camcv

CamCV

Motion detect on webcam.
Learn differences from previous and current image then trigger action when diff ratio greater than tolerance.

Requirements

• Boost lib (v1.78.0)
• OpenCV lib (v4.x)
• C++ compiler (v10.2.1)

Setup

Debian

sudo apt install cmake libboost-all-dev libopencv-dev

Build

./build.sh

Run

./run.sh

Devices

Identify cam devices

#!/bin/bash

for f in /dev/video*; do
    item="$(basename -- $f)"
    name="$(cat /sys/class/video4linux/$item/name)"
    echo "$f  =>  $name"
done

Screenshot

Usage

./run.sh

Display help

./build/camcv --help

Usage camcv [options]:
  --help                display options
  --d arg               device id
  --t arg               threshold
  --w arg               width
  --v arg               verbosity
  --i arg               capture interval
  --s arg               save image
  --ft arg              filter type
  --fv arg              filter value
  --gui arg             use GUI
  --cap arg             capture Path

Sample run options

./build/camcv --d 0 --t 1000 --w 320 --v 1 --i 2 --s 1

In this case we :

• use first video device (/dev/video0)
• set threshold with 1000 as diff value from compare
• set capture dimension to 320x240
• enable verbosity as info
• capture every 2 frames
• save image on motion detect

Gui

• 'm' toggle diff/real mode.
• Filter type 0..4
• Filter value

Perfomances

Check Cpu load

top -c -p $(pgrep -d',' -f camcv)

• On i5-3320M, 320x240, filter activated, without gui, load is 2.7% per instance
• Consider file saving as greedy (+2% cpu load)

Capture consolidate

You can use ffmpeg to consolidate all capture files into a single movie.

ffmpeg -framerate 10 -pattern_type glob -i '*.jpg' -c:v libx264 -pix_fmt yuv420p consolidate.mp4

Looking further

• Zoning there
• OpenCV on esp32