Car detection

asked 2014-03-18 19:31:16 -0500

chelo666 gravatar image

updated 2014-03-18 20:20:12 -0500

Recently I've been working on a program that allows you to detect when ever there's a parking space available. But my algorithm ran into a problem. I'm using simplecv and what I did was crop the parking as close as possible to the car and .meanColor() the crop and comparing that to a .meanColor() taken from and empty parking space. It was working fine with white and dark cars, but when I tried it with others colors that weren't so dark or light, the values were lower than the values of the empty space. I started looking for an alternative solution and stumbled upon the tutorial on car detection that simplecv provides, BUT it worked if you hardcoded the color. I don't want to hard code the color, I need to identify it or try something similar that can help me detect there's a car there.

This is what I tried :

from SimpleCV import *

imgOri = Image("allCars3.jpg")

imgOri = imgOri.scale(400, 400)

corX = 0  #ancho del parking

parking = 0

imgList = []

for i in range(3):

    imgList.append(imgOri.crop(corX, 70, 133, 120))

    avg = imgList[i].meanColor()

      if avg[0] < 123 and avg[1] < 126 and avg[2] < 127:

          parking += 1        

    corX += 133


worked with these type of cars:

https://scontent-a-mia.xx.f... (picture too big to upload here)

Didn't work with these three:


The values I took to detect a car where taken from this picture:


This is the tutorial I found for car detection:


Notice that the person how wrote this tutorial already knew the color. I need the program to figure out that color in order to use that same idea.

Any suggestions, questions, tips or ideas are welcomed.


edit retag flag offensive close merge delete