Hangzhou Zeno-Videopark Import Export Co., Ltd.

Amid the trend of public security informatization, and with the support of different intelligent analysis algorithm models and high-performance computing power, structuralizing and efficient extracting of video data to provide valuable information to help with public security investigation and case resolution has become an important topic for technology manufacturers focusing on video applications. The ZN-VS1308N-AS video structured data analyzer is a deep learning-based technology built by Zeno-Tech that allows real-time detection, tracking, comparison and processing of vehicles and humans in video in a low-delay, highly accurate manner.

High processing power

The Zeno-Tech ZN-VS1308N-AS video structured data analyzer employs HiSilicon’s HI3559AV100 high-performance processor, offering obvious advantages in image processing and software SDK architecture. This device supports the embedded Linux operating system and uses industrial-grade embedded microcontrollers to ensure the stable operation of hardware devices and systems. In terms of image and audio processing, it uses the mainstream H.265 and H.264 encoding and decoding technologies, and four audio compression standards: G.711a, G.711u, PCM and G.726. The hardware device supports a maximum of 8 video streams or 16 image streams. It is also equipped with 2 Gigabit Ethernet ports and supports access by two network cameras with different IP addresses. It is worth mentioning that this device is highly compatible, allowing connection to ONVIF-based third-party cameras as well as cameras from mainstream brands. In addition, it supports a built-in standard SATA port and can accommodate a 2.5-inch hard disk drive. During the test, the performance of various hardware ports was very stable.

The size of this equipment is 215mm x 160mm x 45mm with an elegant exterior that gives a “stable-as-rock” look. The environmental features are outstanding, with operating temperature range from -10°C to +55°C and working humidity from 10 percent to 90 percent, ensuring strong adaptability to different environments. Power consumption is 10W. Overall, the product is suitable for different environments, while hardware specs are excellent and cost-performance ratio is quite high.

Intelligent search and AI preview capabilities

The device has a built-in professional web system. Users can reconfigure the network, device, serial port and user management interface from the system backend and control the display interface in real time. The main page of the system interface is divided into AI preview, multi-channel preview, playback management, intelligent retrieval, intelligent management, storage management, early warning, camera management and traffic statistics. The overall interface is clear and understandable, while operation is easy and friendly. During the actual test, switching between the interfaces is smooth, and the click response is fast.

With the AI preview interface, the user can switch between video channels as needed, view video streams in real time and display structured data with regard to human body, non-motor vehicles and motor vehicles. The playback management interface allows selection of one or more video playback channels, offers flexible screen division, supports 1, 4, 9 and 16 screens, and can simultaneously play back video on multiple channels at the same time to enable convenient viewing. The traffic statistics interface uses two methods, channel-based and flow-based, to analyze total flow in a channel. The camera management interface can modify or adjust the IP address, manufacturer, model, and protocol of cameras on all lanes. At the same time, it supports manual adjustment of image brightness, contrast and saturation to achieve the best video image results.

The software backend supports real-time deployment, control and warning against various types of black and white lists, including vehicles (license plate, model and pictures), tricycles (picture), bicycles, pedestrians and faces, and allows configuration of alarm parameters. In addition, the user can select the query time and search by the name and category of deployment and control. The query result will clearly show the detailed comparison between the snapshot information and the deployment/control information. In addition, the indexing function is relatively intelligent and supports millions of dynamic database comparison and retrieval. It can retrieve images by lane, time, structure attributes and other advanced categories to search for images and play back relevant video. It also supports search by image and allows adjustment of relevant similarity threshold for retrieval.

Real-time video analysis

Through different deep learning-based intelligent algorithms, this device can detect, identify, compare, provide early warnings and do traffic analysis on faces, human bodies, vehicles and non-motor vehicles from 8 video streams or 16 image streams, and structured data retrieval is comprehensive. Human attributes include gender, age, orientation, whether or not glasses are worn, clothes texture, style and color; vehicle attributes include license plate, make, model, year, color, number plate type, plate color, driver behavior, whether they are making calls or wearing seatbelt, hazardous goods, age mark and other information.

And thanks to its excellent real-time video analysis capability, the device can analyze data from relevant video on a particular day in real time, and it can provide high-value intelligence information for case resolution, control, early warning, illegal activity inspection as well as other applications. In addition, the accuracy of detection is relatively high. According to the actual analysis, the target detection rate is >93 percent during the day and >90 percent at night; the false recognition rate is <3 percent during the day and <5 percent at night.

Product features:

1. Built-in professional web system, supports intelligent operations such as video retrieval, video playback and AI preview, which greatly simplifies the time-consuming and tedious process and is easy to use.
2. High-precision extraction of all-purpose structured data, enabling the detection, identification, comparison, early warning and traffic analysis on faces, human bodies, vehicles and non-motorized vehicles. Detection accuracy is relatively high and response is rapid, enhancing public safety’s prevention, case resolution and other law enforcement efforts.
3. Supports embedded Linux operating system and adopts industrial-grade embedded microcontroller; equipment and system run stably.
4. Usage of HiSilicon’s HI3559AV100 high-performance processor allows this product to have obvious advantages in image processing capabilities and software SDK architecture.

Expert review:

This video structured data analyzer by Zeno-Tech shifts the all-purpose structured data extraction function to the edge. Equipped with a high-performance processor and intelligent algorithm models trained according to business needs, the device can efficiently detect in real-time motor vehicles, non-motor vehicles, human bodies and faces in the video and transform them into standard semantics-based structured data. The detection response is fast and the accuracy is high. From the practical application level, it will undoubtedly and significantly improve public security deployment, control and case resolution capabilities. This NVR has a wide range of application scenarios and is mainly used for real-time search, illegal activity inspection, control and early warning by public security. It also has innovative applications in banking, education and energy industries.