Revision history [back]

There are basically two cases:

• the object is detected, you have to perform the prediction and correction step with the new measurements (the current detected position)
• the object is not detected, you can predict the position of the object using the Kalman Filter (of course you cannot predict the position for a very long time and you have to set a thresold where you decide that the object is definitely lost).

I quickly read the code you posted but you don't have to recreate a Kalman Filter at each image.

There are basically two cases:

• the object is detected, you have to perform the prediction and correction step with the new measurements (the current detected position)
• the object is not detected, you can predict the position of the object using the Kalman Filter (of course you cannot predict the position for a very long time and you have to set a thresold where you decide that the object is definitely lost).