Using AI technologies and an integrated platform, Dahua Technology‘s Smart City Solution aims to address safety issues in big cities.
To be more specific, video surveillance using deep-learning performs timely and effective risk detection. The unified platform allows faster emergency response. Data mining helps bring more insights from the front-end systems.
Dahua says that the deep-learning in its Smart City Solution redefines the capabilities of video analysis. Dahua reports that it achieved 99.78 per cent facial recognition accuracy in LFW (Labeled Faces in the Wild, a facial recognition benchmark for academic purposes) in October, 2016. Deep-learning allows classification of people and vehicles appearing in video, and capturing details such as clothing colour, gender, headgear, license plate number, vehicle coluor, size, make and model, cellphone use, and seat belt use in a vehicle.
Dahua says that its Smart City Solution can be viewed in four parts: prevention, detection, response and investigation. By improving each, you can improve the city in general.
Prevention means picking out and focusing on potentially dangerous persons before they do harm to public safety. This requires the collection and analysis of big data. The solution provides forecasting models that give a better chance of predicting events. For example, with analysis of violation data collected by traffic enforcement cameras, the system can generate a watch list for the vehicles with a record of violations.
Detection means real time surveillance in various forms, including deployment of 360-degree panoramic camera and all-weather monitoring thermal cameras. Besides multifunctional cameras, the deep-learning in smart video surveillance can automatically detect risky individuals and activities from the massive data collected. Dahua adds that its traffic enforcement system helped its home city of Hangzhou automatically detect 29,823 vehicles’ illegal behavior during the G20 Summit it hosted in mid-2016, which greatly reduced the burden on policemen who thus could focus on protecting the area hosting the summit.
Response means enabling key organisations in the city to react to security threats in real time, against possible escalation of an incident. Dahua says that it improves communication efficiency through its integrated platform, allowing unified command and central data storage as well as cross-department information sharing. Suppose a natural disaster took place in the city; the Smart City Solution could would report it on a single e-map with available emergency response resources to all relevant departments and agencies, while providing monitoring of the site to all units approaching.
Investigation means being able to quickly analyse all data collected by systems and provide insights to aid the recovery of video of an incident such as a crime after it happened. Technologies such as video synopsis provide a short summary including all moving objects of a long video and supports searches of a moving object according to such characteristics as size, colour, direction and speed. There are other data mining methods like active area analysis, active target analysis and trajectory analysis.
