Next-generation smart lock design integrating visual intelligence and security

Next-generation video smart locks integrating camera and lock into a single device

As homes and businesses place increasing emphasis on security, smart locks have become a vital component of modern access control systems. These devices surpass traditional keyless entry methods by integrating wireless connectivity, mobile app control, and biometric authentication to deliver a seamless and secure user experience. While many existing systems rely on separate doorbell cameras for visual verification, a new generation of video smart locks is redefining this category by merging the camera and lock into one intelligent device.
 
A video smart lock is an innovative access control solution that integrates a camera, a locking mechanism, and smart functionalities into a single device. Traditional smart locks typically offer only keyless unlocking methods (such as fingerprint, passcode, or mobile app unlock). In contrast, next-generation video smart locks further incorporate visual monitoring and intelligent analysis capabilities, transforming them from mere access controllers into devices capable of doorstep surveillance and event awareness.
 
Video smart locks can perform real-time visual monitoring and visitor identification, multi-modal identity verification, remote visitor management and alert notifications, and integration with smart home ecosystems. They are suitable for various scenarios such as home security management, short-term rentals and vacation homes, and office and commercial spaces.
 
These integrated solutions offer more than just convenience. They enable remote monitoring, real-time access logs, and tamper alerts, empowering users to manage property security from anywhere. Whether it's enhancing accessibility or protecting privacy through edge-based image processing, video smart locks represent a significant leap forward in both functionality and security.
 
This new generation of video smart locks achieves the convergence of access control and surveillance functions. It not only enhances the security of entry points and visitor recognition capabilities but also provides more intuitive, controllable, and intelligent access management solutions for modern smart living scenarios. However, challenges such as power supply and battery life management, as well as privacy and data protection, still need to be considered.

Key considerations for selecting image sensors in video smart locks

Video smart locks depend on high image quality, as the accuracy of subsequent actions directly impacts safety and potential property loss. These products and their features incorporate a significant amount of Artificial Intelligence (AI) and Machine Learning (ML) technology. Poor image quality can overload the processing engine, draining power and delaying operations.
 
Optimizing resolution is also crucial to ensure captured images contain sufficient detail for the system to distinguish individual features and take appropriate action. Lower resolutions lack detail and can lead to errors in recognizing people or objects.
Smart locks are typically battery-powered, as providing wired power to the lock area is often impractical or impossible. Low-power devices and systems are highly desirable for extending battery cycle life and ensuring reliable, uninterrupted operation.
 
Smart locks have strict size constraints. They must be compact, with core functions encompassing the locking mechanism and the doorbell camera. Modern electronic systems in these locks are densely packed, containing multiple sensors, a processing engine, connectivity modules, and a battery pack. Consequently, all these components must be incorporated into a highly optimized design that prioritizes not only functionality and reliability but also overall form factor.

Hyperlux LP image sensors with High quality and ultra-low power

The Hyperlux™ LP image sensor family is architected for ultra-low power consumption. Products like the AR0544 and AR0830 lines cover three Color Filter Array (CFA) variants: Mono, RGB, and RGB-IR. Their silicon and pixel architecture is optimized to deliver superior image quality during both day and night, thereby minimizing reliance on high-power Infrared (IR) LEDs. Features such as Wake-on-Motion (WoM) with configurable regions of interest, trigger modes, and inherently ultra-low power consumption contribute to maximizing operational duration on a single charge or battery replacement. This significantly reduces the overall power consumption of the smart lock system and lowers the Bill of Materials (BOM) cost.
 
The RGB-IR variant facilitates easy generation of 1 Megapixel or 2 Megapixel IR images, providing rich depth data - an ideal choice for smart locks with integrated IR pixels. Incorporating IR pixels into the image sensor helps eliminate spoofing attempts using 2D photographs. By scanning an object with structured light, contour variations can be recognized to determine depth, enabling the construction of a 3D map which also assists in detecting the liveness of the object.
 
Another advantage of such an image sensor is the availability of IR data during normal operation, which can be used to correct IR bleed - an inevitable phenomenon when smart locks face outdoor environments. This ensures high-fidelity images and a true representation of the scene.
 
onsemi's Hyperlux LP family offers diverse specifications. The AR0544 is a 1/4.2-inch Back-Side Illuminated (BSI) Stacked CMOS digital image sensor with an active-pixel array of 2592 x 1944. The AR0830 is a 1/2.9-inch BSI Stacked CMOS digital image sensor with an active-pixel array of 3840 x 2160. Both advanced sensors capture images in either linear or enhanced Dynamic Range (eDR) modes with a rolling-shutter readout. The architecture and design of the AR0544 and AR0830 aim to deliver high-quality performance while consuming minimal power. Supported by onsemi's 1.4 µm BSI pixel technology, they provide excellent performance in low-light and NIR wavelengths, offering a powerful edge data capture device programmable through various options to deliver optimal bandwidth at very low power.
 
