Real-Time Multi-Object Recognition Using the Fusion of LIDAR and Camera Data

Mert Can Yaman; Şerafettin Erel

Article Details

Real-Time Multi-Object Recognition Using the Fusion of LIDAR and Camera Data

Authors:

Mert Can Yaman | Şerafettin Erel

Publisher:

Yozgat Bozok University

Main Topics:

Bilgisayar Görüşü | Video İşleme | Derin Öğrenme | Yazılım Mühendisliği (Diğer) | Elektronik

Journal:

Bozok Journal of Engineering and Architecture Vol.2 Issue 2, pp.1-19

Document Type:

Article

Publication Date:

12/2023

Document Language:

English

Document ID (Org No.):

2001307897

pISSN:

3023-4298

Keywords:

Abstract:

 Object recognition is one of the most significant research topics of today. The significance of object recognition, which has extensive use, will increase gradually. In this study, real-time object recognition was performed within the user-definable area with the data taken simultaneously from the built-in camera and LIDAR sensor of iPhone 13 Pro Max. The study was performed using MS-COCO dataset, Swift language and using SwiftUI as the framework. YOLO V5 algorithm is used for object recognition and video processing was performed with Swift Metal narrowing the area by the minimum-maximum distance determined in the interface on each frame depending on the real-time fusion data of camera and LIDAR. The areas outside the contours of objects within the value range are darkened. Thus, object recognition was performed in each darkened frame. In the study, object recognition was performed in the range of 0-15 m, which can be adjusted in the interface.

 Object recognition is one of the most significant research topics of today. The significance of object recognition, which has extensive use, will increase gradually. In this study, real-time object recognition was performed within the user-definable area with the data taken simultaneously from the built-in camera and LIDAR sensor of iPhone 13 Pro Max. The study was performed using MS-COCO dataset, Swift language and using SwiftUI as the framework. YOLO V5 algorithm is used for object recognition and video processing was performed with Swift Metal narrowing the area by the minimum-maximum distance determined in the interface on each frame depending on the real-time fusion data of camera and LIDAR. The areas outside the contours of objects within the value range are darkened. Thus, object recognition was performed in each darkened frame. In the study, object recognition was performed in the range of 0-15 m, which can be adjusted in the interface.

 Object recognition is one of the most significant research topics of today. The significance of object recognition, which has extensive use, will increase gradually. In this study, real-time object recognition was performed within the user-definable area with the data taken simultaneously from the built-in camera and LIDAR sensor of iPhone 13 Pro Max. The study was performed using MS-COCO dataset, Swift language and using SwiftUI as the framework. YOLO V5 algorithm is used for object recognition and video processing was performed with Swift Metal narrowing the area by the minimum-maximum distance determined in the interface on each frame depending on the real-time fusion data of camera and LIDAR. The areas outside the contours of objects within the value range are darkened. Thus, object recognition was performed in each darkened frame. In the study, object recognition was performed in the range of 0-15 m, which can be adjusted in the interface.

 Object recognition günümüzün en önemli araştırma konularındandır. Tarımdan, savunma ve uzay sanayisine kadar her alanda kapsamlı kullanımı olan nesne tanımanın öneminin giderek artacağı öngörülmektedir. Çalışmada, iPhone 13 Pro Max dahili kamerası ve LIDAR sensöründen eş-zamanlı olarak alınan verilerle kullanıcı tarafından sınırları belirlenebilen alan içerisinde gerçek-zamanlı nesne tanıma yapılmıştır. Çalışmada, MS-COCO veri seti ile Swift dili ve framework olarak SwiftUI kullanılarak gerçekleştirilmiştir. Nesne tanıma için YOLO V5 algoritması kullanılmış ve video işleme Swift Metal kullanılarak gerçek-zamanlı veriler üzerinde gerçekleştirilmiştir. Kamera ve LIDAR füzyonuna bağlı olarak gerçek-zamanlı olarak alınan verilerden elde edilen her bir frame üzerinde kullanıcının arayüzde belirlediği minimum-maksimum uzaklığa göre görüntü alanı daraltılarak Swift Metal ile video işleme gerçekleştirilmiştir. Gerçek zamanlı videoda her bir frame içerisinde kullanıcının belirlediği değer aralığındaki nesnelerin konturları dışındaki alanlar karartılmıştır. Böylece karartılmış olan her bir framede nesne tanıma gerçekleştirilmiştir. Çalışmada, arayüzde slider ile ayarlanabilen 0-15 m aralığında gerçek-zamanlı LIDAR ve kamera verileriyle nesne tanıma gerçekleştirilmiştir.

Article Details

Real-Time Multi-Object Recognition Using the Fusion of LIDAR and Camera Data

Authors:

Mert Can Yaman | Şerafettin Erel

Publisher:

Yozgat Bozok University

Main Topics:

Bilgisayar Görüşü | Video İşleme | Derin Öğrenme | Yazılım Mühendisliği (Diğer) | Elektronik

Journal:

Bozok Journal of Engineering and Architecture Vol.2 Issue 2, pp.1-19

Document Type:

Article

Publication Date:

12/2023

Document Language:

English

Document ID (Org No.):

2001307897

pISSN:

3023-4298

LIDAR

Object Recognition

Data Fusion

Video Processing

iOSLIDAR

Nesne Tanıma

Veri Füzyonu

Video İşleme

iOS

Issues

Article Details

Authors: Mert Can Yaman | Şerafettin Erel

Publisher: Yozgat Bozok University

Main Topics: Bilgisayar Görüşü | Video İşleme | Derin Öğrenme | Yazılım Mühendisliği (Diğer) | Elektronik

Journal: Bozok Journal of Engineering and Architecture Vol.2 Issue 2, pp.1-19

Document Type: Article

Publication Date: 12/2023

Document Language: English

Document ID (Org No.): 2001307897

pISSN: 3023-4298

Issues

Authors:

Mert Can Yaman | Şerafettin Erel

Publisher:

Yozgat Bozok University

Main Topics:

Bilgisayar Görüşü | Video İşleme | Derin Öğrenme | Yazılım Mühendisliği (Diğer) | Elektronik

Journal:

Bozok Journal of Engineering and Architecture Vol.2 Issue 2, pp.1-19

Document Type:

Article

Publication Date:

12/2023

Document Language:

English

Document ID (Org No.):

2001307897

pISSN:

3023-4298