The security industry is facing significant cost pressures. As demand for AI computing power grows exponentially and major storage manufacturers prioritize high-bandwidth memory, supply and demand imbalances have emerged in the storage market. This has led to tight hard drive supply and rising procurement costs, complicating large-scale deployments of intelligent security systems. Furthermore, traditional encoding systems cannot distinguish the static, low-value footage that constitutes up to seventy percent of typical surveillance content. The indiscriminate compression used by these systems results in wasted storage resources, further inflating the overall costs of security hardware procurement and data center operations.
Against this backdrop, Hangzhou Hikvision Digital Technology Co.,Ltd. (Hikvision), with over two decades of experience in security, has introduced its Guanlan encoding technology. This innovation is based on the "Guanlan Large Model," integrating scene and object understanding capabilities into the H.265 encoding framework.
According to available information, the Guanlan encoding technology employs a tripartite intelligent encoding system combining "large model scene understanding, ROI-differentiated encoding, and global adaptive bitrate allocation." This system can accurately identify high-value key targets like people and vehicles. Using refined ROI (Region of Interest) protection and segmentation, it maintains high-definition encoding for these critical foreground areas, preserving details without loss. Meanwhile, for large static background areas like walls, greenery, and sky, the video bitrate is reduced to minimize storage resource consumption.
For instance, in an office building lobby, a camera equipped with Guanlan encoding can intelligently perceive human movement. It specifically enhances the bitrate for these target areas, fully preserving details such as body contours and limb movements. This ensures the integrity of core visual information while avoiding redundant bitrate usage across the entire video stream.
Even in complex environments with mixed pedestrian and vehicle traffic, such as commercial streets or campus entrances, Guanlan encoding cameras can simultaneously parse up to 64 targets within the frame. They prioritize image quality protection for critical elements like license plates and pedestrians, ensuring that key feature information is completely retained.
Additionally, Guanlan encoding dynamically optimizes encoding resources based on video content complexity. For example, during the busy morning rush in a cafeteria, it automatically identifies key targets and allocates sufficient bitrate to restore details like facial features and movements. At night, when the scene is completely static, the camera minimizes the bitrate to the greatest extent, achieving extreme compression to reduce redundant storage space.
Based on Hikvision's internal testing, under full-day monitoring conditions, cameras utilizing Guanlan encoding technology can achieve a bitrate saving of 49.81%. Calculated for a system with 100 channels of 4-megapixel cameras storing data for 30 days, this directly saves 57.7TB of storage space, equivalent to purchasing four fewer 16TB hard drives. The advantages are even more pronounced in large-scale security projects. For a 2000-channel system at 1080P@2Mbps with 90-day storage, the Guanlan encoding solution can save 50% on the number of hard drives and 60% on data center space compared to traditional encoding, significantly reducing both storage hardware and facility operation and maintenance expenses.
In conclusion, confronting the industry realities of scarce storage resources and rising hardware investment, Hikvision has leveraged its long-term foundational technology expertise to refine Guanlan encoding. This approach moves beyond the conventional framework of indiscriminate compression, reducing redundancy right at the data generation source. It substantially lowers both the upfront hardware investment and the ongoing operational costs for security projects. Its value extends beyond merely balancing rising storage costs; through the innovative fusion of AI and encoding, it also strengthens the foundational technological base necessary for the large-scale, intelligent transformation of the security industry.
Comments