Predict trajectory of an Object with Kalman filter

Blog: pysource.com/2021/11/02/kalma...
In this video, you will learn how you can predict the trajectory of an orange.
How did this algorithm work?
I threw an orange in the air by drawing a parable. The algorithm was able to predict the trajectory in real-time. I'll explain everything in the tutorial
➤ Courses:
Full Computer Vision course: pysource.com/object-detection...
Training Mask R-CNN PRO (Notebook + Mini-Course): pysource.com/mask-rcnn-traini...
➤ Follow me on:
LinkedIn: / pysource
➤ For business inquiries:
pysource.com/contact
#kalmanfilter #OpenCV #computervision

[@pysource-com]

►► You can Build a Computer Vision software to DETECT and TRACK any Object. → 4-Step FREE Workshop pysource.com/blueprint-workshop-signup/

[@inovotny1412]

Whenever i try to sign up to download the source code it says that my email cannot be added

[@gul156]

Great video explaining how and why predictions are important in tracking objects. I would have liked to understand more on how the Kalman filter works - it is being treated as a "black-box". May be in future you'd consider explaining the inner workings of this filter.

[@buketkaraoglu683]

Thanks this video and other videos. They’re really helpful . And how can i implement extended kalman filter in 3 dimension(x,y,z position). Any examples about that or learn more about this topic?

[@McBobX]

Great video but we will be extremely happy if you explain to us how Kalman Filter works and how we can change the measurement/transition matrices in case it fails to predict the position we want

[@feysalmustak9604]

you can also do a project with an AI such as LSTM to detect a rotating object and predict its impact, such as a golf ball.

[@maxius211]

instead of cx = int((x+x2) / 2) you can use the // operator and write cx = (x+x2) // 2 since x , x2 are integer

[@sachinadnaik8106]

Great work. Thanks for sharing.

[@aiforyounow]

Hello Sergio this a good tutorial and I love it but I will be glad if you can be telling us how to use them practically.Thanks soo much

[@Alessandroale74]

Con quale webcam, che usi per le elaborazioni in opencv, ti stai riprendendo ? Mi pare registri in 1080p a 60fps o sbaglio ?

[@pysource-com]

La webcam che utilizzo di solito è la Logitech c922pro. Però per questo specifico video ho utilizzato la vlog camera Sony ZV1 a 30FPS 1080p

[@Faisalamin01]

Bro thank you so much for the video, really appreatiat it. What are the other algorithms for the same purpose?

[@fabriziofilograna6498]

Ottimo grazie Sergio

[@maheshpatel2005]

Nice

[@arslanmanzoor8716]

can you please make the video to predict multiple objects in different polygon?? i will be thankful to you

[@jaugretler9140]

Well structured video, nice visualization of the concept!

[@gplgomes]

Throwing objects, it does a parabolic curve so instead of Kalman filter, it can be used the basic quadratic polynomials.

[@yjaspal2011]

Is there a way to make this code so that it isn't slow?

[@sercanalnca1962]

Damnnn man. I am beginner at python. This video teach me too much thing about my hobby. I like your explaining all those things. Thank you soooo mucchhh. Greetings from Turkey. Have a good day.

[@imenecharabi8945]

Cool

[@hanshima_]

Thanks for the great video.

[@DenerWerb]

Haven’t watched yet. But I’m sure it’s gonna be another killer demonstration. Thank you for your work.

[@pysource-com]

thanks for the trust ;)

[@lukaszg123]

It is great video! thanks for sharing

[@alexbarnadas]

Thank you so much!

[@lushankshambharkar124]

hello sergio i want to detect and track object continuously in a video even if there is some environmental occlusion or obstruction. For example if a walking man is obstructed by a vehicle or a tree. i want the algorithm to track man continuously in a video frame even in obstruction. How to deal with this problem????🙏🙏🙏🙏🙏🙏🙏

[@pysource-com]

You should use algorithms like Sort and Deep SORT. I explained that in the workshop computer vision blueprint that you can find in my website pysource.com

[@lushankshambharkar124]

@@pysource-com Thank you sergio 🤟✌

[@actechforlife]

Does anyone else finds Sergio’s accent really nice at 1.5speed?

[@sriharibg6168]

Nope

[@olaff667]

it would be better to save ball prediction and draw it on next frame, together with real position. to see the error. great video nonetheless

[@mohammadaliavazpour3268]

Your are doing great

[@dimasalifta1966]

i got error "ValueError: need more than 2 values to unpack"

[@pysource-com]

You might be using an older version of Opencv, 3.x.. If you upgrade opencv to the latest version 4.5 , this problem will be solved

[@PCEBhanuSharma]

Sir, can you please tell me how to consider observation or measurement matrix in Extended Kalman filter for multi-degree of freedom structure?

[@buketkaraoglu683]

Did you learn that ? Did you find source or sample about this?

[@sheetalkumar4579]

Please upload more videos to your Nerual Networks from scratch series, about a network with more than one layer. Thankyou

[@faraz7123]

can this be used for golf ball trecking???

[@hyunew]

how can i get prediction's distribution?

[@gauraomate6587]

How can we implement Kalman filter with yolov5 or YOLOv3

[@pysource-com]

You should change the part of the code I'm currently using to track the orange by it's color with yolo object detection. Normally kalman filter alone is not used, but it's the core of two more advanced algorithms like (Sort and Deep Sort)

[@gauraomate6587]

@@pysource-com so I need to go alone with Kalman or think about Deepsort, which one is better

[@pysource-com]

@@gauraomate6587 I recommend Sort or Deep Sort (instead of Kalman alone)

[@gauraomate6587]

@@pysource-com Thank you,this really helps me

[@gauraomate6587]

Can we draw future trajectory line with Deepsort?

[@pysource-com]

Yes. Now I'm displaying just the future single point predicted, but you can loop through future points as I did at the beginning of the script. and you can connect the points with a cv2.polylines() function to draw the line

[@gauraomate6587]

Can we do it for multiple objects at the same time like predicting the future trajectory lines for multiple objects in a frame

[@pysource-com]

@@gauraomate6587 Yes, it's possible though it's a bit more complex because you need to also associate a univocal ID to each object

[@gauraomate6587]

@@pysource-com Thanks

[@bks637]

Hello sir, how to make face recognition with app lock using python interface

[@azizutkuozdemir]

Can this be applied to stock market ?

[@keepsecret7624]

OK, but a PDF much better to learn - using the video to demonstrate points.