English
English
Views: 0 Author: Site Editor Publish Time: 2026-10-10 Origin: Site
Smart cabinets are moving from a simple automated vending format to a new stage that relies on visual recognition for seamless payment and intelligent operations. A consumer opens the cabinet door, takes out products, and leaves after closing the door. Behind this smooth experience are rigid requirements for the vision solution: the image must be clear enough for the algorithm to accurately identify the category and quantity of each item; transmission latency must be low enough for the checkout action to complete within seconds.
When product packaging is similar, items are densely arranged, or lighting is complex, misrecognition and missed recognition easily occur, directly affecting user experience. Upgrading the visual acquisition end has therefore become an unavoidable issue.
Our company, an officially authorized dealer of Sony (China) and an HD video secondary development solution provider, points out that the combination of the Sony FCB-EW9500H and the 4KP60 AI Network Interface Board is a noteworthy and efficient solution for smart cabinet visual recognition scenarios. The former provides 4K/60p high-resolution image acquisition capability, while the latter encodes and transmits at a high frame rate of 60 frames per second. Together, they can meet the visual input requirements of smart cabinets in two key dimensions: clarity and smoothness.
First, image acquisition clarity from the perspective of hardware performance.
The recognition algorithm of a smart cabinet depends on high-quality image input. The interior space of a cabinet is limited, and cameras usually need to cover multiple shelves within a certain distance. Only at high resolution can small text, barcodes, and pattern details on product packaging be stably extracted by the algorithm.
The Sony FCB-EW9500H uses a 1/1.8-inch type STARVIS CMOS sensor, has approximately 4 million effective pixels, and supports 4K@60fps output. When recognizing small products or distinguishing between packages with similar appearances, it can provide richer image details as a basis for judgment.
With a 30x optical zoom lens, the focal length can be flexibly adjusted. A single Sony FCB-EW9500H camera can cover multiple shelves of an entire smart cabinet, eliminating the need to install a separate camera on each shelf and greatly reducing hardware deployment costs.
The wide dynamic range (WDR) of the Sony FCB-EW9500H can simultaneously suppress the strong-light areas from supplementary lighting inside the cabinet and the shadow areas deep on the shelves. It can perfectly offset specular reflections from transparent packaging and metal cans, avoiding the problem of local overexposure that causes product features to be lost.
Second, real-time performance assurance from the perspective of encoding and transmission.
The sensitivity of visual recognition to latency is often underestimated. In a cabinet scenario, the waiting time from when the consumer closes the door to when they receive a payment notification directly affects their evaluation of the “seamless” experience.
The 4KP60 AI Network Interface Board supports 4K@60fps lossless HDMI-Compatible and network signal reception, fully unleashing the 4KP60 performance of the FCB-EW9500H and ensuring that high-frame-rate video is stably fed into the AI computing unit.
It is compatible with the VISCA protocol. The RS485 serial port can connect to an edge mainboard to enable a host computer to remotely control the camera module’s zoom, presets, and image parameters. It supports power-off memory and automatically restores preset working parameters after restart (some functions are optional).
It supports multi-resolution output adaptation and can output downward to 1080P60 and 720P60, adapting to AI computing devices with different compute capacities and reducing algorithm inference pressure.
The combination of the Sony FCB-EW9500H and the 4KP60 AI Network Interface Board provides a clear technical path for smart cabinet visual recognition: use 4K resolution to solve the problem of image capture, and use 60fps high-frame-rate low-latency encoding to solve the problem of image stuttering. Together, they provide a more reliable input signal for recognition algorithms.
Q: Can this solution directly complete product recognition, or does it require additional AI equipment?
A: The combination of the Sony FCB-EW9500H + 4KP60 AI Network Interface Board mainly solves the problem of “high-resolution, high-frame-rate, controllable video front-end acquisition.” AI inference for product recognition requires additional computing equipment.
Q: Does a smart cabinet scenario necessarily require 4K resolution, or is a 1080P solution sufficient?
A: This depends on the size of the cabinet, the number of shelf layers, and the requirements of the recognition algorithm. For large cabinets with many shelf layers and a dense number of products on a single layer, 4K resolution can provide more abundant image details, helping improve the recognition accuracy of small or similarly packaged products. For small cabinets with a simple structure and fewer product types, a 1080P solution can still meet the requirements provided that the number of cameras is reasonably deployed. It is recommended to select based on the specific cabinet structure and algorithm recognition rate test results.
