OpenCV : Vehicle axle detection

I tried to detect vehicle axles in an image. Here are the steps that I followed.

• Read the image and changed to grayscale.
• Applied median.
• Applied canny edge detection.
• Applied Hough circles.

For canny the threshold1, threshold2 parameters are varied using track bar and for Hough circles param1, param2, minDist are varied. By param1 and param2 has no impact on detection.

Original:

Median: Canny:

Detected cirlce:

Problem : As you can see in the last image, it detects the circles and it is little away from the original axle. Is there any other parameters that I need to change or consider ?

Code

import cv2
import numpy as np

def readAndResize(image,a):
    imag = cv2.imread(image,a)
    org = cv2.imread(image)
    #resixe the image
    small = cv2.resize(imag, (0,0), fx=0.3, fy=0.3)
    small2 = cv2.resize(org, (0,0), fx=0.3, fy=0.3)
    height, width = small.shape
    img = small[height - height/3:height,0:width]
    org2 = small2[height - height/3:height,0:width]
    return img, org2


img,org2 = readAndResize('5.jpg',0)
cv2.imshow('Original',org2)
gray = cv2.cvtColor(org2, cv2.COLOR_BGR2GRAY)
med = cv2.medianBlur(gray,5)
canny = cv2.Canny(med,100,50)
circles = cv2.HoughCircles(canny,cv2.HOUGH_GRADIENT,1,92,20,10,minRadius=4,maxRadius=0)
circles = np.uint16(np.around(circles))

for i in circles[0,:]:
    cv2.circle(org2,(i[0],i[1]),i[2],(0,255,0),2)
    cv2.circle(org2,(i[0],i[1]),2,(0,50,255),3)
    cv2.imshow('Median',med)
    cv2.imshow('canny',canny)
    cv2.imshow('Detected',org2)
cv2.waitKey(0)
cv2.destroyAllWindows()

Sorry but i didn't used your code. Here is some other which can help you :

import cv2
import cv2.cv as cv
import numpy as np
import sys

img = cv2.imread('5.jpg',0)
img = cv2.medianBlur(img,5)
cimg = cv2.cvtColor(img,cv2.COLOR_GRAY2BGR)
circles = cv2.HoughCircles(img,cv.CV_HOUGH_GRADIENT,1,240,
                               param1=250,
                               param2=50,
                               minRadius=5,
                               maxRadius=200)
circles = np.uint16(np.around(circles))
for i in circles[0,:]:
    cv2.circle(cimg,(i[0],i[1]),i[2],(0,255,0),2) 
    cv2.circle(cimg,(i[0],i[1]),2,(0,0,255),1)
print circles
cv2.imshow('circles',cimg)
cv2.waitKey(0)
cv2.destroyAllWindows()

Play with parameters 240 param1 param2 minRadius maxRadius Additionally what i got with this code Sample