Adaptive Threshold Object Detection for Noisy Surfaces
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image segmentation methods, particularly in industrial applications, face challenges with real-time detection of elongated or line-like objects on noisy surfaces due to susceptibility to non-uniform illumination and complex backgrounds, leading to inaccurate and time-consuming results.
Innovation Solution
An accelerated logical morphological method based on principles of human visual perception, using adaptive thresholding and logical morphological operations to fuse object fragments, enabling rapid detection of elongated objects and improving accuracy and speed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional segmentation methods are used on noisy surfaces with non-uniform illumination, then object detection can be performed, but detection accuracy deteriorates and computing time increases
Solution Approach 1:
The patent applies segmentation by dividing the image processing task into distinct stages: initial binarization to separate marked pixels from background, followed by morphological closing to connect fragments. This multi-stage segmentation approach improves detection accuracy by systematically handling different aspects of object reconstruction separately, rather than attempting single-step detection that fails on noisy surfaces.
Solution Approach 2:
The patent performs preliminary binarization and morphological closing operations before final object detection. By pre-processing the image to mark and connect potential object regions in advance, the system prepares the data structure for more accurate subsequent detection, reducing computing time for the actual measurement phase while maintaining high detection accuracy.
2Stability of the object's composition
If morphological closing is used to connect object fragments, then object completeness improves, but false connections increase on complicated backgrounds
Solution Approach 1:
The patent applies local quality by performing morphological closing operations specifically on marked pixels identified through adaptive thresholding, rather than applying uniform processing to the entire image. This localized approach connects fragments only where object features are actually present, preventing false connections between unrelated background elements while maintaining object completeness.
Solution Approach 2:
The patent uses dynamic adaptive thresholding that adjusts to local image characteristics, allowing the morphological closing operation to be more aggressive in regions with actual objects while being more conservative in noisy background regions. This dynamic adaptation reduces false connections while maintaining object completeness through context-aware processing.
3Measurement precision
If adaptive thresholding is used to handle non-uniform illumination, then segmentation accuracy improves, but computational complexity increases
Solution Approach 1:
The patent segments the thresholding process into multiple passes: an initial coarse thresholding to identify potential object regions, followed by refined adaptive thresholding only in those regions. This segmented approach maintains high segmentation accuracy through adaptive methods while reducing overall computational complexity by limiting expensive adaptive operations to relevant image areas only.
4Productivity
If conventional methods are used for real-time detection, then implementation simplicity is maintained, but detection speed decreases
Solution Approach 1:
The patent implements preliminary binarization and morphological closing operations that prepare the image data in advance, creating an optimized structure for rapid subsequent detection. This preliminary processing, while adding some complexity, enables real-time detection speed by pre-computing the difficult parts of the algorithm and simplifying the final detection step.
Solution Approach 2:
The patent replaces conventional mechanical morphological operations with optimized logical operations and adaptive algorithms that execute faster on modern processors. By substituting traditional iterative morphological methods with more efficient computational approaches, the system achieves real-time detection speed while maintaining the necessary processing complexity.
Data Source
Figure 1
Figure 2
Figure 3A~3B
AI summary
A method for detecting an object on an image represented by pixels comprises the following steps: "Determining a first and a second adaptive threshold for pixels of the image, which depend on a mean intensity in a region around the respective pixel," "Identifying sub-objects from pixels of a first type, which are obtained based on a comparison with the first adaptive threshold," "Identifying pixels of a second type, which are obtained based on a comparison with the second adaptive threshold," and "Combining a first and a second of the sub-objects using pixels of the second type to form an extended sub-object, if there is a minimum distance between the first and the second of the sub-objects, wherein the object to be detected can be described by a sum of the sub-objects from pixels of the first type and/or the obtained extended sub-objects."