Improving the Accuracy of an R-CNN-Based Crack Identification System Using Different Preprocessing Algorithms.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Civil Engineering

Background:

• Accurate road crack detection is crucial for effective road maintenance.
• Existing methods face challenges with complex road conditions.

Purpose of the Study:

• To develop an intelligent deep learning-based method for accurate road crack detection.
• To improve the efficiency and robustness of crack identification systems.

Main Methods:

• Utilized image sparse representation and compressed sensing for data preprocessing.
• Applied various algorithms including linear smooth, median filtering, Gaussian smooth, and grayscale threshold.
• Trained the Faster Regions with Convolutional Neural Network features (Faster R-CNN) model with optimal parameters.

Main Results:

• The proposed method demonstrated significant improvements in accuracy and efficiency.
• Achieved higher detection rates in the presence of road markings, shallow cracks, multiple cracks, and blurring.
• Showcased robustness across diverse and challenging crack image datasets.
• Reported an improvement in mean average precision (mAP) of up to 5% compared to original methods.

Conclusions:

• The deep learning approach effectively enhances road crack detection.
• The method offers a robust and efficient solution for intelligent road maintenance.
• Sparse representation and compressed sensing are valuable preprocessing techniques for crack detection.