How to Use OpenMV Cam for Real-Time Object Tracking with Arduino Integration?

Project Discussion
1

Hey guys… :smiling_face_with_three_hearts:

I am trying to use it for real-time object tracking. The goal is to detect and track a specific colored object (like a red ball) and then send commands to an Arduino board based on the object’s position in the frame.

Also, I have managed to set up the OpenMV IDE and get some basic scripts running, like face detection and color tracking examples. However, I’m struggling with understanding how to make the OpenMV Cam continuously track an object and communicate this data to the Arduino in real time. Here are a few specific questions I have:

  • I’m using the find_blobs function to detect the object by color. However, I’m not sure how to fine-tune the color thresholds for better accuracy. The object sometimes goes undetected, especially under different lighting conditions. Is there a recommended method or tool to calibrate these color thresholds more effectively?

  • I know the OpenMV Cam can communicate with an Arduino via UART, but I’m a bit lost on how to structure this communication. Should I send data in a specific format (e.g., JSON, CSV) to make parsing easier on the Arduino side?

  • Right now, the tracking feels a bit sluggish. I’ve heard that optimizing the image resolution and frame rate can help, but I’m unsure how to balance these settings to achieve both speed and accuracy.

I also check this: https://forums.openmv.io/t/realtime-color-tracking-with-arduino-and-openmendix But I have not found any solution. Could anyone guide me about this?

Thanks in advance! :innocent:

Respected community member! :blush:

2

Hi Nisha02:

  • I’m using the find_blobs function to detect the object by color. However, I’m not sure how to fine-tune the color thresholds for better accuracy. The object sometimes goes undetected, especially under different lighting conditions. Is there a recommended method or tool to calibrate these color thresholds more effectively?

Please see the Tools->Machine Vision->Threshold Editor in OpenMV IDE. Color thresholds are somewhat invariant across lighting if you make the thresholds very loose. However, generally, you may need to use multiple thresholds to achieve your goal. Find_blobs() supports this.

  • I know the OpenMV Cam can communicate with an Arduino via UART, but I’m a bit lost on how to structure this communication. Should I send data in a specific format (e.g., JSON, CSV) to make parsing easier on the Arduino side?

Typically you use binary data. See our Arduino RPC library: openmv/openmv-arduino-rpc: Remote Procedure/Python Call Library for Arduino (github.com)

  • Right now, the tracking feels a bit sluggish. I’ve heard that optimizing the image resolution and frame rate can help, but I’m unsure how to balance these settings to achieve both speed and accuracy.

What camera model are you using? For any resolution at VGA or under you should see a maxed out FPS.