Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

Is there any example of cv2.KalmanFilter implementation?

I'm trying to build a veeery simple tracker for 2D objects using python wrapper for OpenCV (cv2).

I've only noticed 3 functions:

  • KalmanFilter (constructor)
  • .predict()
  • .correct(measurement)

My idea is to create a code to check if kalman is working like this:

kf = cv2.KalmanFilter(...)
# set initial position

cv2.predict()
corrected_position = cv2.correct([measurement_x, measurement_y])

I've found some examples using the cv wrapper but not the cv2...

Thanks in advance!

like image 723
Jairo Vadillo Avatar asked Mar 12 '15 14:03

Jairo Vadillo


People also ask

What is Kalman filter in OpenCV?

OpenCV Kalman filter is a class of method used to implement the standardized Kalman filter. Let us first have a look at what is the use of the Open CV Kalman filter. It is predefined, which is used to equate for an algorithm that is known to use a series of observed measurements taken over an observational time period.

How use Kalman filter in Python?

A Kalman Filtering is carried out in two steps: Prediction and Update. Each step is investigated and coded as a function with matrix input and output. These different functions are explained and an example of a Kalman Filter application for the localization of mobile in wireless networks is given.


1 Answers

if you're using opencv2.4, then it's bad news: the KalmanFilter is unusable, since you cannot set the transition (or any other) Matrix.

for opencv3.0 it works correctly, like this:

import cv2, numpy as np

meas=[]
pred=[]
frame = np.zeros((400,400,3), np.uint8) # drawing canvas
mp = np.array((2,1), np.float32) # measurement
tp = np.zeros((2,1), np.float32) # tracked / prediction

def onmouse(k,x,y,s,p):
    global mp,meas
    mp = np.array([[np.float32(x)],[np.float32(y)]])
    meas.append((x,y))

def paint():
    global frame,meas,pred
    for i in range(len(meas)-1): cv2.line(frame,meas[i],meas[i+1],(0,100,0))
    for i in range(len(pred)-1): cv2.line(frame,pred[i],pred[i+1],(0,0,200))

def reset():
    global meas,pred,frame
    meas=[]
    pred=[]
    frame = np.zeros((400,400,3), np.uint8)

cv2.namedWindow("kalman")
cv2.setMouseCallback("kalman",onmouse);
kalman = cv2.KalmanFilter(4,2)
kalman.measurementMatrix = np.array([[1,0,0,0],[0,1,0,0]],np.float32)
kalman.transitionMatrix = np.array([[1,0,1,0],[0,1,0,1],[0,0,1,0],[0,0,0,1]],np.float32)
kalman.processNoiseCov = np.array([[1,0,0,0],[0,1,0,0],[0,0,1,0],[0,0,0,1]],np.float32) * 0.03
#kalman.measurementNoiseCov = np.array([[1,0],[0,1]],np.float32) * 0.00003
while True:
    kalman.correct(mp)
    tp = kalman.predict()
    pred.append((int(tp[0]),int(tp[1])))
    paint()
    cv2.imshow("kalman",frame)
    k = cv2.waitKey(30) &0xFF
    if k == 27: break
    if k == 32: reset()
like image 92
berak Avatar answered Sep 21 '22 20:09

berak