Feasibility of subtle light change detection in dark environment

Using Arduino Project Guidance
1

I'm researching a way to detect small changes in light hitting a sensor as a trigger.

The environment is an underground mine where there is no ambient light, save the camp lamps of workers and lights affixed to equipment. The goal is to be able to detect the headlights of a vehicle coming towards the sensor at the greatest distance possible, think 300-500 meters. There is a glow of the headlights before the lights themselves are visible, think of how fog glows at night when there is an oncoming vehicle. I suspect the distant glow of headlights would be far too subtle to overcome any noise on a simple sensor like a CDS cell. I wonder if a camera with a simple comparison of images every few seconds would do it, but I worry that the low light environment would cause too much noise. Perhaps a comparison of the average of the 100 darkest pixels with the average of the 100 brightest pixels in grayscale image captures?

I'd appreciate if anyone had any ideas that would be simpler or points out any obvious problems with ideas I've presented.

2

Why not detect thermals? Or use a multi detection system of ambient light and thermal.

Use a thermal camera and ML, machine learning, to detect the car. A KNN algorithm would work well, nearest known neighbor.

3

Would thermals be detectable at a distance of 300-500m? The ambient temperature is 30 degrees C

4

All such detection requires integration of values over time. Basically accumulating values over a time period and comparing that value to a previous value. So, the question is how much time do you require before the headlights are detected? Is a second ok or do you need to integrate over a longer time?

5

The TSL2591 can easily detect and measure starlight, on a moonless night.

6

The device is intended to provide early warning to the crew at an active mining face of others entering the work area for safety. In a perfect world, the glow of the headlights in the distance would be detectable before the headlights themselves had line of sight. The goal is detection at the greatest possible range, balanced with as few false positives as reasonable manageable.

7

I might order one of these for testing, I've been meaning to put together an adafruit order. Thanks for the suggestion

8

So, a 1 second delay is not acceptable? What is acceptable?

9

Sorry, I'm not 100% sure I understand your question.

If you're asking for the delay between the first glow of the lights appearing and the device registering a positive result, then 1-10 seconds is totally acceptable.

10

And that device varies the dynamic range by varying the integration time. Part of the setup communications with the device.

11

To achieve that remarkable dynamic range, the TSL2591 has four gain settings, and six integration time settings (100 to 600 milliseconds).

600 milliseconds and highest gain is adequate to measure starlight on a moonless night, so there should be no problem to detect faint headlight reflections in less than 1 second.

In fact, in tests I have done, the device can measure starlight reflected by lawn grass, on a moonless night.

12

This topic was automatically closed 180 days after the last reply. New replies are no longer allowed.