System introduction

　　Based on the current world's leading in-depth learning technology, we will capture images and videos of urban vehicles passing by, recognize vehicle vision depth, and accurately identify more than 7,000 vehicle brands, sub-brands, and annual models. At the same time, the number of annual inspection marks on the vehicle, the internal details of the driving window, the situation of the front passenger, the safety belt test, and whether there are skylights and other vehicle features are identified.

　　The system uses in-depth learning vehicle model analysis technology to extract up to 4,000 dimensions of subtle features of vehicles, and can perform large data analysis based on multi-dimensional vehicle feature information and information such as vehicle location and passing time. Applied to criminal investigation, traffic law enforcement, intelligent transportation, investigation and control, and data analysis, it is an indispensable application system for public security actual combat.

System function

Vehicle Licence Recognition

Vehicle license plate identification: common blue card, ordinary yellow card(single-layer), double-layer yellow card, new military vehicle license plate, new armed license plate, embassy license plate, agricultural license plate new energy number plate and other common types, vehicle identification rate is high: daytime recognition rate ≥ 99, Night recognition rate ≥ 98 %.

Vehicle type recognition

　　Nearly ten types of vehicles that are currently common can be identified, such as: minibuses, vans, SUVs, MPVs, pickup trucks, trucks, passenger cars, and other custom vehicle types. The identification rate reaches ≥ 99 %.

Vehicle brand recognition

　　Nearly 200 common vehicle brands can be identified, such as: Mercedes, Volkswagen, BMW, Lotus, Saab, Viscount, Fiat, etc., with an identification rate of ≥ 99 %.

Annual vehicle identification

Nearly 7,000 types of vehicle styles and sub-models that are currently common can be identified, such as: Volkswagen Jetta(X year), Volkswagen Santana(X year), Volkswagen Polaris(X year), Passat, POLO, etc.. The recognition rate reached ≥ 95 %.

Body Color Recognition

It can distinguish between dark and light-colored vehicles and can identify ten common vehicle color types. Such as: white, gray, purple, black, yellow, green, blue, red, orange, brown, the recognition rate reached ≥ 95 %.

Off-vehicle feature recognition

Can identify and accurately distinguish targets such as cars, people, and bicycles, and automatically distinguish the characteristics of the front and rear images. The characteristics of the vehicle that can be identified include: lights, car faces, rearview mirrors, light masks, window glass, skylights, luggage racks, internship signs, and other body stickers.

In-vehicle feature recognition

The annual inspection signs and environmental protection signs on the windshield of the vehicle can be identified, and the complex image features such as hangings, paper smoking, car pets, driver and co-driver heads, whether to wear seat belts, and cell phones can be identified.

Passenger Face Recognition

The face of the front passenger in the car can be located and identified. Including face location, face critical point detection, face recognition, face verification.

Technical characteristics

1、Using the original virtual number plate technology in the industry, the label management of unlicensed vehicles can provide effective management means for the public security traffic to detect cases such as unlicensed vehicle hit-and-run, unlicensed vehicle traffic violations, and illegal evidence collection.

2、Efficient recognition of complex images and vehicle patterns such as public security bayonet, and pan-bayonet.

3、Automatic extraction and identification of front and rear image features, weak light and blocking vehicles, unlicensed vehicles, defaced vehicles can still be effectively identified.

4、The original 360-degree vehicle brand year recognition technology can still be recognized under the monitoring image.