Final year project for Electrical and Electronic Engineering degree. Platform with a mounted laser pointer to track a person. Using OpenCV 3.1 and a Raspberry Pi 3.
From what it seems, the recognition is based in a kind of motion detection. You have an open-cv function for that and there's an youtuber called sentdex that has some very nice tutorials on that! About making the laser pointing to your target I don't know much on the topic, sorry...
@@VascoCC95 yeah sure, I didn't want to insinuate a comparison. In that case, do you know which is the optimizer way to program object detection in rp???
@@alvaroboven Well... You are lucky because I JUST ran a test on a RP and I got average 12,7 FPS. I used OpenCV's HAAR Cascade for face recognition in a RP 4B with 2GB RAM. I installed OpenCV via the link's tutorial and placed the haar file "manually" in the project's folder. www.pyimagesearch.com/2018/09/26/install-opencv-4-on-your-raspberry-pi/
The object is about 1.7 meters high, and it is on the ground, so from en.wikipedia.org/wiki/Pinhole_camera_model, calculate intersection of a line (from camera positon calibration) and ground plane , you can get the positon of the man. then point your laser to him.
Did you make your own closed loop system? Looks like you used DC motors and pots. Nice work! Though, the music seems a little off for a person tracker :)
Hi Andy, This project look's very cool and perfect :). How you established the connection between the raspberry pi module and arduino to share the object coordinate information.
Hi there, I was wondering how did u make the laser point to the real position of a person. I'm doing sth sort of similar with urs, projecting images on the back of one's palm. For the projection part ,I selected the area of projection in camera and then corrected the perspective of this area. SO I can just draw at the perspective corrected coordinate in my projection area. But how did u get the right rotation for a real world position. Thanks.
How was ur experience with rasp 3 in terms of speed while working with opencv ? I have rspb 2 model b and using a simple color detection it would slow the processing around 3-5 fps . I hope u get my question sorry for my English .
the fact that this video showed up on my recommended just know. I'm literally making a yt series on making a nerf gun turret that uses open cv to aim and shoot darts at ppl.
So I see that you've used C++... I've been trying to code with OpenCV and C++ for quite some time but I can't seem to figure out what IDE to use (if any) that will make it easy to use with OpenCV (building/path to the libraries gets really annoying). Are you using Code::Blocks, Geany, the Linux Terminal, or just your favorite text editor? Just curious if you're around 🤔
The cool thing to do is put the camera where the laser pointer is and have the camera move so as the object that is tracked it is always drifting it back to the center of the screen. Now you have an auto tracking gimbal like soloshot without a tag. I don't think distance is being calculated here, because the camera is fixed the image the movements of the laser are limited to the fov and so the laser moves in relation to the center of the image. It probably took trial and error of some sort to move the laser to the correct extents of the image and distance is irrelevant. You would need additional data (gps) to calculate distance.
funkyfreestyler thanks for your comment! True, distance was not calculated here and was irrelevant. The aim was to only have the laser point at the middle of an object in the fov of the camera. Having a moving camera would mean I would have to change my method of image processing as I made use of the ‘steady’ background.
DROIDii1 Yeah, on a serious note on your number 4... how would a system like that distinguish between hostile and friendly individuals? One way I see it is by marking friendlies with an IR emitter, but that wouldn't do in a military situation where this would denounce the operative's position on the enemy's NVGs. It could be used though if it is stationary and it has to protect a perimeter AFTER a certain limit (as this can be easily programed in OpenCV), limit within which there can be safe movement of troops.
@DROIDii1 this would be true for a static camera facing a fixed scene. But won't it be problematic for a moving camera which has to deal with a changing background scene?
Hi, I am designing a motion tracking project with raspberry pi. I will use it as a finishing project. I want to get information about the laser project. can you help me.
+Willyam Adhi Hi, i recently bought two optical scopes, i already made an adapter to mount my smartphone on it (digi-scoping), works great. Next step in the project is to make a pan & tilt for the scope(s) and use the pi camera instead of the phone. Then stream the image via a web server on the pi. Then i created a remotely controlled extreme zoom webcam... i live in a high building with great view... So i need 2 (or 4) servo's for the pan & tilt, the spotting scope weight is 988 gram, but in balance, i dont think it's hard work for the servo's. What's your plan with the pan & tilt system? Regards, Bastiaan
Bastiaan Klaasing hi, that is cool project. Mine is just like yours but it's focused on surveillance. So I kinda need smooth pan & tilt but also compact in size. I'm still working on it
Hi guys, nice work! I'm trying to get OpenCV installed on a Pi3 and keep getting errors. Are there any docs or tutorials on the install that you could point me towards?
In case you still need it or anyone else is looking for a step by step: www.pyimagesearch.com/2016/04/18/install-guide-raspberry-pi-3-raspbian-jessie-opencv-3/
Easy man.Substract background picture from recent picture than get moment of substraction.Can you do it moving background with two similar object or more :)
I am using shape descriptors or their combinitions.Some times i am coding some times iam using opencv But the key is to find invariant features of the shape.In opencv you can find some of them like sift,surf,blob.Then elemination of redundant unneccessary features.Some moments like hu,zernike usefull too.Depents on situation.Good like my friend.