Every frame captures the lighting at the same point in its cycle.
On May 22, 2026, Vadzo Imaging introduced the Innova-678CRS, a 4K HDR GigE camera built around Sony's STARVIS 2 sensor — a device designed not merely to capture images, but to reconcile the ancient tension between light and shadow that has always limited what machines can reliably see. In industrial and urban environments where illumination is neither stable nor cooperative, this camera attempts to remove light itself as a variable, offering engineers and system designers a more faithful window onto the physical world. It is a quiet but consequential step in the longer story of teaching machines to perceive with something closer to human adaptability.
- Industrial inspection and traffic monitoring systems routinely fail not because of mechanical faults, but because light — flickering, uneven, or extreme in contrast — defeats the cameras meant to observe them.
- LED flicker invisible to human eyes corrupts detection algorithms frame by frame, causing robots to misread barcodes and automated systems to confuse luminance noise with actual surface defects.
- The Innova-678CRS answers with dual HDR modes reaching 120+ dB dynamic range and shutter synchronization to LED cycles, attacking both the contrast problem and the flicker problem simultaneously.
- Back-illuminated STARVIS 2 sensor architecture and low-noise readout circuitry extend reliable performance into near-darkness, removing the need for supplemental lighting in warehouses and nighttime traffic environments.
- A single GigE PoE cable per camera, combined with ONVIF Profile S through M compliance, allows multi-camera deployments to scale across factories and smart city infrastructure without proprietary hardware or custom middleware.
- Engineering samples are available now, positioning the camera as an immediate option for OEM integrators and system designers navigating the growing demand for machine vision that performs under real-world, uncontrolled conditions.
Vadzo Imaging's Innova-678CRS, announced in May 2026, is built around a deceptively simple premise: that light itself is often the primary obstacle to reliable machine vision. Industrial floors mix task lighting with ambient glare, creating scenes where polished metal surfaces blow out while recessed areas vanish into shadow. Traffic cameras must simultaneously resolve a sunlit car hood and a pedestrian standing under a bridge. Standard cameras are forced to choose one extreme or the other. The Innova-678CRS does not choose.
The camera's two HDR modes address this directly. Clear HDR merges multiple exposures into a single frame for static scenes, while Digital Overlap HDR interleaves short and long exposures at the sensor level for moving subjects — together achieving up to 120 decibels of dynamic range. Alongside this, the camera synchronizes its shutter to the cycling frequency of LED lighting, eliminating the frame-to-frame luminance variation that is invisible to human observers but catastrophic for detection algorithms. A robot reading barcodes or an inspection system hunting for surface defects no longer has to contend with lighting noise masquerading as meaningful data.
The underlying Sony STARVIS 2 sensor uses back-illuminated architecture, routing light directly to the photodiode rather than through metal wiring layers first. Combined with low-noise readout circuitry, this extends usable performance into near-darkness — nighttime traffic monitoring, dimly lit warehouses — without requiring supplemental illumination.
Connectivity is handled through a single Gigabit Ethernet cable carrying both data and power, removing the need for separate power routing across factory floors or along traffic poles. ONVIF compliance across Profiles S, T, G, and M means the camera integrates directly into standard video management systems and outputs structured analytics metadata without custom middleware or proprietary drivers. Support extends to C, C++, C#, and Python APIs, with browser access across Windows, Linux, Android, and embedded edge platforms. Engineering samples and evaluation kits are available now, with OEM customization handled through direct contact.
Vadzo Imaging announced the Innova-678CRS on May 22, 2026—a camera built around Sony's STARVIS 2 sensor that solves a problem most people never think about: how to capture usable images when light itself is the enemy.
The camera delivers 8.4 megapixels at 4K resolution, but the real story lives in what it does with that light. Industrial inspection floors mix task lighting with ambient light, creating scenes where polished metal surfaces blow out into white while recessed areas disappear into shadow. Traffic cameras face the opposite problem: direct sunlight on a car hood while a pedestrian stands in deep shadow under a bridge. Standard cameras choose—they expose for the bright areas and lose the dark ones, or vice versa. The Innova-678CRS doesn't choose. It captures both simultaneously through two different HDR modes. Clear HDR combines multiple exposures into a single frame, useful when nothing is moving. Digital Overlap HDR interleaves short and long exposures at the sensor level itself, maintaining detail in both extremes even when subjects are in motion. The result: up to 120 decibels of dynamic range in a single frame.
