Hangzhou City Brain Releases AI Vision Products - part 1

On April 8th, the City Brain released the AI product Scorpio, upgraded the monitoring dome intelligent perception system, and provided the city's large-scale real-time cameras with real-time analysis capabilities through automatic tracking recognition and intelligent perception technologies. The use efficiency is further broken through. The traffic accident can be found in only 20 seconds and the accuracy rate was over 95%. It truly replaces traffic police patrols with machine vision.

In order to solve the problem of traditional traffic patrols, Hangzhou City Brain attaches great importance to video intelligent analysis technology and will become the eye of the city's running state in the future. In the fields of target detection, tracking, scene segmentation and behavior identification, they have achieved a number of technological breakthroughs and innovations. It is like a huge smart gsm alarm system at your home. In particular, in pedestrian and non-motor vehicle testing, the Hangzhou City Brain machine vision team has finally launched the mature vision product Scorpio.

20 seconds recognition with an accuracy of 95% or more

In terms of the scene, "Scorpio" possesses comprehensive traffic incident identification capabilities, autonomously identifies different scenes such as intersections, road sections, motor vehicle lanes, non-motor vehicle lanes, and crosswalks in urban roads, and accurately identifies motor vehicles from dynamically changing mixed traffic. Traffic participants such as non-motorized vehicles and pedestrians can realize automatic recognition of various types of traffic anomalies such as congestion, reverse driving, non-motorized vehicles and motorized lanes, and can adapt to the working environment at night in low light conditions. The system can detect traffic jams in a timely manner, calculate the duration of congestion, and discover traffic accidents and non-motorized illegal driving.

In terms of immediacy and accuracy, Scorpio took only less than 20 seconds from the event to the alarm, and the accuracy rate was over 95%, providing strong support for early detection and early disposal.