Backlit Tray Object Identification via Lab Color Space
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object identification technologies face difficulties in accurately separating objects from their background, especially when objects and backgrounds share similar colors or when shadows complicate the distinction, as seen in the case of pastries on a white tray, leading to challenges in automated pricing and identification.
Innovation Solution
An object identification apparatus employing backlighting and image processing techniques to separate objects from their background by converting digital color images into brightness components, using Lab color space to extract contour lines and distinguish regions based on brightness and chromatic dispersion, allowing for accurate identification of objects even when they share similar colors with the tray.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If ambient lighting is used for image capture, then the setup is simple and cost-effective, but objects with similar colors to the background (e.g., white pastry on white tray) cannot be accurately separated
Solution Approach 1:
The patent changes the illumination parameter from ambient lighting to structured backlighting, transforming the lighting condition to create brightness differences between objects and background. This parameter change enables accurate separation of similarly-colored objects by exploiting the transparency differences under backlit conditions.
Solution Approach 2:
The patent effectively changes the apparent brightness/color characteristics of the background by using backlighting through the transparent tray. The tray appears brighter in regions where light passes through it, creating contrast with opaque objects placed on it, thereby enabling separation based on brightness rather than original color.
2Measurement precision
If active illumination is used to create shadows for object separation, then some contrast is achieved, but shadow portions become ambiguous and cannot be reliably distinguished as object or background
Solution Approach 1:
Instead of using front lighting that creates shadows, the patent inverts the lighting approach by using backlighting. This inversion causes transparent regions (background) to appear bright while opaque regions (objects) appear dark, reversing the expected shadow pattern and eliminating ambiguity in region classification.
3Measurement precision
If backlighting is used to illuminate the tray, then transparent regions appear bright and opaque objects appear dark enabling separation, but the tray material must be transparent or translucent
Solution Approach 1:
The patent changes the optical parameter of the tray from opaque to transparent/translucent to enable the backlighting method to work. This material parameter change allows the tray to transmit light, creating the necessary brightness contrast for accurate object separation.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables definitive extraction and identification of objects from their background, even under conditions where ambient lighting complicates the distinction, ensuring accurate and stable processing without misidentification, and allows for efficient processing of diverse pastry types.
Implementation Method 1
The tray is formed so as to be at least semitransparent such that the backlighting is transmitted therethrough to the image capture means side thereof
Data Source
AI summary
Identification apparatus is capable of automatically performing processing for separation of objects from background to definitively extract objects to be identified. Backlighting is used to cause illumination from behind a tray on which pastry serving as object has been placed, and while in this state an image of the pastry and the tray are captured. The captured image is converted into an L image representing the L-axis component in Lab color space (S11). The L image is used to create a Canny edge image (S12), and contour extraction is carried out (S13). Pastry region(s) are extracted from background region(s), and a mask image is output (S14, S15). The extracted pastry region(s) are used to carry out identification of pastry type by comparison with registered characteristic quantities. Determination of pastry region(s) is carried out based on magnitude of chromatic dispersion and distance between colors in Lab color space.