Catfood

Catfood is a Python script working in a RaspberryPi 3 with Raspbian. Catfood recognizes one cat on two different cats with Opencv and opens access to the food for the recognized cat.

Presentation available on Youtube

Prerequisites

To recognize the right cat, you must first generate an xml file Cascade Classifier with opencv.

-- For example --

While the good cat was eating, I started a script detecting the cat's face and cropped the image in the center of his forehead so that he had his own marks.

for frame in camera.capture_continuous(rawCapture, format="bgr", use_video_port=True):
		image = frame.array
		gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
		rects = detector.detectMultiScale(gray, scaleFactor=1.25,
			minNeighbors=20, minSize=(75, 75))
		if rects < "[" :		#Si detection
      x, y, w, h = rects[0] 
      pw = (w * 15)/100 #percent width
      ph = (h * 28)/100 #percent hight
      img = frame[y+pw:(y-pw)+w,x+ph:(x-ph)+h] #cropped image

Then with about 200 good images cropped, i could start the training of classifier with opencv_traincascade.

Installing

Raspberry OS :

• Raspbian

Package :

• Python 2.7 or 3
• OpenCV 2

Materials :

• Raspberry Pi
• Two steppers motors 5v with ULN2003 driver and alimention 5v
• Raspberry Pi Night Vision Camera
• IR Break Beam Sensor
• White LED Board Specific For Raspberry Pi Camera x2
• Pushbutton x2 (one for end plate postion and one for manual opening)
• Wires

Deployment

Bash Script :

The python script is start as deamon during Raspbian start, so you have to create a bash file.

#!/bin/sh
### BEGIN INIT INFO
# Provides:          MonaFood
# Required-Start:    $remote_fs $syslog
# Required-Stop:     $remote_fs $syslog
# Default-Start:     2 3 4 5
# Default-Stop:      0 1 6
# Short-Description: Start Python program at boot time
# Description:       Mona Food
### END INIT INFO

PATH=/usr/sbin:/usr/bin:/sbin:/bin
DESC='Startup script python monafood'
NAME=monafood
DAEMON=/usr/sbin/$NAME
PIDFILE=/var/run/$NAME.pid
SCRIPTNAME=/etc/init.d/$NAME

case "$1" in
   start)
       python /home/pi/camera/monafood.py
   ;;

   *)
        echo 'Usage: /etc/init.d/monafood {start}'
       exit 1
   ;;
esac
exit 0

Install

 sudo chmod 0755 /etc/init.d/monafood

sudo update-rc.d monafood defaults

Next Update

• The success rate is around 95%, the bad cat can be detected. The black pixels are counted and list when there is a detection. So to put a second identification with the number of black pixel counted.

nb = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
ret,thresh = cv2.threshold(nb,127,255, cv2.THRESH_BINARY)
blackp = cv2.countNonZero(thresh)

• Problem detect during the night. The detection is longer and difficult with low light. A third source of light is to consider.