As scenarios like access control, time attendance, smart parks, smart retail, and rail transit fully enter the “face recognition era,” the market demand for high-cost-performance, customizable AI cameras continues to grow. Rockchip’s RV1126B, a high-performance SoC designed for machine vision and intelligent security, has become a mainstream platform for edge-based facial recognition cameras. Smartgiant’s NeoCAM series AI camera/core board solution, built on the RV1126B, features deep optimizations in hardware, algorithms, and system integration, serving as an ideal standard platform and customization starting point for face and person recognition devices.

Industry Trend of “Face Recognition Cameras” Moving to Edge Computing

The edge AI camera market is currently expanding rapidly. The industry is transitioning from “front-end capture, back-end recognition” to an edge architecture where cameras possess built-in AI capabilities. The main reasons are: enhanced privacy and compliance requirements necessitate local processing of facial data; higher demands for real-time response, where uploading only results reduces bandwidth and latency; and the increasing scale of project deployments, making edge devices easier to install, maintain, and scale. This trend makes edge AI chips like the RV1126B standard equipment, giving vendors with comprehensive, integrated hardware and software solutions a greater market advantage.

Core Performance Advantages of RV1126B AI Camera

Given this trend, a chip truly suitable for front-end facial recognition must strike a balance between computing power, image quality, and power consumption. The RV1126B is precisely designed around this goal for “edge face/person recognition cameras”:

• Processor: Quad-core ARM Cortex-A53processor for running Linux, application logic, and protocol stacks.
• NPU: Approximately 3.0 TOPS of neural network computing power, accelerating deep learning models for face detection, face feature extraction, human detection, etc.
• ISP: Supports up to 12+8MP image processing with features like HDR, 3A (AE/AF/AWB), 2D/3D noise reduction, distortion correction, and fisheye correction. This ensures reliable face capture in complex environments like backlit entrances, brightly lit lobbies, and nighttime passages.
• Encoding/Decoding: Supports 4K@45fps H.264/H.265 encoding/decoding and multi-stream output. One high-resolution main stream is used for evidence storage, while one or more sub-streams are allocated for preview and platform analysis.

For typical access control, attendance, retail, and branch scenarios, 1080p@25/30fps is sufficient. This means that in practical applications, the RV1126B retains ample computing headroom for additional functions like liveness detection and multi-target tracking.

Modular Hardware Configuration Solution for RV1126B

1.  With the right “chip,” suitable “eyes” and a “housing” are needed for the complete device to truly meet diverse scenario requirements. Smartgiant adopts a modular architecture based on the RV1126B, consisting of a main board, sensor board, PoE board, and housing. This allows a single main control platform to flexibly work with various sensors and device form factors, providing customers with a highly scalable and customizable combination solution.

    Regarding sensor configuration, the solution supports mainstream sensors like the IMX327, IMX335, IMX415, and IMX678, covering different resolutions and application scenarios:

    GC2093 (2MP):  Excellent low-light performance, suitable for indoor access control, time attendance, and close-range recognition scenarios with high night vision requirements.

    IMX335 (5MP):  Higher resolution and wider field of view, ideal for lobbies, wide-area monitoring, and environments requiring preserved facial details.

    IMX415 (8MP 4K): Targets ultra-high-definition images and detailed recording, often used in high-end access control, commercial spaces, and精细分析 (fine analysis) scenarios.

IMX678 (4K STARVIS 2): High sensitivity and wide dynamic range, enabling stable and accurate recognition under complex lighting conditions like backlighting and high contrast.

    By pairing different sensors with lenses of various focal lengths and apertures, the platform can flexibly adapt to various application modes, such as close-range fast passage, mid-range multi-person simultaneous recognition, and wide-area large-scene people counting.

2.  In terms of physical form, the platform supports multiple housing options suitable for indoor and outdoor scenarios:

    Bullet Style: Suitable for outdoor wall mounting or pole mounting, providing ample space for heat dissipation and supplementary lighting, with stronger environmental adaptability.

    Conch / Dome Style: Compact structure and clean appearance, often used in indoor scenarios like office buildings, shopping malls, and subway stations.

    Through unified interface specifications for the main board and sensor board, different housing styles can be freely combined with multiple sensors. This allows brands to quickly launch multiple models and grades of finished products without modifying the core solution.

Engineering Deployment Advantages of RV1126B AI Camera

When a product moves from a lab demo to actual engineering projects, power supply, network connectivity, and installation complexity often determine success or failure. Smartgiant provides a mature PoE solution on the RV1126B platform, supporting standard Power over Ethernet. A single network cable handles both power and data transmission, making it ideal for retrofit projects and large-scale park deployments.

Furthermore, features like network port surge protection, power isolation, and short-circuit protection enhance the device’s reliability in complex electromagnetic environments. The pigtail cabling can be customized according to project habits, including DC plug type, reset button, and cable length, facilitating construction and subsequent maintenance. For projects requiring the unified deployment of hundreds or thousands of devices, such as access control, campus security, and commercial chain stores, these “detailed designs” significantly impact overall operational and maintenance costs.

