To help pan-security products land, Rockchip launched a complete front-end and back-end solution

The pan-security industry is far broader and has potential than imagined. Especially with the innovation of upstream technology in the security industry, security products have gradually expanded to a long series of security products covering intelligent monitoring, alarm, access control, building intercom, etc., which are suitable for various scenarios. With upstream new technologies leading the way, the pan-security industry is becoming more and more broad.

According to the forecast of the China Security Products Industry Association, the demand for basic construction of security products, the need for system upgrades and the expansion of new formats will maintain a steady growth trend in the next few years. Among them, the total revenue of the security market in 2020 will reach about 800 billion yuan, and it is expected to exceed one trillion yuan in 2022.

In the rapid development of the security market, as the key to the industrial chain, the importance of security chips has become increasingly prominent. Recently, Rockchip released complete solutions for the pan-security industry: RV1126/RV1109 two smart vision processors focusing on front-end products, RK3568 and RK3588 focusing on back-end NVR product solutions. Under the coverage of multiple scenarios, a new pan-security industry pattern is being reshaped, moving smart security products from “seeing” and “seeing clearly” to “seeing deeply”, “understanding” and “predictable”. ” of the multidimensional intellisense horizon.

To help pan-security products land, Rockchip launched a complete front-end and back-end solution

Focus on front-end security products RV1126/RV1109

On the front end, in order to meet the needs of different security scenarios and industry segments, Rockchip has newly launched two high-low matching chips, RV1126 and RV1109, both of which use 14nm process technology and have built-in smart vision processors with NPU and ISP.

The CPU of RV1126 is quad-core ARM Cortex-A7, NPU is 2.0Tops, built-in 14 million high-performance, 3-frame high dynamic ISP; RV1109’s CPU is dual-core ARM Cortex-A7, NPU is 1.2Tops, built-in 5 million high Performance, 3 frames of high dynamic ISP. In terms of video codec, RV1126 supports 4K H. 264/H. 265, RV1109 supports 5 million H. 264/H. 265, RV1109 and RV1126 hardware are pin 2 pins, and product upgrades are seamless. Has the following six advantages:

One, see clearly

RV1126 and RV1109 have technologies such as multi-level noise reduction and 3-frame HDR, which can ensure the dynamic range of the scene and meet the needs of full-color black light.

Two, listen clearly

Built-in HD AEC algorithm and support for microphone voice array, RV1126 and RV1109 make high-definition voice fully implemented, and enhance sound collection and pickup distance.

Three, takes up little space

Low bit rate, low-latency encoding, and perceptual encoding make the space ratio smaller and the user experience better.

Fourth, AI computing power is high

Heterogeneous computing, hybrid quantification, improved tools and supporting AI algorithms, and network model architecture support more abundantly, enabling more AI function expansion. At the same time, the power consumption is less than 10% of the required GPU, and supports Tensorflow, PyTorch, Caffe, MxNet, DarkNet, ONNX, etc. for direct conversion and deployment.

Five, software-defined camera

Fast Boot, dual-system architecture, open source SDK, etc., customers conduct in-depth customization according to product requirements to avoid product homogeneity.

Six, rich encryption and decryption engine

Supports TrustZone technology, multiple cryptographic engines including national secrets, and secure boot to ensure data security and integrity.

In addition to the above six points, RV1126 and RV1109 have rich interfaces, including Gigabit Ethernet interface, and TSO network acceleration; USB 2.0 OTG and USB 2.0 host; two SDIO 3.0 interfaces; 8-channel I2S with TDM/PDM, 2-channel I2S…meet the needs of various security scenarios.

Three front-end application product advantages

The application scenarios of the pan-security industry are very extensive. At the front end, it is divided into three major fields according to the product form: IPC intelligent network camera, battery IPC intelligent doorbell/cat’s eye, gate/access control/attendance products.

In view of the differences in front-end product forms, scenarios and needs, while maintaining full support, RV1126 and RV1109 also perform more precise subdivision processing to show more targeted feature advantages.

For IPC smart network camera products, RV1126 and RV1109 have three major highlights:

1. 3F-HDR technology can ensure the detailed performance of high dynamic range scenes, and rely on multi-level noise reduction technology to achieve high-quality black light full-color pictures

2.Intelligent coding technology solves industry pain points, low bit rate coding, more efficient

3.Built-in NPU, mature AI tools, rich AI ecological algorithms, stronger computing power, greatly expand application scenarios

For battery IPC, smart doorbell, and cat’s eye products, customers’ needs focus on long battery life, quick snapshots, and high-definition voice intercom. RV1126 and RV1109 have four technical advantages:

1.Support quick start, 150ms cold start quick snapshot

2.Dual systems, low-power architecture, 14nm process, and support for LP3/LP4 low-power DDR, all effectively extend the life of battery-powered security products

