An Integrated Deep Learning and Fuzzy Logic System for Road Crack Severity Analysis, and Pedestrian Fall Risk Prediction

Cracks on pedestrian sidewalks and walkways pose a significant safety hazard, increasing the risk of trips, falls, and injuries, particularly for vulnerable groups such as the elderly and children. Traditional crack detection and severity assessment methods are usually manual, time-consuming, and subjective, especially in the Indian context, where road andsidewalk inspections are still conducted mainly through visual surveys due to cost and infrastructure constraints. This paper proposes an integrated framework of deep learning and fuzzy logic to analyze sidewalk crack severity and predict pedestrian fall risk automatically. A novel crack quantification method using edge detection and adaptive segmentation is proposed to measure crack width accurately. A fine-tuned deep learning model is employed for automating crackseverity prediction, which achieved ۹۵% accuracy and demonstrated robustness to noise, blur, and lighting variations. To estimate fall risk, a fuzzy inference system is developed considering four inputs: crack severity, road condition, weather, and pedestrian age, and a set of expert-defined fuzzy rules is applied to estimate risk levels. The outcomes show the effectiveness of the proposed FIS scheme, which achieved ۹۵% accuracy and outperformed non-fuzzy baseline approaches.

کلیدواژه ها:

Road Crack Quantification ، Crack Severity ، Deep learning ، Pedestrian Falling Risk ، Fuzzy logic ، Triangular Membership Function