Deep Learning-Based System for Monitoring Pavement Distresses

基于深度学习的路面病害监测系统

• 批准号: 571245-2022
• 负责人: Easa, Said
• 金额: $ 1.46万
• 依托单位: Ryerson University
• 依托单位国家: 加拿大
• 项目类别: Idea to Innovation
• 财政年份: 2021
• 资助国家: 加拿大
• 起止时间: 2021-01-01 至 2022-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=729099
• 关键词: Deep; Learning; Based; System; Monitoring

Maintaining highway and airport pavement quality at an acceptable level is challenging for transportation authorities. Manual, semi-automated and automated methods have been developed to evaluate pavement distresses, such as cracks, viscoplastic deformation, and surface defects. However, the manual process is costly, labour-intensive, and time-consuming and is prone to human error. The semi-automated and automated methods lack accuracy and involve high costs. Furthermore, road and airport authorities lack the screening tools to make real-time evaluations associated with simultaneous visualizations. Our proposed automated vision-based system overcomes these limitations by detecting and evaluating all pavement distresses mentioned above more efficiently and accurately. The pavement distress monitoring system includes two parts: hardware and software. The hardware is an off-the-shelf camera with certain features used for image data collection. The software is a Python-based code that simultaneously analyzes the collected images using computer vision, image processing, and deep learning algorithms to detect and assess pavement distresses. The generated outputs include pavement distress type, dimensions, severity, and location. The proposed product can be installed at the back of a moving vehicle or on a drone. It can capture images remotely at different moving speeds and accomplishes the assessment process with reasonable cost and without traffic interruption. In addition, this technology can scan all road lanes by capturing successive images with just one passage along the road of interest and can be used during both day and nighttime. Based on our primary market size estimation, the serviceable available market is about 60,000 municipalities, airports, and provincial governments, and capturing 10% of this market amounts to 6,000 sales. Thus, given typical product values, the annual income might reach US$21M. Using the NSERC fund, we will work on our product market assessment to validate our business model elements and create a more realistic and reasonable development process in product specifications and capabilities from the customer's point of view.

将高速公路和机场路面质量保持在可接受的水平对交通部门来说是一项挑战。人工，半自动化和自动化的方法已经开发出来，以评估路面损坏，如裂缝，粘塑性变形，和表面缺陷。然而，手动过程是昂贵的、劳动密集型的和耗时的，并且容易出现人为错误。半自动化和自动化方法缺乏准确性并且涉及高成本。此外，道路和机场当局缺乏筛选工具，无法进行与同时可视化相关的实时评估。我们提出的自动视觉系统克服了这些限制，检测和评估所有路面损坏上述更有效，更准确。路面病害监测系统包括硬件和软件两部分。硬件是一个现成的相机，具有用于图像数据收集的某些功能。该软件是一个基于Python的代码，可以使用计算机视觉、图像处理和深度学习算法同时分析收集到的图像，以检测和评估路面损坏。生成的输出包括路面损坏类型、尺寸、严重程度和位置。拟议的产品可以安装在移动车辆的后部或无人机上。它可以在不同的移动速度下远程捕获图像，并以合理的成本和不中断交通的情况下完成评估过程。此外，该技术可以通过捕获沿着感兴趣的道路沿着的仅一个通道的连续图像来扫描所有道路车道，并且可以在白天和夜间使用。根据我们的主要市场规模估计，可服务的市场约为60，000个市政当局，机场和省级政府，占据该市场的10%相当于6，000个销售额。因此，考虑到典型的产品价值，年收入可能达到2100万美元。利用NSERC基金，我们将致力于我们的产品市场评估，以验证我们的商业模式元素，并从客户的角度出发，在产品规格和功能方面创建一个更现实、更合理的开发流程。