Scenario-based Applications of Face Recognition + Person Recognition

Focusing on the core object—”people”—the RV1126B AI camera can create clear value in multiple scenarios:

1.  Access Control & Time Attendance: Face Recognition Terminals for High Passage Efficiency

    In access control and attendance scenarios, the camera needs to complete the entire process—from face detection, capture, and feature extraction to local 1:N comparison—and issue an unlock or record attendance command within an extremely short time. The RV1126B’s NPU maintains smooth recognition even in scenes with multiple people, while the ISP ensures accurate face exposure in complex environments like backlit entrances or glass lobbies. At night, IR illumination coupled with an IR cut filter enables clear imaging. Additionally, front-end cameras support offline whitelist storage and local comparison, ensuring basic passage and recording even during network fluctuations or short interruptions. This is particularly crucial for sites with unstable network conditions, such as construction sites or factories.

2.  People Trajectory & Passenger Flow Analysis: From “Seeing” to “Understanding”

    In scenarios like station passageways, mall entrances, or campus main roads, the RV1126B AI camera can not only recognize facial identities but also combine human detection and Re-identification (ReID) technology to perform structured analysis of people’s appearance frequency, dwell time, and movement paths. This data can be used for:

    *   Passenger flow statistics and heat map analysis in shopping malls and chain stores.

    *   Passage efficiency evaluation and optimization in stations and campuses.

    *   Entry/exit alerts and trajectory (tracking) for persons of interest.

Since most computation is done at the edge, only structured metadata needs to be uploaded, making back-end system pressure and bandwidth usage more manageable.

3.  Smart Retail & Financial Branches: Connecting “Recognition” to “Service”

    In smart retail and financial branch scenarios, the RV1126B edge camera can handle tasks like VIP recognition, customer segmentation, and queue monitoring. By integrating with CRM or business systems, recognition results can trigger welcome screens, personalized service prompts, or optimize teller scheduling. These applications typically demand high image quality and recognition accuracy while being highly sensitive to terminal device costs and installation/maintenance complexity—aligning perfectly with the RV1126B + selectable sensor + PoE (integrated) product form factor.

Smartgiant‘s One-Stop Customization Capability for Software and Hardware

To truly realize these scenarios, the key lies in a comprehensive solution combining hardware and software, going beyond just the chip and module. Smartgiant has built a complete product and service chain around the RV1126B:

• Hardware Layer: Main board, sensor boards supporting multiple sensors (IMX327/335/415/678), PoE board, and various engineered housings (bullet/dome).
• System Layer: Fundamental functions like sensor drivers, ISP tuning, encoders, network communication, and PoE management are pre-adapted and ready-to-use upon delivery.
• Algorithm Layer: Can integrate algorithms for face detection, face recognition, human detection, and passenger flow analysis, provided via SDK/API for easy integration with customer-specific business logic.
• Platform Layer: Supports standard protocols (like ONVIF/GB28181) and allows customization of lightweight HTTP/SDK protocols per project, facilitating integration with existing access control platforms, attendance systems, or SaaS services.
• Security & Maintenance: Supports secure boot, encrypted firmware updates, and remote maintenance, ensuring the device remains secure, iteratively improvable and manageable throughout its lifecycle.

This allows brands and solution integrators to focus more effort on product definition, industry solutions, and market expansion, while delegating underlying hardware, software, and complex integration to a mature platform and team.

How to Build Your Own AI Face Camera Step-by-Step Using the RV1126B

If you have a clear application direction (e.g., access control, attendance, retail passenger flow, or branch recognition), you can take the first step by following this logic:

1.  First, Define Your Application and Installation Scenario

    Indoor or outdoor?

    Is it close-range recognition (0.5–1 meter) or medium-range/wide-angle recognition (2–5 meters)?

    Is the focus “fast passage” or “detailed capture and statistics”?

2.  Next, Select the Appropriate Sensor and Housing Combination

    Low-light Priority: Favor GC2093 / IMX678, suitable for weak light and nighttime scenes.

    Resolution & Detail Priority: Favor IMX335 / IMX415 / IMX678 for wide-angle and high-precision scenarios.

    Indoor Aesthetics Friendly: Choose Dome/Conch style; **Outdoor Installation Flexibility:Choose Bullet style.

3.  Finally, Engage in Hardware/Software Customization and Platform Integration**

   Quickly run a Proof of Concept (PoC) using the general firmware and SDK.

   Customize interfaces, UI, protocols, and branding based on project needs, expanding to a multi-model product line.

   Throughout this process, Smartgiant provides comprehensive support from samples and evaluation boards to small-batch trials and mass production, helping you accelerate your product launch while managing risk.

You can learn more and start your project through the following channels:

*   Visit the Smartgiant official website to explore RV1126B AI cameras and related core board/complete device solutions: Official Website: `https://www.smartgiant.com`

*   Summarize your application requirements (e.g., “Face Access Control Terminal,” “Mall Passenger Flow Statistics Camera,” “Branch VIP Recognition Camera”) into key points, including the required sensors (GC2093/IMX335/IMX415/IMX678, etc.), housing solution, required interfaces, power supply method (PoE/DC), platform integration needs, face/person recognition algorithms, SDK, and subsequent upgrade/maintenance plans. Contact the Smartgiant sales and technical team: `info@smartgiant.com`

Starting from defining your scenario and selecting the sensor and housing, all the way through hardware/software customization and mass production, the RV1126B Rockchip platform combined with Smartgiant’s AI camera solution can help you build stable and differentiated own-brand products in less time during this “face recognition era.”