Anomaly Detection

This service takes as input video/photos with metadata (in KLV or Mavlink format) of drone flight (live or recorded) and generates alerts or incidences of anomalies (e.g. major alteration of soil, movement of structure, etc).

The service implements a collection of algorithms that have main functionalities of background subtraction in dynamic sequences of images, segmentation of objects and morphological operations on images allowing the identification of any kind of changes in a specific scenario.

  • Background segmentation using K-Nearest Neighbours
  • Background segmentation using Gaussian Mixture
  • Segmentation through Deep Neural Networks
  • Utilities for morphological transformation of images, erosion, dilation, opening, closing, binarization, etc.


anomaly components
  1. Enhanced Background Subtractor: This module reads the video streaming images and applies analysis algorithms described. Its output are the 2D coordinates of a rectangle containing the identified anomaly and the frame where the identification has been made.

Execution and deployment

raspberry pi
On-Board Edge device
  • NDIVIA Jetson Nano
  • Raspberry PI 4 – 2GB RAM
  • Raspberry PI 3 B+