right angles at each corner. You will have the ability to filter based on rectangle area, aspect ratio, or corner angle tolerance.

Example use cases: product pack counting, feature validation, defect detection


Text

The Text tool leverages Samsara’s optical character recognition (OCR) algorithms to read and verify text in your image. The bounding box can be adjusted to search for specific text within an image, and the results can be compared against expected text to determine a pass or fail. Advanced settings are also available to optimize for a variety of character styles.

Example use cases: SKU validation, label verification, track & trace

Similar posts like this

Optimizing Digital Settings for Machine Vision

Learn how to fine tune Machine Vision camera settings for your use cases.

Lighting 101 for Samsara Machine Vision

Learn how to select the best lighting options for your Machine Vision application.

3 Tips for Better Machine Vision Lighting

Troubleshoot these common lighting challenges for your Machine Vision applications.