How can I implement kalman prediction model on my data

I trained a custome cascade classifier to detect an object from a video file with dynamic background. I then acheived the centroid of each detected bbox from each video frame and based on the tickenss of a known parameter of the object, i then plot the object trajectory in millimetre vs time. The problem that I am facing is lost tracks at certain video frames, mostly due to lack of enough illumination and reflection of the sun into the camera lens. I want to predict the object location based on its last location detected before the lost tracks and the first detected location after the lost tracks.

I attached part of the trajectory,

( according to the application of this task, the importance of the existing false positive is insignificant

I appreciate any help very much in advance