Tracking of rotating objects using opencv

Question

I need to track cars on the road from top-view video.

My application contain two main parts:

Detecting cars on the frame (Tensorflow trained network)
Tracking detected cars (opencv trackers)

I have troubles with opencv trackers. Initially i tried to different trackers, but only MOSSE is fast enough. This tracker works almost perfect for case with straight road, but i faced problems with rotating cars. This situation appears on crossroads.

As i understood, bounding box of rotated object is bigger that bbox of horizontal or vertical object. As result bbox contains big part of static background and the tracker lose target object.

Are there any alternative trackers which can track contours (not bounding boxes)? Can i adjust quality of existing opencv trackers results by any settings or by adjusting picture?

Schema:

Real image:

What you want is the bounding rotated box. However, I am not sure how to use that with opencv trackers yet. — cuongptnk
– cuongptnk, Commented Nov 15, 2019 at 12:34
@SlawomirOrlowski Objects are small and have low resolution, so i had errors with tracking using SIFT — Max
– Max, Commented Nov 18, 2019 at 6:41
@Max: OK. My next shot would be GOTURN: Generic Object Tracking Using Regression Networks. — Slawomir Orlowski
– Slawomir Orlowski, Commented Nov 18, 2019 at 10:42
do you detect in each frame? Does the detector have the same problem (axis aligned bounding box)? Is the camera stationary? Should be size of the (rotated) box be constant for one vehicle? Is the actual box relevant for you? Maybe it's enough to regard the center point of the (too big) box? — Micka
– Micka, Commented Nov 20, 2019 at 14:14

ma.mehralian · Accepted Answer · 2019-11-21 09:57:24Z

2

If your camera is stationary the following scenario is feasible:

use ‌background subtraction methods to separate background image from foreground blobs.
Improve the foreground results using morphological operations.
Detect car blobs and remove other blobs.
Track foreground blobs in video i.e. binary track (simply use this or even apply KF).

answered Nov 21, 2019 at 9:57

ma.mehralian

1,3052 gold badges15 silver badges34 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

code-lukas · Accepted Answer · 2019-11-21 13:39:46Z

2

A very basic but effective approach in this scenario might be to track the center coordinates of the bounding box, if the center coordinates only change along one axis (with a small tolerance for either axis), its a linear motion (not a rotation). If both x and y change, the car is moving in the roundabout.

This only has the weakness that it will detect diagonal motion, but since you are looking at a centered roundabout, that shouldn't be an issue.

It will also be very efficient memory-wise.

answered Nov 21, 2019 at 13:39

code-lukas

1,68113 silver badges20 bronze badges

Comments

Ben Parry · Accepted Answer · 2019-11-15 13:35:19Z

1

You should use PCA method, which can calculate the orientation of an detected object and which way it is facing. You can change the threshold of detection to select objects more like the cars (based upon shape and colour - a HSV conversion which in your case is red) in your picture.

Link to an introduction to Principal Component Analysis (PCA)

answered Nov 15, 2019 at 13:35

Ben Parry

351 silver badge9 bronze badges

2 Comments

Max Over a year ago

objects are detected only on first frame. After detection they should be tracked using tracker and input bbox. I added real image (cars can have any colors)

Ben Parry Over a year ago

After, an object is detected using MOOSE, apply a mask and then apply the PCA method per frame. The mask shall blank out all of the additional parts of the frame leaving only the cars in the image. Thus allowing the PCA method to process the image and find the orientation of the cars.

Ziri · Accepted Answer · 2019-11-22 09:05:04Z

1

Method 1 :

      - Detect bounding boxes and subtract the background to get blobs rotated rectangles.

Method 2 :

      - implement your own version of detector with  rotated boxes.

Method 3 :

      - Use segmentation instead  ... Unet for example.

answered Nov 22, 2019 at 9:05

Ziri

7469 silver badges21 bronze badges

Comments

fogx · Accepted Answer · 2019-11-22 14:21:43Z

1

There are no other trackers than the ones found in the library.
Your best bet is to filter the image and use findcontours. Optical flow and background subtraction will help with this. You can combine optical flow with your car detector to rule out false positives. https://docs.opencv.org/3.4/d4/dee/tutorial_optical_flow.html https://docs.opencv.org/3.4/d1/dc5/tutorial_background_subtraction.html

answered Nov 22, 2019 at 14:21

fogx

1,8302 gold badges19 silver badges46 bronze badges

3 Comments

Max Over a year ago

1. Optical flow doesn't work properly for me, because objects are small and points which returned from goodFeaturesToTrack() are not good :/

Max Over a year ago

2. I already used background subtraction but it doesn't work correct with static object

fogx Over a year ago

you can pass in custom features as well. Background subtraction needs multiple frames to detect a background.This should be no problem, if you have a camera mounted above

Collectives™ on Stack Overflow

Tracking of rotating objects using opencv

5 Answers 5

Comments

Comments

2 Comments

Comments

3 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

5 Answers 5

Comments

Comments

2 Comments

Comments

3 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related