But there's another invisible problem the camera addresses. Modern LED lighting—in warehouses, factories, and street installations—cycles at frequencies that interact with standard camera shutters in ways that create frame-to-frame flicker invisible to human eyes but catastrophic for detection algorithms. A robot trying to read a barcode sees the lighting intensity shift between frames. An automated inspection system looking for surface defects gets confused by luminance variation that has nothing to do with the part being inspected. The Innova-678CRS synchronizes its shutter to the LED cycle, eliminating this variation entirely. Every frame captures the lighting at the same point in its cycle.
The sensor itself uses back-illuminated architecture, a design that inverts the traditional layer stack so light reaches the photodiode directly rather than passing through metal wiring first. This matters most when light is scarce—nighttime monitoring, dimly lit warehouses, environments where adding more illumination isn't practical or permitted. The STARVIS 2 generation adds low-noise readout circuitry on top of this architecture, further improving signal quality when cameras are running at high gain in near-darkness.
Connectivity is handled through Gigabit Ethernet with Power over Ethernet compliance, meaning a single cable carries both data and power to each camera. This eliminates the need for separate power routing across factory floors or along traffic poles. Multiple cameras can be networked together through standard switches without frame grabbers or proprietary hardware at each node. The camera supports ONVIF Profiles S, T, G, and M—a standards hierarchy that starts with basic video streaming and culminates in Profile M, which allows the camera to output structured analytics metadata directly into any ONVIF-compliant video management system. No custom middleware, no proprietary drivers.
The applications span industrial production lines where micro-defects must be caught under inconsistent lighting, robots navigating between fluorescent and LED zones, traffic infrastructure monitoring pedestrians and license plates across extreme contrast ranges, and distributed warehouse surveillance systems that need to scale without dedicated hardware at each location. The camera is supported by Vadzo's NXT software platform and VISPA NXT SDK, with APIs available in C, C++, C#, and Python, and browser access through standard browsers on Windows, Linux, Android, and embedded edge systems.
Engineering samples and evaluation kits are available now through Vadzo's website, with volume pricing and OEM customization handled through direct contact. The camera represents a shift in how industrial and surveillance systems approach the fundamental problem of capturing reliable images in environments where light is neither constant nor controllable.
Notable Quotes
DOL-HDR interleaves short and long exposure lines at the sensor level within a single readout cycle, eliminating inter-frame motion between the two exposure components.— Vadzo Imaging technical documentation
LED flicker mitigation synchronizes shutter timing to the LED cycle, ensuring every frame captures the same point in the illumination waveform and eliminating this variation.— Vadzo Imaging technical documentation
The Hearth Conversation Another angle on the story
Why does LED flicker matter so much? It's just lighting, right?
LED cycles at twice the mains frequency—100 Hz in 50 Hz regions, 120 Hz in 60 Hz regions. A standard camera's shutter doesn't always align with that cycle, so different frames capture the LED at different points in its brightness waveform. Humans don't see it, but a detection algorithm looking for consistency between frames sees the luminance jump around. It causes false positives, missed detections, barcode read failures.
And the back-illuminated sensor—that's just a marketing term, or does it actually change what the camera can do?
It changes what the camera can do in low light. Front-illuminated sensors have metal wiring on top that blocks some light before it reaches the photodiode. Back-illuminated inverts that—wiring goes behind the photodiode. More photons reach the light-sensitive area, especially critical when you're working with very few photons. Add STARVIS 2's low-noise readout and you get usable images at night without flooding the scene with infrared illumination.
The two HDR modes—Clear and DOL. Why not just pick one?
Clear HDR takes multiple sequential exposures and combines them. Works great for static scenes. But if something's moving between exposures, you get motion artifacts that confuse detection systems. DOL-HDR interleaves the exposures at the sensor level within a single readout cycle, so there's no motion between the short and long exposure components. For traffic monitoring or robots in motion, DOL is the right choice.
ONVIF Profile M—that sounds like alphabet soup. What does it actually unlock?
Profiles S and T get you video streaming and basic metadata. Profile M lets the camera output structured analytics data—detected objects, classifications, events—in a standardized format. Any ONVIF-compliant video management system can consume that directly. For smart factories or smart cities where camera output feeds into automation systems, Profile M is what makes integration seamless instead of requiring custom code.
So you're saying a factory could deploy dozens of these across a facility without buying specialized hardware?
Exactly. Standard Gigabit Ethernet switches, standard network infrastructure. No frame grabbers, no proprietary vision hardware. ONVIF compliance means any VMS can discover and manage the cameras. You add a camera the way you'd add any networked device.