As the world of IoT (internet of things) expands, we begin to see that many of our connected devices need to ability to see the world around them to better understand their surroundings. Much of the machine learning abilities IoT devices have today are a direct result of computer vision capabilities which have evolved into the ability to recognize objects in the real world. To recognize objects in the real world, these devices and computers need to have the computer equivalent of eyes, also known as cameras. For the Internet of Things there is going to be an increased demand for quality optics to improve the accuracy and power consumption of these devices.
Today's camera systems don't cut it
Much of the camera systems that we see today vary wildly in image quality which makes it more difficult for the computers to properly identify the objects around them. This inability to properly and clearly see what is around them usually requires said IoT devices and the computers inside them to have to spend more compute resources to keep processing frames until it can recognize an object. The faster a computer can identify an object, the faster it can determine what it can do and the faster it can appropriately act. By shortening the time to identification you can make a smarter and more intelligent machine that uses less power and reacts more quickly.
Automated drones and vehicles
These capabilities are especially important in certain IoT applications that are already popular today, the first two that come to mind are automotive and drones. Both verticals are extremely popular areas of innovation for the machine learning and artificial intelligence fields and both require multiple cameras. The reason that both cars and drones require multiple cameras is because no individual camera can see 360 degrees around an any object and to have full environmental awareness, you need multiple cameras. This means having to constantly keep all the cameras on and constantly trying to detect objects to avoid or use for contextual purposes.
When it comes to self-driving and piloted driving cars and drones the role that optics plays in safety is immense. Most of these vehicles heavily rely on cameras around them to be able to see the lines, other vehicles, street signs and most importantly pedestrians. With quality optical systems, a good self-driving car can far outperform a human driver in accident avoidance including avoiding other vehicles and humans. People can sometimes be hard to see as they are partially obstructed by other objects, but a well-tuned precise optical system would be able to help the computer vision algorithms inside of a self-driving car to see the person before a human even with only a single appendage showing. With a lower quality optical system the computer may not be able to identify certain objects quickly enough as body parts and would not be as effective or safe.
VR and AR
Outside of smartphone photography, quality optics are also extremely valuable for wearables, namely the head-worn wearables that we are likely to see in the future. While it remains to be seen whether these devices will be VR or AR, they all require a certain type of optics to relay images to the human eyes. This can be done through complex wave guides, prisms or using Fresnel lenses that sit in front of displays. These future wearable devices are going to gain popularity as VR and AR continue to grow in popularity and the complexity of their optical systems will only increase as they continue to get thinner. The quality of the optics will become an even greater factor in ensuring that visual fidelity remains at the same levels as the thicker, older generation head-worn wearables of today.
Finally, is the world of security cameras, which had been mentioned earlier. Security cameras, both consumer and commercial are starting to gain intelligent functionalities enabled by machine learning and AI that allow for them to better recognize what they are looking at. By doing so, cameras today will be able to tell you who is at your door and who is inside your house. With precision optics, these cameras will be able to better and more quickly identify faces and at the same time capture better quality video of visitors or intruders. These cameras can also save power and storage space by not always recording and choosing to record depending on what the camera recognizes the object as. This can be especially valuable for cameras that are battery operated and stream wirelessly to the internet, every milliwatt counts and having quicker and better object recognition is extremely valuable in the long term.
As you can see, there are a lot of applications in IoT today that can benefit greatly from precise quality optics in a multitude of ways. Safety, performance, battery life and accuracy all come into play in addition to the obvious better quality images that you can get from these IoT devices. As many of these devices get smaller and thinner, the need for precise optics will only go up, not down. Money and lives can be at stake if a device isn’t able to see something that it should be able to, it will also be much harder to bring many of these IoT products to market if they don’t work reliably due to poor optics.
Anshel Sag and I will be drilling into different areas of optics over the next few months.