Beijing's Xicheng district has implemented an artificial intelligence-powered system to streamline street inspections, allowing for more efficient monitoring of its over 1,400 alleys and streets.
This new approach uses specially designed backpacks equipped with AI to automatically identify and report urban management issues.
Instead of traditional manual inspections, patrol officers now carry a discreet backpack weighing just over 3 kilograms. The backpack, equipped with a protruding sensor, houses a sophisticated AI system capable of rapidly assessing the condition of narrow streets and alleyways.
The system features a 48-megapixel ultra-high-definition panoramic camera that captures 12K panoramic photos and 8K panoramic videos in real time, even at speeds up to 60 kph. This allows inspectors to cover more ground on foot, by bicycle or in a slow-moving vehicle.
The integrated AI technology uses multimodal video recognition and various large model algorithms to analyze images. It can automatically identify environmental problems like improperly placed items, littering, damaged facilities and incorrect waste sorting. The system achieves over 90% accuracy for common issues and averages over 65% across 280 indicators. This eliminates the need for inspectors to take manual photos and write reports, significantly boosting efficiency.
Community workers now receive automatic notifications from the AI system immediately after inspections. These include panoramic videos, problem images, and precise QR codes for location tracking that are accurate to the centimeter. The system can even measure data such as unpleasant smells.
An official from the Urban Management Committee of Xicheng District of Beijing said the AI system has drastically reduced the workload for front-line workers. Previously, tasks like downloading case files, locating problem areas and coordinating rectification could take at least a full day. The new system not only reduces this burden but also continuously improves its accuracy and efficiency.
Go to Forum >>0 Comment(s)