The AR0544 and AR0830 feature the lowest power consumption among comparable offerings in the industry, enhanced NIR response at 850nm & 940nm wavelengths, and Wake-on-Motion (WoM) functionality. They support multiple subsampling modes including Binning, Summing, Skipping, and Windowing, and offer various functional modes such as Global Reset Release (GRR), Trigger, and Context Switching. They also feature enhanced Dynamic Range (eDR) and Line Interleaved HDR (LI-HDR) modes.

A comprehensive next-generation smart lock reference design

To realize next-generation smart locks, onsemi has collaborated with SoC/FPGA OEM partners. In this collaboration, onsemi, together with AMD and MakarenaLabs, has introduced a hybrid 2D/3D smart lock reference design utilizing onsemi's AR0830 image sensor. This RGB-IR image sensor captures both 2D and 3D data, which is fed into processing and machine learning engines to enable face detection and recognition.
 
The 2D data is input into AI models to analyze whether a human face is present and verify against a database if the person is an authorized user. The 3D image is key to determining if a real person is standing before the camera, helping prevent spoofing with 2D images. IR 3D data from the AR0830 and a VCSEL projector is used to construct a point cloud of the real human face, aiding in 3D facial reconstruction. The door unlocks only after both 2D and 3D information are successfully verified.
 
This reference design provides a rapid development pathway for smart locks before committing resources to production. onsemi's image sensors are readily integrated with the necessary hardware and software for designing these smart locks. Smart locks are evolving into integral components of home and business smart ecosystems, blending security, technology, and user experience into intelligent solutions.
 
As biometric authentication becomes more accurate and accessible, the adoption of smart locks as next-generation security infrastructure is growing. Image sensors play a vital role in capturing the data required for AI-driven facial and behavioral detection and recognition. Lock providers can now focus on differentiating their applications by leveraging the capabilities of image sensors, SoCs/FPGAs, and AI/ML models to infuse intelligence into their solutions.
 
The Machine Learning solution in the smart lock design reference combines 2D and 3D information processing to identify specific individuals attempting to access a restricted area. The system is based on a Machine Learning engine that uses AI models to extract features from both 2D and 3D acquisition data. The 2D and 3D branches each have their own dedicated AI model for feature extraction. The extracted information is then fused to compute the authorization process.
 
The 2D branch operates two different AI models in a cascade: first Face Detection, then Face Recognition. The Face Detection model analyzes an image from the video source and extracts bounding boxes for any faces present. These bounding boxes are cropped versions of the original image containing only the face region. The Face Recognition model takes this cropped face image as input and returns a facial descriptor as output. It determines if the person in front of the smart lock matches an authorized user stored in the database.
 
The 3D branch is designed to prevent spoofing of the 2D system. Its function is to determine whether the 3D mapping of the person in front of the smart lock corresponds to a real human face. This rejects attempts using images, photographs, or other methods that might deceive the 2D system. This is achieved by building a point cloud of the real human face using IR data from the onsemi AR0830CS sensor and a dot projector (ams OSRAM BELAGO1.1 VCSEL dot projector), enabling infrared 3D facial reconstruction.
 
The hardware for the SmartLock solution utilizes the AMD Kria™ KV260 development board, designed for advanced vision application development without requiring deep hardware design expertise. It is paired with an RGB-IR camera based on a solution from onsemi and a VCSEL dot projector from ams OSRAM. This dot projector is a VCSEL (Vertical Cavity Surface Emitting Laser) projector, essential for supporting the reconstruction of the real human face. Additional accessories include a MicroSD card, a MicroSD to USB adapter, a Full HD HDMI monitor and cable, and a keyboard and mouse set.

Conclusion

Examining the development trends of next-generation smart locks, the integration of visual intelligence with multi-layered security mechanisms has become pivotal for enhancing overall protection capabilities and user experience. Through the deep integration of image sensing, edge AI recognition, and security algorithms, smart locks have evolved from passive switching devices into intelligent security nodes capable of active perception, real-time judgment, and risk warning. Looking ahead, with continuous optimization in low-power design, privacy protection mechanisms, and system interconnectivity, onsemi's smart lock solutions will play an increasingly central role in residential, commercial, and public spaces, providing a new generation of access control solutions that balance security, reliability, and intelligent management for smart environments.