For all the emphasis on cloud systems and centralised servers at ISC West, a lot of innovation in security video systems is happening at the edge.
New advancements inside video cameras are boosting capabilities at the edge, from advancements in processing power to artificial intelligence (AI) and machine learning (ML) algorithms that can now be deployed directly on the cameras or edge devices.
Advancements in AI algorithms
The progress of video systems becoming smarter at the edge is driven by the need for real-time insights, lower latency, bandwidth efficiency, enhanced privacy, and improved reliability.
Advancements in edge computing hardware and AI algorithms are enabling a range of intelligent video applications across various industries, including physical security. Smarter functionality at the edge is a benefit of new computer systems-on-chips (SoCs) that are driving new heights of performance for today’s cameras.
Axis Communications’ ARTPEC-9
Axis Communications’ new ARTPEC-9 SoC offers advanced video compression to reduce bandwidth
Axis Communication’s new ARTPEC-9 system-on-chip (SoC) offers advanced video compression to reduce bandwidth and storage needs. With a low bitrate, the SoC helps deliver high-quality imaging with outstanding forensic detail.
ARTPEC-9 also offers enhanced deep learning capabilities to allow users to leverage the latest video analytics and accelerate the implementation of AI technology.
Axis maintains control over all aspects of the chip’s development to ensure high quality and cybersecurity. Among the benefits of ARTPEC-9 are better AI and deep learning, better image quality, better cybersecurity, and AV1 license-free video compression (see below).
Hanwha Vision’s Wisenet 9
Hanwha Vision has launched Wisenet 9, its most advanced AI-powered System on Chip (SoC). Wisenet 9’s enhanced edge AI capabilities increase performance as the volume and complexity of security threats demand real-time, accurate analysis. By elevating edge-device performance, AI empowers systems to quickly analyse vast amounts of video data and discern crucial patterns and anomalies.
A key differentiator driving Wisenet 9 is deployment of two Neural Processing Units (NPUs), which improve performance three-fold compared to Wisenet 7, the previous SoC generation. While one NPU handles image processing, the other focuses on object detection and advanced analytics. This dual NPU concept was introduced to ensure video quality and analytics have independent resources, thus preventing one function from impacting the performance of the other.
The latest from Ambarella
Ambarella is a supplier of edge AI systems-on-chips to multiple video camera manufacturers
Off the ISC West trade show floor in a nearby meeting room, semiconductor company Ambarella demonstrated how it will continue to push the envelope of what is possible with generative AI at the edge. Ambarella is a supplier of edge AI systems-on-chips to multiple video camera manufacturers and recently achieved the milestone of 30 million cumulative units shipped.
The demonstrations highlight Ambarella’s ability to enable scalable, high-performance reasoning and vision AI applications across its ultra-efficient, edge-inference CVflow 3.0 AI SoC portfolio. The company’s DeepSeek GenAI models run on three different price/performance levels of its SoC portfolio. In addition to advancements in GenAI processing at the edge, Ambarella integrates image processing, encoding and system-level functions into all its AI SoCs.
New standard for video encoding: AV1
AV1 compression is a next-generation video coding technology that offers significant improvements in compression efficiency and video quality, especially at lower bitrates. Its royalty-free nature positions it as a crucial codec for the future of internet video. AV1 compression is a next-generation video coding technology.
Axis Communication’s ARTPEC-9 chip now supports the AV1 video encoding standard. By embracing this standard, which is new to the physical security market although it was introduced in 2018, Axis sets the stage for AV1 compression to eventually become the industry standard, replacing H.264 and H.265.
Network video transmission
AV1 is an open-source, license-free coding format designed mainly for efficient network video transmission
AV1 is an open-source, license-free coding format designed specifically for efficient network video transmission. It delivers high-quality video at low bitrates, reducing bandwidth consumption and storage costs. The codec was developed by the Alliance for Open Media (AOM), a nonprofit organisation founded in 2015 by Google, Intel, Amazon, Microsoft, Netflix, and Mozilla (among others), to provide open-standard, next-gen video coding technology.
AV1 is ideal for cloud solutions—making streaming applications more robust, scalable, and capable of delivering real-time insights. Now the ARTPEC-9 chipset brings these benefits to the surveillance industry, and AV1 is currently supported by AXIS Camera Station. Providers of major video management solutions (VMS) such as Genetec and Milestone will be adding support for AV1, with further developments already underway.
More intelligence at the edge
Intelligence inside video cameras comes from the processing power and algorithms that enable them to perform tasks beyond simply capturing and recording images. This "intelligence" allows cameras to analyse the video stream in real-time, identify objects, detect events, and make decisions or provide alerts based on what they "see."
New and improved SoCs are driving performance improvements at the edge. The increasing power of embedded processors and advancements in AI are continuously expanding the capabilities of intelligent video cameras.