3.Built-in NPU to filter invalid alerts

4.Support Pre-Roll recording function

Gate / access control / attendance products have higher requirements for the accuracy of face recognition. Based on this, RV1126 and RV1109 have been specially designed to make face recognition faster, safer and more reliable. Face recognition does not wait and access control quickly becomes a reality:

1.Support dual MIPI sensor input, more sensitive perception, live verification method can prevent photo verification and improve security

2.ISP optimization for backlight and front light scenes, accurate identification in environments with poor lighting conditions

3.Rich AI algorithm ecology, expandable product application scenarios

Focus on back-end NVR + XVR products RK3568 and RK3588

Among the back-end NVR+XVR products of security monitoring, Rockchip also launched two mid-range and high-end solutions – RK3568 and RK3588.

RK3568 is equipped with 4-core Cortex-A55 CPU, G52 GPU, and NPU is 0.8Tops. The software interface kit is designed to be friendly. In accordance with the NVR industry’s large platform practice, the multimedia processing API is uniformly packaged by module, and customer application software can be quickly developed and imported into AI/AO/VO /VDEC/VENC.

1.Support 10-channel 1080P30 video decoding

2.Support TSO/UFO hardware acceleration, and at the same time powerful CPU performance, improve the access capability of network cameras

3.Built-in independent 0.8T NPU hardware accelerator; support face comparison, face search, alarm, classification of people and vehicles; support one-click conversion, support Caffe/TensorFlow/TF-Lite/ONNX/PyTorch/Keras/Darknet mainstream architecture model

4. Rich Display interface: support dual/three-screen display, HDMI2.0b, output 4K image quality; support eDP/MIPI DSI/LVDS/RGB888/BT. 1120/BT. 656

5.Powerful post-processing: support 3D-LUT/GAMMA/CSC/BCSH/DITHER/CABC; support image post-processing, denoising, color enhancement

6. Strong encoding ability: support 4K H. 264/H. 265/VP9 HD decoding in multiple formats; support ROI encoding

7.Compatible with a wide range of DDR chips, support LP4/LP4x/LP3/DDR4/DDR3, support DDR3 and DDR4 ECC, meet the high reliability requirements of industrial control and other scenarios, 32bit bit width, maximum frequency 1600Mhz, ample bandwidth design, more suitable for security Smart NVR Scenario Requirements

8.Rich peripheral interfaces: support 3xSATA3.0, dual Gigabit Ethernet ports, TSO technology, dual PCIE, to meet the expansion requirements of 4G/5G/WIFI6, multi-network ports and NPU

RK3588 adopts 8nm process, equipped with 4-core Cortex-A76 + 4-core Cortex-A55, adopts ARM’s latest 4-core GPU, and NPU is 6Tops, which can enable high-end NVR + XVR products. Insufficient power, few display interfaces, and few high-speed interfaces are like no SATA products waiting to be improved. RK3588 has the following four advantages:

1.Support 32-channel 1080P video decoding

2.Support 4 screens with different display

3. With multiple high-speed interfaces, it supports multiple PCIe3.0/PCIe2.0, SATA3.0, USB3.1 and Gigabit Ethernet.

4.With independent 6TOPS NPU super computing power, 8-core CPU + 4-core new GPU architecture

Both software and hardware upgrades make developer tools more efficient

At present, it is a new era that emphasizes the coordination of software and hardware. While developing hardware products, the tech giants are also improving their “soft power” by creating development tools, systems, etc. Rockchip is also actively promoting the dual improvement of “soft and hard”, launching more efficient and accurate products. Developer tools RKNN Toolkit.

The functions of RKNN Toolkit are extremely rich, mainly including: custom operators, model multi-input, multi-accelerator support, performance and memory estimation, simulation debugging, quantitative accuracy analysis, visualization function, model pre-compilation, model segmentation, batch inference , model conversion, quantization/hybrid quantization, model optimization switches, etc. With its own visualization features, the RKNN Toolkit makes the developer’s job easier.

In addition, the RKNN Toolkit supports two quantization mechanisms – Post training quantization, Quantization-aware training. Moreover, RKNN Toolkit uses the quantization parameters of the model itself, and there is no loss in quantization accuracy.

RKNN Toolkit and flexible multimedia development interface can increase the flexibility of developers to customize applications, make development no longer difficult, and help terminal partners to expand more AI functions and application scenarios.

Expand a new layout to empower more industries

Rockchip is committed to building multi-scenario computing unit SoCs, and has accumulated rich experience in image vision processing, video codec, SoC chip design, and large-scale system development, and has formed multi-level and multi-platform professional solutions. By digging deep into the pain points of security products and the needs of different application scenarios, Rockchip will greatly optimize the performance of terminal products and enhance the market competitiveness of terminal products by virtue of its technical advantages.

The Links:   LQ121S1LG42 G065VN01-V1