Darwinian engineering
Keywords: UV imaging, Thermal imaging, Nature
Our themes for this issue are heat sensing, thermal imaging and low-light imaging, and in addition to the technologies involved I am also personally interested in the new view of the world that these technologies can give us.
People have trichromatic vision (red, green, blue) while many animals have tetrachromatic vision because their eyes have cone receptors that can see in the near UV as well as the red, green and blue that we are familiar with. Bees are trichromatic but see UV, green and blue.
At the other end of the spectrum some snakes, notably the pit viper, are able to see in the infrared. The resolution of this IR vision is pretty poor but it is adequate for detecting and catching prey – which is what it is all about.
Nocturnal animals are largely monochromatic, but tend to have proportionately larger eyes to attain sufficient resolution. Speaking of resolution, birds of prey such as hawks have visual resolution that we find difficult to comprehend;sufficient it has been calculated to allow them to read a newspaper at 25m,assuming of course that they had the inclination to do so.
If you have ever marveled at how birds fly quickly between branches and grab fast moving prey then part of the answer is that they are able to process images about eight times faster than we can and at moments of stress adrenalin kicks into increase the rate even further. The effect of this high-frame rate is that everything appears to be moving much slower to them than us. Incidentally this same adrenalin effect works with people as well and accounts for why people often report that “everything happened in slow motion” when describing the course of accidents and other high-stress situations.
Some aspects of vision are identifiable enhancements of our own abilities;better resolution, higher sensitivity, faster frame rates, etc. but others, such as the ability to see in parts of the electromagnetic spectrum that are invisible to us open the door on whole new worlds. Nature has, over the last 300 million years or so, evolved mechanisms that are perfectly adapted to specific tasks. Eyes have developed to be very fit for purpose. This does not mean that a hawk's eyes are better than ours but they have been developed to meet different criteria. Bees, for example, see a very simplified version of the world they live in but it is sufficient to allow them to find nectar and to find their way back to their hive, and they do this with brains that can hardly rate too highly in the whole scheme of things.
For me this raises two questions when it comes to using machine vision for some automated task. The first is whether we are actually looking at it with the correct part of the electromagnetic spectrum and the second is whether we could develop a very simple system specifically for a particular task.
There is always a strong tendency to throw computing power at a problem and solve it in software using fancy algorithms rather than investigate optical techniques that simplify the problem to the point that a few logic gates provide the answers we are looking for.
The need to produce something at low cost is a great promoter of ingenuity. For example, security systems need to have ways of detecting the movement of people. One approach would be to interface a camera to a computer with image acquisition hardware and a bundle of imaging software and you would soon have a working system. An alternative, and one that would be approved of by Darwin, is the PIR. A simple device that uses crude plastic mirrors to focus infrared radiation on the surface of a silicon detector. It knows nothing regarding the nature of the person or what they are wearing or even how tall or fat they are– but it does detect the movement of people and is therefore fit for the purpose for which it was designed.
Clive Loughlin
