Product Overview
Centering on video in-depth application, the TIANYAN·Video Structured Server is deployed in the video surveillance center. It is capable of connecting to the streaming media server of the existing surveillance platform or directly connecting to the front-end camera to receive real-time video streaming (supporting offline video resources simultaneously), carry out multiplex concurrent analysis and structured recognition, and store structured information and captured images in the data engine and file system separately for efficient retrieval by users.
Product Appearance
Technical Features
High index & performance
The deep learning framework is adopted to take the recognition item and rate of subdivided features of motor vehicles/pedestrians/non-motor vehicles to the industry’s leading level.
Stand-alone 2U server is capable of processing 40-path full target analysis concurrently.
Strongly compatible
Capable of connecting to the platforms, front-end and storage of mainstream manufacturers.
Stand-alone is capable of processing online/offline video and image resources.
Highly integrated & expandable
Stand-alone supports the structuring of target and inquiry into structured information simultaneously; and supports the application of reverse image search for the targets of automobiles, pedestrians and non-motor vehicles.
It is flexibly deployed and supports cluster expansion and the elastic expansion of calculation and storage resources; and it is equipped with smart load balancing and unified resource management.
Easy to integrate
A rich array of second development interfaces are in place to provide algorithmic support for the third-party manufacturers.
Functional Features
Data retrieval
Vehicle feature recognition:
It is capable of recognizing the subdivided vehicle features in the captured image, such as license plate, type, model, body color, annual inspection sticker, sunshade, tissue box, pendant, ornament, failure to wear a seatbelt, and using phones while driving. It supports 16 subdivided vehicle models, more than 200 main vehicle brands, over 8,000 sub-brands and annual models.
Personal feature recognition:
Supporting the following features of human body:
1. Gender, nationality (Han nationality or minority) or age;
2. Head: glasses, hat, helmet or mask;
3. Upper body: color or texture;
4. Lower body: color or classification;
5. Belongings: handbag, shoulder bag/messenger bag, backpack, suitcase, trolley or goods carried.
Feature recognition of non-motor vehicles:
Supporting the classification, color or posture of non-motor vehicles;
Supporting the gender, head features, belongings or the color of tops of the driver
Reverse image search
It supports the reverse image search for pedestrians, automobiles, and non-motor vehicles directly in search results or uploaded images.
Reverse image search on pedestrians:
Analysis into violations of laws
Failure to wear a seatbelt:
Supporting the screening of vehicles in which the main and second drivers fail to wear a seatbelt based on time and place information
Facial occlusion:
Supporting the screening of vehicles involving driver’s facial occlusion based on time and place information
Using phones while driving:
Supporting the screening of vehicles in which drivers use phones while driving based on time and place information
Performance indexes
Real-time video analysis
The stand-alone equipment supports the real-time analysis of full targets (motor vehicles, non-motor vehicles, and people) on the 40-path 1080P HD video at most.
Previous video analysis
For full targets (motor vehicles, non-motor vehicles, and people), the stand-alone equipment supports the 40x speed of analysis into 1-path 1080P HD previous video or 1x speed of analysis into 40-path 1080P HD previous video at most.
