Problem

This is a question my friend in mechanical engineering asked me. However, I am not an expert in computer vision/image processing, so I come here to seek help.

Data

The following are two frames of a video, which records the positions of a 3D printer arm (the vertical bright curve).

Objective

The objective is to track the positions of the arm and do some computation based on this. I know there are basically two steps to do tracking

• Step 1: identify the object in each frame of the video and draw a bounding box around the object.
• Step 2: measure the relative position difference of bounding boxes of two consecutive frames.

Difficulty

The difficulty here is objective detection.
More specifically,

• If I try to use model-based approach. It is not easy to model the shape of the curve since it is deformable.
• If I try to use some object detection package (deep learning-based approach). I am not sure which package supports the tracking of a curve like this.

Could someone help me on this. Thank you in advance.