Stepwise Pavement Crack Detection via Adaptive Threshold Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automatic crack detection methods for pavement images face challenges such as poor real-time performance, low identification rates, and inability to handle complex environments with uneven illumination, shadows, and textures, leading to difficulties in detecting small, weakly contrasted, and 'white cracks'.
Innovation Solution
A stepwise refinement detection method involving the extraction of suspected crack regions based on geometric and grayscale features, followed by confidence region determination and region growth using directional feature weighting, which includes adaptive threshold segmentation and compensation to enhance crack detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional crack identification methods (grayscale threshold, morphological processing, machine learning) are used, then crack detection can be performed, but the detection accuracy is low for small, weakly contrasted, and 'white cracks' in complex environments
Solution Approach 1:
The patent segments the crack detection process into multiple stages: initial crack region extraction, suspected crack region identification, and refined crack detection. This multi-stage segmentation allows each stage to focus on specific features, improving overall detection accuracy for subtle cracks while adapting to complex environmental conditions through progressive refinement.
Solution Approach 2:
The patent performs preliminary image preprocessing including illumination correction, shadow removal, and texture normalization before crack detection. This preliminary action prepares the image by eliminating environmental interferents, thereby improving the detectability of weak cracks and enhancing adaptability to complex environments beforehand.
2Measurement precision
If manual crack identification is used, then high detection accuracy can be achieved, but the workload is heavy and efficiency is low
Solution Approach 1:
The patent introduces an automated image processing system with multiple refinement stages as an intermediary between manual inspection and final crack identification. This intermediary system performs preliminary and refined detection automatically, reducing manual workload while maintaining high accuracy through multi-stage verification, thereby improving productivity without sacrificing precision.
3Reliability
If the same identification method is applied to all pavement images regardless of damage presence, then comprehensive coverage is achieved, but the processing time complexity is greatly increased
Solution Approach 1:
The patent applies partial action by performing full refined crack detection only on images suspected to contain cracks, while applying simplified methods to clear images. This selective application of detection depth maintains comprehensive coverage for damaged images while reducing processing time for the majority of undamaged images, optimizing the balance between reliability and time efficiency.
Solution Approach 2:
The patent performs preliminary classification to identify images containing cracks before applying the time-consuming refined detection algorithm. This preliminary action filters the dataset, ensuring that comprehensive detection is applied only where necessary, thereby maintaining reliability for damaged images while significantly reducing overall processing time.
4Ease of manufacture
If grayscale threshold methods are used for crack identification, then simple processing is achieved, but the method fails to identify cracks with low contrast and poor continuity
Solution Approach 1:
The patent segments the detection process into initial simple thresholding followed by refined multi-stage detection. The segmentation allows the simple method to provide initial results quickly, while subsequent refinement stages improve identification rate for low-contrast and discontinuous cracks, maintaining processing simplicity at the initial stage while enhancing precision through structured refinement.
Solution Approach 2:
The patent performs preliminary grayscale thresholding to quickly identify potential crack regions, then applies refined detection methods specifically to these regions. This preliminary action maintains processing simplicity for the initial pass while enabling improved precision for difficult-to-detect cracks in the refinement stage, balancing simplicity and identification rate.
Data Source
AI summary
The present invention discloses a stepwise refinement detection method for pavement cracks, processing pavement images comprising the following primary steps: extracting and processing marking lines, extracting ROA, adaptive threshold segmentation based on ROA, extracting ROC and region growth based on ROC direction feature weighting. The present invention rapidly extracts ROA based on that cracks are a set of pixel points which have similar grayscale and have distinct spatial accumulation features; rapidly locates the possible grayscale intervals and spatial locations of the cracks by ROA, realizes an adaptive threshold segmentation for images, establishes a confidence evaluation criterion, and accurately extracts ROC; accurately evaluates the development trend of the cracks based on the evaluation method for crack growth direction by segment weighting; utilizes an improved region growth method, uses the regions of confidence as seed regions, grows along the development trend of the seed regions, and sufficiently guarantees the accuracy of the crack growth and the completeness of the crack detection.