|
|
|||||
|
New Product
MotCam: Crossing-base objects detector
Hi all,
I’m proud to present MotCam, my development for planes motion detection. I’ve been working in this project for a whole year, trying several different approaches to find the most consistent and fastest detection system as possible. After rebuilding it from scratch two times, I think I have a very reliable and flexible detection algorithm that can run fast enough for the most exigent disciplines (I hope so!). MotCam is a complete software for crossing-base objects detection based on optical analysis. Images comes from one or more cameras connected to PCs or laptops that analyse frame by frame all the appearing objects and their trajectories, and detects when one of them crosses the base (vertical line in the middle of the screen) from one side to the other in the correct direction. Every time an object is detected crossing the base, MotCam informs about it: • Acoustically, by using the PC’s speaker. • Optically, by drawing special information on screen. • Remotely, by sending a TCP/IP message through the network (in case of connected to it). • Serially, by sending user defined data through a serial port (in case of connected to it). • Electronically, by activating an USB Relay (in case of connected to it). With MotCam, it is possible to configure all the camera settings (exposure, gain, frames per second, resolution and rotation), as well as all detection parameters (algorithms and its characteristics, image processing, video recording, image saving, and direction for detection) in three ways: • By using the interface, • By modifying the configuration file, • Through simple TCP/IP messages. Also, MotCam can be configured to send automatically the detection images through the network to other third-party software that could make use of them for specialized applications (F3F, F3B, etc.). These applications can be connected to as much MotCams as needed, managing them remotely by a complete set of commands that can be invoked through simple TCP/IP messages. MotCam has been specially designed to work exclusively with Sony® PS3 Eye camera. This camera is inexpensive and has an excellent processing time of only 1/120th second. It provides 2 resolutions (320x240 and 640x480) with very high frame rates (up to 180 and 75 FPS respectively). The camera features a two-setting adjustable fixed focus zoom lens that can be selected manually by rotating the lens barrel: • 56° field of view (red dot) for close-up framing. • 75° field of view (blue dot) for long shot framing. MotCam can run in any PC or laptop with Windows 7, 8 and 10 in both x86 or x64 versions. This PC or laptop should be selected accordingly with the processing speed needed. For simple normal speed detections, cheap PCs based in Atom Z8300 CPUs should be enough. For high speed objects with concurrent real time video recording, it is recommended the use of higher-end ones (Intel® Core™ i5 or i7). 1 GB RAM will be enough for most applications. To connect many MotCams with a third-party application, it’s strongly recommended the use of Cat5e or Cat6 Ethernet cables and a router. Managing MotCams remotely and receiving their messages through Wi-Fi connections is possible, although additional delays will be introduced on the transmission. After testing it intensely for my own during the last 3 months, I finally did a “real” test on a F3F competition last weekend. The result is fantastic, 100% of success detecting planes and 100% of success ignoring artifacts, clouds moving, sun appearing and disappearing, and bees buzzing around. In this case, the detection was from left to right, ignoring the planes in opposite trajectories. You can see the video here:
The video is a "honest" editing, I left the parts where a plane is appearing and cut the rest. The delay on detection is between 2 and 4 frames. As I was analysing 75 frames per second, the delay on detection was about 50-75 milliseconds, much faster than human reaction. I was very lucky because I found a bug. Not a computer bug, but a real one buzzing in the same direction than the plane. It can be seen on minute 1:04 of the video (pay attention, it’s only appearing on 3 frames!). The software detects it, but it decides it is not a “compatible trajectory”, so it ignores it. There's another bug on minute 1:08, but it flies on the opposite direction and it's ignored quickly. Here you can see another video from outside the camera, so you can hear it buzzing. It could be faster, since I was also recording the video with the detecting PC, but I think it's good enough. The delay buzzing is 3 frames on 50 FPS, but try to do a "honest buzzing" by snapping the fingers when the plane is in front of the camera to see if you are faster than it ![]() ![]()
MotCam has been strongly tested in order to find and correct bugs and defects on its code. Actually, the version is in Release Candidate 1 status. As this is a very special software based in optical recognition, and it can be used in a lot of different situations with different kind of objects to detect, the beta program has been launched to allow final users to test it for their own purposes and have feedback with their results. If you are interested on joining the beta program, send me an e-mail or a private. The price of the beta version is 60 euros per camera. Once the final version is launched, beta testers will receive a registered copy of this version for free. You can find the complete user manual attached in this URL: https://www.dropbox.com/s/qnp26kikn4...anual.pdf?dl=0 I hope you like it. Best regards, Alvaro. |
||||
|
Last edited by Alvaro Silgado; May 15, 2017 at 08:56 AM.
|
||||
|
|
|
|
Hi Alvaro, this looks very impressive. From my side I fly F5B and was curious if testing had been done when planes are 200+m away?
