Back to Blog

Nvidia Jetson/Orin +FPGA+AI High-Compute Edge AI Box: Smart Road Inspection Solution

#人工智能#边缘计算

Project Background

China's road network is vast, with a total mileage exceeding 5.35 million kilometers, and its expressway mileage ranks among the highest in the world. Facing the continuous growth of infrastructure stock, highway maintenance and management has entered a new era of an integrated approach to construction, management, and maintenance. As maintenance expenditures increase year by year, improving maintenance efficiency and reducing management costs have become urgent issues.

Challenges Faced

  1. High-Frequency Inspections and Labor Costs: Daily inspection work is arduous, leading to high labor costs.

  2. Data Collection and Management: Manual data collection and recording are highly subjective, inefficient, and pose safety risks to personnel.

  3. Limitations of Annual Inspections: While reliable, annual inspections are costly and infrequent, failing to meet real-time demands.

Policy Directives

The "14th Five-Year Plan for Highway Maintenance and Management Development Outline" emphasizes promoting facility digitalization and enhancing the digital level of maintenance management, especially AI-based automated inspections and IoT-driven maintenance project quality management.

Solution

For such application scenarios, Xinmai Technology, in collaboration with a software company, jointly developed a lightweight smart highway inspection solution. The inspection equipment and system collect road surface image data via vehicle-mounted cameras; utilize edge computers for AI image analysis to efficiently and accurately identify road safety risks and defects; upload images via object storage HTTP(s) protocol interfaces; upload metadata via MQTT protocol; simultaneously, combining 5G/4G communication and high-precision positioning technology, data is uploaded to the cloud in real-time, providing comprehensive and real-time assessment and display of road conditions. In cases of poor 4G/5G network conditions, the system caches data and resumes transmission later. The device has a built-in media server, allowing mobile apps to access and play real-time inference videos via RTSP protocol.

Technical Advantages

  1. High Efficiency: AI intelligent analysis improves inspection efficiency by 10 times;

  2. High Accuracy: Detection rate > 91.6%, false positive rate < 5.4%, area calculation error < 8.6%, ensuring data accuracy;

  3. Real-time Response: Real-time detection, real-time reporting, second-level response;

  4. High Reliability: Provides mechanisms such as caching and resumable transmission, ensuring secure and stable data transmission regardless of network environment limitations;

  5. Remote Controllability: Built-in VPN and mirrored deployment enable remote upgrades and maintenance.

  6. Rich Algorithm Models: Supports integrated management of multi-scenario, multi-version AI algorithms, including but not limited to: road debris detection, road water accumulation and icing detection, slope rockfall monitoring, roadside weather observation, and road asset inventory.

Customer Benefits

  1. Comprehensive Road Condition Overview: Real-time monitoring and assessment of road conditions through daily inspections.

  2. Leap in Inspection Efficiency: Significantly improves work efficiency, reduces labor costs, and mitigates safety hazards.

  3. Real-time Risk Warning: Real-time detection of road risks through vehicle-mounted edge intelligent analysis, allowing for proactive preventive measures.

  4. Scientific Decision Support: Provides scientific decision-making basis for maintenance management based on accumulated daily effective data, reducing maintenance costs.

Implementation Results

The lightweight smart highway inspection solution has been successfully implemented in 12 provinces and cities nationwide, gaining widespread recognition from customers.