I need some guidance in selecting the right sensor for a project I like to start.
Some background info of the purpose:
I do a sport called Agility, this is mainly guiding a Dog through an obstacle course with jumps and tunnels as fast as possible. (fast dogs can run up to 36kph.)
Purpose of the project:
I am starting a project where I can detect the dog passing a staring line and a finish line, to calculate the time from start to finish:
there are already (expensive) products available to do this, but all of them are using sensors with two components (like “Retroreflective” or “Thrubeam”) Detection method and feature of Photoelectric Sensors | Sensor Basics: Introductory Guide to Sensors | KEYENCE
I want to have a system which uses only “one” component, so it’s more portable. (eventually I want to be able to add extra sensort between start and finish to calculate meantimes). the measure distance needs to be between 10cm and 100cm.
To be clear, it’s not to measure the distance between the sensor and the dog, but only the passing of the sensor.
I have considered TimeOfFlight sensor "adafruit VL53L0X " and Infrared Proximity Sensor “adafruit GP2Y0A02YK” but I don’t know if they are fast and accurate enough since the dogs can have a speed up to 36kph.
does any of you have suggestions for the best sensors to use?
Let's say the reflective patch of the dog's collar is 2cm wide. (Maybe you put a 1-inch collar on the dog but it's partly obscured by the fur.)
At 36kph, what is the time period available to detect that collar passing the sensor? 36kph is 1000cm/sec. A 2cm collar will be in front of the sensor for 2 milliseconds. That's a really long time for an Arduino. You would not have to do anything special to examine several sensors 5-10 times each in a 2ms window.
Is this one of those where the dog's owner runs alongside? Then it will be difficult to detect only the dog and not the handler. A radio transponder on the dog will discriminate it easily but then you need a bit more effort to constrain the radio reception zone into a straight line.
I think you're going to have to go retro-reflective and carefully adjust the levels until the reflective collar is identified more strongly than any other light source. (Daylight makes it tough.) Maybe you will need a modulated transmitter like a TV remote so that the return signal is identifiable.
The optics might be a pair of cardboard tubes mounted over-and-under, so the transmitter and receiver only cover a narrow field of view. You will increase sensitivity by using lenses. Buying pre-buit sensors will get expensive because you are getting into industrial sensors, not consumer-level.
Maybe you could orient an optical detector (such as a photodiode) so it is pointing at the sky and when the dog runs over it s/he creates a shadow?
My concern is that the detector would get dirty.
The advantage of a break-beam sensor is it can be placed out of the way of dirty paws. An optical detector on one side and a battery powered lamp on the other side does not seem difficult to carry around or install. There would be no need for an electrical connection between the lamp and the detector.
You could probably use an infra-red LEDs for the "lamp" if you don't want the light to be visible. However I don't know if dogs can see infra red.
Hi, thanks for your reply MorganS.
unfortunately it’s not allowed for the dogs to wear a collar during a run.
And yes it’s the sport where the handler is also in the course. but only the dog takes the obstacles
So you have to distinguish the dog at the start and finish with nothing on the dog to assist the computer in determining it's the dog and not the handler?
Yes MorganS, but distinguishing the dog from the handler should not be that hard since the handler is not passing the obstacle. I have to detect the moment the dog is passing the sensor/equipment. like the image in the link below
Yes MorganS, but distinguishing the dog from the handler should not be that hard since the handler is not passing the obstacle.
What about my suggestions in Reply #2 ?
Break beam is really the way to go here.
Those proximity sensors may do the job - if you're OK with an error of about 0.1 seconds, at least the VL53L0X has a minimum reaction time of about that duration due to the required measurements and integration done on board.
An IR based sensor has the additional issue that a white dog will be detected much better than a black dog (which, for lack of reflections, may be missed altogether).
A sensor under the track, looking for a shadow of the dog, is not reliable as you don't know when the shadow will be cast onto the sensor: this depends on the current angle of the sun. If it's an overcast day you don't have much of a shadow to begin with. A strong light above it may be the solution (artificial sun) but that basically tuns it into a break beam sensor again.
@Robin2, thanks for you suggestions, Break-beam for me is the last resort solution, a photodiode detecting the shadow will not be accurate enough i guess.
however your suggestion for an upward facing sensor is a nice approach.
since the measurement (almost) always is for a jump, I am thinking of putting some upward facing sensors in the horizontal bar of the jump (using object avoidance sensors). I am just not sure if they will intervene with eachother.
As i mentioned in my reaction to Robin2, Break-beam for me is the last resort solution.
Is this reaction time of 0.1 sec for VL53L0X consistent?
because if it is, the same reaction time at the finish line will cancel out the start-line
Quote "An IR based sensor has the additional issue that a white dog will be detected much better than a black dog (which, for lack of reflections, may be missed altogether)."
does this apply to all "measure" distances, or only for distances over x centimeters?
Minimum measurement time of that sensor is 33 ms, higher accuracy requires longer times.
That is your timing resolution. The dog may pass through at any moment.
Quote “An IR based sensor has the additional issue that a white dog will be detected much better than a black dog (which, for lack of reflections, may be missed altogether).”
does this apply to all “measure” distances, or only for distances over x centimeters?
Of course closer distances are less affected. See data sheet p.26 for an indication of maximum range indoor and outdoor. They don’t even mention black targets, just grey and white.
A quick experiment shows that indoors at 30 cm distance it seems to have no problem detecting a black t-shirt, that’s the most matte thing I have at hand. So it seems that if you can guarantee those dogs to run by the sensor at that distance, you should be fine.
Place a few above one another and you can measure dogs of different size, and be better able to tell when the first part of the dog crosses the line. Reading them gets tricky as you have to enable them one by one to take a reading (only one I2C address).
Those Sharp sensors you were looking at have an analog output so it’s easier to mount a number of them above one another - a Nano or Pro Mini has 8 analog inputs for total eight such sensors. Should be enough to make a nice gate. Response time as given in the data sheet of these sensors is 39 ms. For a good signal the dogs should be passing that sensor at a distance of 15-60 cm. Maybe a little further but not too much. I’ve looked at those myself before; no hands on experience with them.
Another consideration is the viewing angle. That of the VL53L0X is about 15°, that of the GP2Y0A02YK is not given in the data sheet other than that it’s “a wide angle sensor”. This means that the further away from the sensor, the more before the finishing line the dog will be detected. Darker coloured dogs likely need more of their body in front of the sensor to be detected. In contrast, break beam sensors use a very narrow beam, and their field of vision is limited further by the size of the reflector on the other side.
You can purchase curtain style motion detectors.
You can use EXP-NOW and put in an end point that sends out it's signal and have a dozen slaves to send back to one controller.
A portable unit should be as simple and inserting a stake in the ground at the start and one at the finish, and one anywhere else you like.
battery power and even a display or act as webserver.
as with all the sensors, you will need to test.
at first thought a reflective collar is X distance back from the nose. each dog will have a different distance.
not sure if you are working in nano-seconds or whatever, but every sensor will have some drawback. I think that is why they have a camera for visual verification.
Have a look at this paper, especially p.4.
The Sharp sensors have a near-60° wide field of vision, but it's highly lopsided. One side of the sensor has an angle of 5-10°, the other side of the sensor has the rest of the field of vision. This would make it quite suitable for your dog-detection. Have the "blind" side of the sensor on the side where the dogs come from, and your sensing is quite close to the actual finish line.
@wvmarle: Thanks for you elaborate answer and investigation.
I am going to look into that asap. I think I am going for the sharp sensor and go from there.