Python 是否有cv2.KalmanFilter实现的示例？

我正在尝试使用python wrapper for OpenCV（cv2）为2D对象构建一个非常简单的跟踪器

我只注意到3个功能：

• 卡尔曼过滤器（建造商）
• .predict（）
• .正确（测量）

我的想法是创建一个代码来检查kalman是否像这样工作：

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

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

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()

我发现了一些使用cv包装器的示例，但没有使用cv2

---

提前谢谢

如果您使用的是opencv2.4，那么这是个坏消息：KalmanFilter无法使用，因为您无法设置转换（或任何其他）矩阵

对于opencv3.0，它可以正常工作，如下所示：

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

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

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()

---

感谢您提供的信息，不幸的是需要使用opencv 2.49。。。我们正在尝试使用比其他代码简单得多的pykalman:）。但是有一个问题，最初我们必须先预测/估计，然后修正测量值，不是吗？