Scribd
Upload
323 views32 pages

Drone Based Window Cleaner Robot

This document discusses using a drone to clean solar panels. It outlines using image processing and object detection to identify and localize solar panels. Initial trials used edge detection and histograms to detect objects. A custom object detector was developed using TensorFlow to identify solar panels. Simulations were conducted using the Parrot-Sphinx tool to test drone control and localization algorithms for cleaning solar panels. A research paper on this project was selected for presentation at the ICCAS 2019 conference.

Uploaded by

Kumar Saurav
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
323 views32 pages

Drone Based Window Cleaner Robot

This document discusses using a drone to clean solar panels. It outlines using image processing and object detection to identify and localize solar panels. Initial trials used edge detection and histograms to detect objects. A custom object detector was developed using TensorFlow to identify solar panels. Simulations were conducted using the Parrot-Sphinx tool to test drone control and localization algorithms for cleaning solar panels. A research paper on this project was selected for presentation at the ICCAS 2019 conference.

Uploaded by

Kumar Saurav
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 32

Drone Controlled Solar Panel

Cleaner

Kumar Saurav
Objective of the Project

Image Processing using OpenCv


Object Detection (any open source module
available can be modified and the libraries
can be changed according to the need)
Object Identification using Deep learning(In
this case of course it’s a Solar Panel)
Drone control and localisation
and finally, merging the sub-parts together
Real TimeImage-processing(Initial trials)

Face recognition without using Deep Neural Networks


(Works but not accurate)
Initial Trials (Edge Detection using OpenCv)

Detecting edge of a sample using Open cv(comparing the consecutive gradients and the peak sho
Perception of Image Processing and face detection

A human face as observed by computer

How we think of Image processing vs How the computer actually observes the human face(using the
The Histogram method explained
(how gradient is calculated using the adjacent members)
Storing the Calculated Gradient Values in Histogram
Image Segmentation
The observed Outputs:

Initial Picture taken: After Gray-Scaling: After OTSU’s Binarisation:


Noise reduction: Sure-areas: After Applying watershed algorithm:
Raspberry Pi and Altimetre

For Stand-alone implementations of


Height monitoring and control
Working Video of Altimeter(BMP 388) using Raspberry-
pi
Custom Object Detector using
TensorFlow
TensorFlow is a free and open-source software library for dataflow and
differentiable programming across a range of tasks. It is a symbolic math
library, and is also used for machine learning applications such as neural
networks
Steps involved in Creating Solar panel detector
Tensor board Graph of training

This Graph depicts all the steps involved in the training purpose and its so complicated so at the sim
Detection technique involved in SSD models

The CNN(Convolution Neural network method uses the convolution layers to extract features from th
Neural Networks layer diagram and output
(TensorFlow)

The layers shows different nodes and paths of the neural network graphs along with the corresp
Testing the developed detector
Real-time Solar panel detection using the developed system

The Video shows real-time detection of solar panel place in the lab
Flow Chart of Complete detector
Drone
Images Captured using drone
Proposed Flow-chart for drone localisation
Complete Bebop Drone flow chart
Simulations for drone applications

Parrot-Sphinx is a simulation tool initially thought to cover the needs of Parrot engineers developing d
The main concept is to run a Parrot drone firmware on a PC, in an isolated environment well separate
It comes with several friendly features:
• Visualize flight data at runtime via your web browser
• Modify drone’s behaviour at runtime
• Entirely scriptable (using command lines)
• Build your own scenes
• Video sensor and streaming
• Connect to any kind of controller, such as Freeflight running on a smartphone.
• Can be run remotely (e.g. on a distant server)
Simulation in Sphinx(Gazebo based) environment
Different Environments(Worlds) for simulation with different objects
Parrot-sphinx Dashboard(to monitor data)
Controlling Bebop2 in virtual environment using
Parrot Free flight pro
Research Paper Selected for Presentation at ICCAS 2019
End

You might also like