Edge Orientation Analysis for Partially Detached Adhering Substance Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional adhering substance detection apparatuses face challenges in accurately detecting adhering substances on camera lenses, particularly when the substance is partially detached, due to increased granularity and frequent transitions in angle classes, making it difficult to distinguish between covered and uncovered conditions.
Innovation Solution
The apparatus employs two types of angle classifications: 'up/down/left/right four classification' and 'tiled four classification' to accurately detect changes in edge orientations, counting transitions within unit regions to determine the presence of adhering substances, and uses edge feature values in cells composed of predetermined pixels to enhance processing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If conventional temporal difference detection is used to detect adhering substances, then the detection process is simple, but the detection accuracy deteriorates when substances are partially detached
Solution Approach 1:
The patent transforms the detection approach by changing from temporal difference analysis to spatial edge orientation analysis. By calculating edge vectors and classifying their orientations into angle classes, the system detects adhering substances based on the statistical distribution of edge orientations rather than temporal changes, thereby improving accuracy for partially detached substances
Solution Approach 2:
The patent divides the detection process into local analysis units (cells with multiple pixels) and evaluates edge orientation characteristics within each cell. By analyzing the local edge orientation distribution and transition counts within unit regions, the system achieves more precise local detection while maintaining overall detection accuracy
2Measurement precision
If edge orientation analysis with angle classification is used to improve detection accuracy, then the processing complexity increases
Solution Approach 1:
The patent segments the image processing into discrete steps: dividing the image into cells, calculating edge vectors for each pixel, classifying edge orientations into discrete angle classes, and analyzing transition counts within unit regions. This segmentation transforms a complex continuous analysis into manageable discrete operations, improving both accuracy and processing efficiency
Solution Approach 2:
The patent introduces a new dimension of analysis by classifying edge orientations into discrete angle classes and analyzing the statistical distribution and transitions of these classes. This transforms the problem from direct pixel-level comparison to a higher-level statistical analysis of orientation patterns, enhancing detection capability while managing complexity
Data Source
AI summary
An adhering substance detection apparatus according to an embodiment includes a calculating unit and a determining unit. The calculating unit calculates, for each cell composed of a predetermined number of pixels in a captured image, an edge feature value that is based on edge vectors in the pixels, and that classifies an edge orientation that is included in the edge feature value, into two types of angle classes. The determining unit determines a condition of adhering substance adhering to an image capturing apparatus that captured the captured image, based on a transition count representing number of transitions the angle class goes through within a unit region that is a predetermined region composed of a predetermined number of the cells.