Thanks |
|
|
|
|
|
Dear Alan,
To be honest, I've never tried it that far with small planes. I had false positives with real planes passing really far away, and delta wings at about 1000m distance. The point is, if you can see the planes ON A VIDEO, the program will probably detect them. Another important point is that you can run as many cameras as you need and interconnect all of them with Ethernet wire, activating or deactivating all of them at the same time. The first one buzzing will be the one to consider. I hope it helps. Al. |
|
|
|
|
|
Actualized versions of the user manual can be downloaded from this URL: https://www.dropbox.com/s/qnp26kikn4...anual.pdf?dl=0
|
|
Last edited by Alvaro Silgado; Apr 25, 2017 at 11:06 AM.
|
|
|
|
Coming soon...
|
|
|
|
|
|
New program "ReMotCam Tester", now included with MotCam!
ReMotCam Tester is a free utility that communicates with all MotCams connected to the local network. Found remote cameras will appear automatically on the combo box. Selecting one of the MotCams, it's possible to send remote commands to manage it, receive its messages and images, do image streaming, and beep on a remote PC each time a MotCam detects an object. User manual has a new chapter explaining this useful utility: https://www.dropbox.com/s/qnp26kikn4...anual.pdf?dl=0 I hope you like it! Álvaro. |
|
Last edited by Alvaro Silgado; Apr 25, 2017 at 11:06 AM.
|
|
|
|
Impressive support Alvaro!
Waiting for our PC to arrive so we can start using MotCam! |
|
|
|
|
|
|
|
|
|
Had a chance to try out the system today in the shop .... set up a cam at each end and flew the "course" with my Mini Whoop quad racer ... worked flawlessly with a nice BEEP at each end ...
![]() We'll hopefully get out this weekend and test it with EF1 and F5D planes. Thanks Alvaro for this awesome system!!!! Mark |
|
|
|
||
|
Quote:
Dave |
|
|
||
|
Quote:
![]() I'm looking forward to have your feedback with racers and F5Ds. Just some tips for setting the cameras for fast and small planes: - Try to have a helper flying his plane so you can configure it parameter by parameter and see how every change affects to detection. - If possible, use the 56º FOV position for the lens. Detection area will be smaller, but planes will be seen bigger and easier to detect. It is better to have a smaller detection area and place cameras redundantly every certain distance than using a single camera with a wider detection area. - If the planes are being seen small by the cameras, you’ll probably need the 640x480 resolution to have better detail on every frame. But, if you’re passing near the cameras, try using 320x240@120FPS. You’ll be amazed of having an almost-immediate reaction time! - Use the blob-tracker algorithm with a distance factor of 3 (or maybe 4), and a 1% of minimum size of objects. - Just for fun, try the photo-finish algorithm too. It’s thought for indoor use and it doesn’t reject clouds, insects, or wrong trajectories, but it’s incredibly fast on detections. Use a 15% of size of detection area, and about a 5% or less percent of area detected to start with. - If possible, record video locally (set it to “Detection videos” so you'll have a video file for every detection). It will reduce a little bit the reaction time (you won’t notice it), but it will be useful to analyse mistakes buzzing and will help you (and me!) to configure it better. - Try to place the cameras avoiding backlight. Direct sunlight is much better for detecting objects. Thanks a lot for your support. Have fun! Álvaro. |
|
|
||
|
|
|
Thank you, brother!
I've heard you'll probably see it working by yourself next weekend... Have fun, I miss you all! Al. |
|
|
Thread Tools | |
Similar Threads | |||||
Category | Thread | Thread Starter | Forum | Replies | Last Post |
Discussion | 17th century detector lock | dll932 | Life, The Universe, and Politics | 0 | Mar 08, 2017 10:17 AM |
Question | 1s Batteries and Low Voltage Detectors | BrrrIce | Batteries and Chargers | 1 | Jan 08, 2017 12:47 PM |
Careful! | "Smoke Detector Disconnector" Good idea to help prevent LiPo fires? | thatjoshguy | Batteries and Chargers | 8 | Oct 07, 2016 11:54 AM |
Discussion | Important illusion- Motion of an object can mask a change in the object | CNY_Dave | Beginner Training Area (Aircraft-Electric) | 15 | Jan 12, 2011 01:26 AM |
Discussion | Ground Base Thermal Detector Needed | John Boren | DIY Electronics | 8 | Mar 27, 2010 06:00 AM |