litterbugcam/litterbug-algorithm

Abandoned Object Detection & Illegal Dumping Detection

FYI: OpenCV is only used for image aquisition and preprocessing steps. The core routines are written in C/C++. The readability of the code will be improved soon !

LitterDetection

This a program for Litter detection events in video surveillance. An object (of any type) is detected secondes after its owner has left it in the camera FOV. The algorithm is an improved version of the algorithm we first proposed in :

Ilias, D. A. H. I., et al. "An edge-based method for effective abandoned luggage detection in complex surveillance videos." Computer Vision and Image Understanding 158 (2017): 141-151.

LitterBug Project

This program is part of the LitterBug project, a battery powered camera system for litter prevention.

For more information, please visit our website https://litterbug.cam/

Prerequisites

1. OpenCV

Detection example

check this video !

Configuration

You can modify the Parameters.txt file to adapt the algorithm to the scene you are running the algorithm on and to the type of abandoned object you are trying to detect.

Setup

Just run make command in the source code folder

Usage

./Litter_Detect videopath

Contact

For any suggestions or comment, please feel free to contact Ilias Dahi ([email protected])

litterbug-algorithm's Issues

How to work this code best

Hello,author! very good work. I noticed that u annotated some code in litter_detect.cpp,which makes me can't run the code in the way like the demo did. which lines of code shoud I unannotate to make the code work best? I'll be very grateful if u can help me solve this problem.

Test on Google AIY platform (incl Pi Zero)

Prepare installation routine for MH in Ireland

RPI crash when heavy loading

Ability for Server to request a live image

Interface Items

• Favicon
• Typo in Title
• Logo
• Colour Scheme -like Litterbug
• Remove all the filters on right side

Add Logging

Making small Webserver on RPI3

For local configuration and camera visualization

Optimise to use low frame rate source video

add people detection module to optimize precision

Web Interface v1 (Server)

• Login as a user
• See images from S3
• See list of alerts

(Choose whatever platform you want, but now in OK we're using Vue.JS , Phoenix, and some dashboard e.g. https://coreui.io/vue/ )

Further Optimisations

We suspect that there are many optimisations that could be made in the Matrix Operations:

• Sobel
• Blur
• Canny
• Resize()
• CvtColor()
• CartToPolar()
• Accumulation operator ()

Possible Solutions:

• Parallelisation
• GPU Optimisation