(For USM Staff/Student Only)

EngLib USM > Ω School of Aerospace Engineering >

Drone based multi geometry target detection using opencv and python

Drone based multi geometry target detection using opencv and python / Mohamad Hanif Mukhlis Muhamad Hanizar
Pada masa kini, pengesanan sasaran menggunakan visi komputer semakin meningkat kerana kemudahan teknologi yang mudah digunakan dalam sektor komersial. Walau bagaimanapun, pengekstrakan ciri sasaran sedemikian sentiasa mencabar bagi pengaturcara. Tujuan tesis ini adalah untuk melaksanakan satu algoritma untuk logo berasaskan dron mengikut sempadan persegi menggunakan Python dan buka visi komputer (OpenCV). Tesis ini menyediakan kaedah terperinci dalam melaksanakan algoritma untuk pengecaman logo berasaskan objek menggunakan pengesanan ciri berasaskan bucu dan anggaran kontur. Algoritma dibangunkan supaya ia mampu mengesan pelbagai objek berdasarkan logo dengan menggunakan kotak bingkai pada imej frame. Projek ini menggunakan Python sebagai bahasa pengaturcaraan dan OpenCV sebagai Perpustakaan sumber terbuka untuk pengaturcaraan. Imej diambil dari DJI Mavic air. Teknik pra-pemprosesan yang berkaitan telah digunakan sebelum peringkat pengesanan bucu. Pengesan “Edge Canny” digunakan untuk pengiktirafan pinggir dalam logo tertentu. Selepas pelaksanaan, kontur telah dikeluarkan untuk pengiktirafan sempadan logo. Pangkah silang menyiarkan telah dilukis pada bingkai untuk tujuan pengecaman objek. Ujian ini dilakukan untuk saiz logo yang berbeza dengan mengubah ciri spesifikasi sempadan untuk mengesahkan ciri yang diperlukan dan keputusan masalah. Ujikaji ini dilakukan dengan menggunakan pelbagai nilai perimeter sempadan (epsilon) untuk menganggarkan ketepatan pengesanan logo dengan menggunakan masa pelaksanaan yang lebih rendah oleh mengeksploitasikan Python berasaskan multiprocessing. Sudut unjuran dan jarak antara dron dan objek dimasukkan dalam pemerhatian untuk menganggarkan ketepatan pengesanan logo berdasarkan perimeter epsilon dan nisbah aspek. Hasil daripada projek ini mencerminkan teknik pengesanan objek logo, yang akhirnya akan diubahsuai dengan pelaksanaan visi mesin dan seterusnya akan membangunkan pembangunan pemprosesan imej dalam kecerdasan buatan. _______________________________________________________________________________________________________ Nowadays, target detection using computer vision has been gaining prominence due to its ease of deployment in the commercial sector. However, the feature extraction of such targets has always been challenging for the programmers. The purpose of this thesis is to implement an algorithm for a drone-based square boundaries-based logo using Python and Open Computer Vision (OpenCV). This thesis provided a detailed method in implementing the algorithm for an object-based logo recognition using edge-based feature detection and contour approximation. The algorithm is developed so that it is capable to detect various objects-based logos by applying a bounding box on the frame image. This project uses Python as its programming language and OpenCV as an open-source library for programming. The images were taken from the DJI Mavic Air drone. Relevant pre-processing techniques were applied before the edge detection stage. The Canny edge detector was used for the recognition of the edges in the proposed logo. Upon implementation, the contours were drawn for the recognition of the border of the logo. The crosshairs were drawn on the frame for aiming purposes. The testing was done for different logo sizes by varying its border specification characteristics to verify the required features and the problem outcome. The experiments were done by using the various value of boundary perimeter (epsilon) to estimate the accuracy of the detection of the logo using lower execution time by exploiting python-based multiprocessing. The angle of projection and the distance between the drone and the object was included in observation to estimate the accuracy detection based on the epsilon perimeter and aspect ratio. The outcome of this project reflects the logo object detection technique, which eventually will improvise the machine vision deployment and will subsequently contribute to the development of the image processing in artificial intelligence.
Contributor(s):
Mohamad Hanif Mukhlis Muhamad Hanizar - Author
Identifiers:
Accession Number : 875007945
Language:
English
Subject Keywords:
target; detection; drone-based
First presented to the public:
8/1/2020
Original Publication Date:
10/6/2020
Previously Published By:
Universiti Sains Malaysia
Place Of Publication:
School of Aerospace Engineering
Citation:
Extents:
Number of Pages - 81
License Grantor / Date Granted:
  / ( View License )
Date Deposited
2020-10-06 15:07:14.616
Submitter:
Mohd Jasnizam Mohd Salleh

All Versions

Thumbnail Name Version Created Date
Drone based multi geometry target detection using opencv and python1 2020-10-06 15:07:14.616