A Brief Analysis of ROI Video Surveillance Technology

　The variable quality transmission feature of the video surveillance area of ​​interest is very practical in the field of monitoring. The images collected by monitoring cameras have more or less useless areas, which also occupy transmission bandwidth and storage. The ROI function of video surveillance only transmits the video information of the area of ​​interest with high quality, and transmits less or no video that does not change, so as to improve the video quality of the scene of interest and the accuracy of video analysis. However, it is not easy for developers to implement these technologies.

　　Q: Security & Automation A: Cheng Wei from Hikvision

　　Q: What is the ROI technology of video surveillance and what are its advantages?

　　A: ROI (Regions of Interest). In actual surveillance, people tend to pay attention to certain specific areas in the surveillance image, such as license plates and faces, but not the blue sky and grass in the background. These specific areas are called "regions of interest".

　　ROI coding technology is one of the biggest highlights of the new generation static image compression coding standard JPEG2000 established by the ISO organization in 2000, and is also a research hotspot in the field of image coding today. This technology can perform lossless or near-lossless compression with a low compression ratio on the region of interest (ROI) in the image, and perform lossy compression with a high compression ratio on the background area. In this way, without losing important information, the data volume can be effectively compressed without changing the bit rate, which well solves the contradiction between compression ratio and image quality.

　　Q: What are the technical risks and difficulties in current ROI research and development?

　　A: There are currently two ways to implement ROI encoding technology: one is fixed area, and the other is dynamic tracking.

　　1. Fixed area

　　In the ROI configuration interface of the network camera, the fixed area ROI encoding function is enabled, and up to 4 regions of interest can be manually drawn on the preview image. The intensity level of ROI encoding can be set from 1 to 6.

In fixed area mode, 512Kbps bit rate

　　2. Dynamic tracking

　　First, in the face recognition configuration interface of the network camera, turn on the face detection function and dynamic analysis function, and then turn on the dynamic tracking ROI encoding function in the ROI configuration interface. The intensity level of ROI encoding can be set from 1 to 6, which can realize automatic recognition and dynamic tracking of faces in the monitoring screen, and concentrate resources on encoding the face area.

512Kbps in dynamic tracking mode

　　Note: 1. ROI supports three streams to set 4 fixed areas or dynamic tracking;

　　2. Currently, in the dynamic tracking mode of ROI encoding, only automatic detection and recognition of faces (not license plates, etc.) and ROI encoding are supported.

　　Q: In terms of cost, is the input-output ratio cost-effective?

　　A: Due to limited network bandwidth, only a low bit rate can be used to transmit video, but when you want to see the key target clearly, ROI encoding technology can sacrifice the image quality of non-interested areas and concentrate resources on the area of ​​interest to obtain key target information without increasing the bit rate, which can effectively avoid the increase in storage and bandwidth costs. ROI technology is currently supported as a "standard feature" on IPC/ball machine equipment from some mainstream manufacturers, and the hardware cost of purchasing the equipment will not increase. For users, it is a cost-effective choice in appropriate scenarios.

　　Q: What impact does the implementation of ROI function have on the industry?

　　A: ROI is not a new technology. Last year, mainstream manufacturers such as Hikvision launched this technology, which can solve the problem of insufficient bandwidth in some scenarios but still want to see the details clearly. However, this is only an improvement at the software algorithm level. With the upgrade of hardware platforms such as H.265, the bit rate can be effectively reduced, which fundamentally solves the bandwidth and cost of high-definition. It is the mainstream trend in the future, and ROI can be combined with H.265 products to further reduce the bit rate. By then, people will enjoy ultra-high-definition surveillance images with very little bandwidth and storage costs.