REAL-TIME PERSONAL PROTECTIVE EQUIPMENT AND WAREHOUSE SAFETY DETECTION WITH DEEP LEARNING-BASED WORKPLACE CAMERA

Abdil Karakan; Yüksel Oğuz

Makale detayları

REAL-TIME PERSONAL PROTECTIVE EQUIPMENT AND WAREHOUSE SAFETY DETECTION WITH DEEP LEARNING-BASED WORKPLACE CAMERA

Yazarlar:

Abdil Karakan | Yüksel Oğuz

Yayıncı:

Adıyaman University

Ana Başlıklar:

Elektrik Mühendisliği (Diğer)

Dergi:

Adıyaman Üniversitesi Mühendislik Bilimleri Dergisi Vol.11 Issue 24, pp.402-414

Belge Tipi:

Makale

Yayın Tarihi:

12/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001329187

pISSN:

2149-0309

Anahtar kelimeler:

Özet:

 The majority of work accidents can be prevented with simple precautions. The most important of these is the personal protective equipment that employees must use. In the study, personal protective equipment and warehouse security were detected in real time with images taken from a workplace camera. For this purpose, a data set was created from images taken from the workplace camera. This data set consists of 6125 photographs. Additionally, grayscale, tilt addition, blurring, variability addition, noise addition, image brightness change, color vibrancy change, perspective change, resizing and position change have been added to the photographs. With these additions, the error that may occur due to any distortion that may occur from the camera is minimized. With the changes made to the photographs, the number of photographs forming the data set increased to 21079. The created data set was run on YOLOv8 architecture. In the study, 9 types of personal protective equipment and warehouse safety were determined: helmet, shoes, vest, on the road, not on the road, without vest, without shoes, apron and without helmet. As a result of the study, average stability was 97.3, mAP was 93.8 and recall was 91.7.

 The majority of work accidents can be prevented with simple precautions. The most important of these is the personal protective equipment that employees must use. In the study, personal protective equipment and warehouse security were detected in real time with images taken from a workplace camera. For this purpose, a data set was created from images taken from the workplace camera. This data set consists of 6125 photographs. Additionally, grayscale, tilt addition, blurring, variability addition, noise addition, image brightness change, color vibrancy change, perspective change, resizing and position change have been added to the photographs. With these additions, the error that may occur due to any distortion that may occur from the camera is minimized. With the changes made to the photographs, the number of photographs forming the data set increased to 21079. The created data set was run on YOLOv8 architecture. In the study, 9 types of personal protective equipment and warehouse safety were determined: helmet, shoes, vest, on the road, not on the road, without vest, without shoes, apron and without helmet. As a result of the study, average stability was 97.3, mAP was 93.8 and recall was 91.7.

 The majority of work accidents can be prevented with simple precautions. The most important of these is the personal protective equipment that employees must use. In the study, personal protective equipment and warehouse security were detected in real time with images taken from a workplace camera. For this purpose, a data set was created from images taken from the workplace camera. This data set consists of 6125 photographs. Additionally, grayscale, tilt addition, blurring, variability addition, noise addition, image brightness change, color vibrancy change, perspective change, resizing and position change have been added to the photographs. With these additions, the error that may occur due to any distortion that may occur from the camera is minimized. With the changes made to the photographs, the number of photographs forming the data set increased to 21079. The created data set was run on YOLOv8 architecture. In the study, 9 types of personal protective equipment and warehouse safety were determined: helmet, shoes, vest, on the road, not on the road, without vest, without shoes, apron and without helmet. As a result of the study, average stability was 97.3, mAP was 93.8 and recall was 91.7.

 İş kazalarının büyük bir çoğunluğu basit tedbirlerle önlenebilecek seviyededir. Bunların başında çalışanların kullanması gereken kişisel koruyucu ekipmanları gelmektedir. Yapılan çalışmada bir iş yeri kamerasından alınan görüntüler ile gerçek zamanlı olarak kişisel koruyucu ekipmanlarının tespiti gerçekleştirilmiştir. Bunun için iş yeri kamerasından alınan görüntülerden bir veri seti oluşturulmuştur. Bu veri seti 6125 tane fotoğraftan oluşmaktadır. Ayrıca fotoğraflar üzerinde gri tonlama, eğim eklenmesi, bulanıklaştırma, değişkenlik eklenmesi, gürültü eklenmesi, görüntü parlaklığı değişikliği, renk canlılığı değişikliği, perspektif değişikliği, boyutlandırma ve konum değişikliği eklenmiştir. Bu eklemeler ile kameradan meydana gelebilecek herhangi bir bozulmaya karşı oluşacak hata en aza indirilmiştir. Fotoğraflar üzerinde yapılan değişiklikler ile veri setini oluşturan fotoğraf sayısı 21079a yükselmiştir. Oluşturulan veri seti YOLOv5 mimarisinde çalıştırılmıştır. Çalışmada kask, ayakkabı, yelek, yolda, yolda değil, yeleksiz, ayakkabısız, apron ve kasksız olmak üzere 9 çeşit kişisel koruyucu ekipmanın tespiti gerçekleştirilmiştir. Çalışma sonucunda ortalama doğruluk 97.3, mAP 93.8 ve recall 91.7 gerçekleşmiştir.

Makale detayları

REAL-TIME PERSONAL PROTECTIVE EQUIPMENT AND WAREHOUSE SAFETY DETECTION WITH DEEP LEARNING-BASED WORKPLACE CAMERA

Yazarlar:

Abdil Karakan | Yüksel Oğuz

Yayıncı:

Adıyaman University

Ana Başlıklar:

Elektrik Mühendisliği (Diğer)

Dergi:

Adıyaman Üniversitesi Mühendislik Bilimleri Dergisi Vol.11 Issue 24, pp.402-414

Belge Tipi:

Makale

Yayın Tarihi:

12/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001329187

pISSN:

2149-0309

Deep learning

Personal protective equipment

Real-time object detection

YOLO architecture

Warehouse security.Derin öğrenme

Kişisel koruyucu ekipman

Gerçek zamanlı nesne algılama

YOLO mimarisi

Depo güvenliği.

Sayı

Makale detayları

Yazarlar: Abdil Karakan | Yüksel Oğuz

Yayıncı: Adıyaman University

Ana Başlıklar: Elektrik Mühendisliği (Diğer)

Dergi: Adıyaman Üniversitesi Mühendislik Bilimleri Dergisi Vol.11 Issue 24, pp.402-414

Belge Tipi: Makale

Yayın Tarihi: 12/2024

Belge Dili: Türkçe

Document ID (Org No.): 2001329187

pISSN: 2149-0309

Sayı

Yazarlar:

Abdil Karakan | Yüksel Oğuz

Yayıncı:

Adıyaman University

Ana Başlıklar:

Elektrik Mühendisliği (Diğer)

Dergi:

Adıyaman Üniversitesi Mühendislik Bilimleri Dergisi Vol.11 Issue 24, pp.402-414

Belge Tipi:

Makale

Yayın Tarihi:

12/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001329187

pISSN:

2149-0309