Optronics and vision systems are now at the heart of many innovations, from autonomous robotics to defence, industry, space and security. But how have these technologies evolved in recent years? And how far can they go?
To gain a clearer understanding, we spoke with Vincent Carrier, CEO of Nexvision, an engineering firm specialising in the design of optronics and embedded vision systems.
First of all, what is a vision system?
Vincent Carrier:
A vision system is primarily a system capable of analysing its environment using an image sensor, usually a camera, but not exclusively (e.g. lidar, radar). It is not just a matter of ‘seeing’, but above all of understanding what is being seen.
In concrete terms, these systems either augment human capabilities or completely replace them through autonomy. They are found, for example, in drones, robots, autonomous vehicles and critical embedded systems.
The issues addressed are very varied:
- Detecting a whale to avoid a collision with a ship (ex : Seadetect project),
- Making driving safer by assisting or replacing the driver,
- Giving perception to a humanoid robot, in what is now called physical AI,
- Improving soldiers’ perception through advanced optronic helmets (ex : eSAS),
- Processing and analysing images on board satellites,
- And many more!
We are talking here about systems capable of interacting with the real world, moving around in it, manipulating it and making decisions in real time: vision then becomes a fundamental pillar of autonomy.
What major developments have been observed in recent years?
Vincent Carrier:
We have clearly seen a major acceleration in the capabilities of vision systems, driven by several technological breakthroughs.
1. The rise of artificial intelligence
The ‘deep learning’ branch of AI has profoundly transformed the way images are used. Where deterministic algorithms were previously used, AI now offers:
- Greater robustness in real-world conditions,
- Learning capabilities,
- And a much more detailed interpretation of complex scenes.
These advances are particularly significant for physical AI, where vision is no longer used solely to analyse an image, but to drive an action: moving, anticipating, avoiding an obstacle, grasping an object or interacting with a human.
2. The massive increase in embedded computing power
Modern platforms now combine CPU, GPU, NPU, FPGA and ASIC on a single architecture.
A good example is what we find today on NVIDIA AGX Thor-type processors, capable of analysing complex video streams in real time, directly on board.
This embedded computing power is a key prerequisite for autonomy, particularly in humanoid, industrial or service robotics, where decisions must be made locally, without relying on the cloud.
3. Advances in optics and manufacturing
Advances in optics, particularly with metalenses and freeform lenses, enable the design of:
- More compact optics,
- More efficient optics,
- And optics that are better suited to mechanical and environmental constraints.
This paves the way for vision systems that can be integrated into mobile platforms, robots or portable equipment, further enhancing their level of autonomy and integration.
And what role does light play in all this?
Vincent Carrier:
We often forget it, but “in vision, good lighting solves half the problem“.
Light-emitting components (LEDs, lasers) have made enormous progress:
- Better performance in terms of range and accuracy,
- Better energy efficiency of photonic components,
- And finer integration into systems.
This is particularly true for active sensors: LIDAR, 3D vision, structured imaging, etc. These technologies are essential for enabling autonomous systems, particularly robotic ones, to perceive their environment reliably, regardless of conditions.
Have image sensors kept pace?
Vincent Carrier:
Absolutely. Sensors have improved on all fronts:
- Resolution and spectral bands,
- Sensitivity, particularly in low light conditions,
- Reading speed (e.g. event-based vision sensors),
- And robustness in challenging environments (e.g. space).
Combined with AI and embedded computing power, they now enable robots – whether humanoid, industrial or service robots – to perceive, understand and act in an increasingly autonomous manner, paving the way for their gradual integration into our daily lives.
Can we say that this is a rapidly growing market?
Vincent Carrier:
Definitely yes. Vision systems are solving more and more practical problems, which is leading to widespread adoption in many sectors.
Vision is being used everywhere:
- In manufacturing,
- Defence,
- Space,
- Aeronautics,
- Transport,
- Energy,
- Agriculture,
- and now at the heart of autonomous systems and advanced robotics.
The rise of physical AI and autonomy is greatly accelerating this trend: the more machines interact with the real world, the more they need high-performance vision systems.
Are there limits, despite everything?
Vincent Carrier:
Yes, and that’s a key point. We are gradually reaching physical limits, particularly in microelectronics.
Today, engraving technologies are around 50 atoms, and we are aiming for thresholds close to 20 atoms.
We still have some leeway, but we won’t be able to miniaturise indefinitely.
The market’s response is therefore clear: multiply the number of chips and parallelise processing, specialise architectures and optimise the entire system.
This is a major challenge for autonomous platforms, which must balance performance, energy efficiency and compactness.
What is Nexvision’s role in this ecosystem?
Vincent Carrier:
At Nexvision, our added value lies in thinking about the vision system as a whole: optics, sensors, electronics, mechanics, computing, image processing and field constraints.
We support our customers on projects where vision is at the heart of autonomy and physical AI, whether in robotics, critical embedded systems or innovative platforms designed for long-term integration into society.
A conclusion?
Optronics is at a turning point: vision technologies have never been so powerful, and the stakes have never been so high. Between technological breakthroughs, the rise of physical AI, system autonomy and physical limitations, system expertise is becoming the real differentiating factor. This is a playing field that Nexvision has fully mastered!
