A convolutional neural network-based approach for the yellowrust disease detection model and comparison of methods for diseaseseverity assessment
Başlık çevirisi mevcut değil.
- Tez No: 772193
- Danışmanlar: Belirtilmemiş.
- Tez Türü: Yüksek Lisans
- Konular: Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve Kontrol, Ziraat, Computer Engineering and Computer Science and Control, Agriculture
- Anahtar Kelimeler: Belirtilmemiş.
- Yıl: 2021
- Dil: İngilizce
- Üniversite: The University of Reading
- Enstitü: Yurtdışı Enstitü
- Ana Bilim Dalı: Belirtilmemiş.
- Bilim Dalı: Belirtilmemiş.
- Sayfa Sayısı: 82
Özet
Wheat is one of the most produced cereals worldwide and is an indispensable part of the human diet. In order to ensure sustainability and food security, it is necessary to protect wheat against diseases. Yellow rust (YR) is the most serious disease of wheat, and early detection of the disease or production of disease-resistant varieties is very important in YR control. In recent years, the use of artificial intelligence technologies and automatic image analysis have increased to facilitate the detection of YR disease or identify the wheat varieties with the resistance gene. In this research, a new artificial intelligence model was developed to detect the disease on wheat leaves. A Convolutional Neural Network (CNN) based training created with the augmentation technique was conducted for the YR disease detection model. Scanned leaf images with and without YR disease were used for model training and testing. As a result of the CNN training, the validation accuracy value of the YR disease detection model was measured as 0.91. The model made an average of 86% correct predictions in the analysis made with test images. The results showed that creating a successful YR disease detection model using augmentation tools and CNN architecture is possible. During the training, Artificial intelligence was able to learn many of the parameters that would distinguish the symptoms of YR. In addition, as long as the model continues to be trained, the success rate will increase even more. ImageJ software-based methods of measuring pustule severity on diseased leaves were used. As the pustule severity measurement methods, ZymoMacro_Scanner_v2.1.1 macro was developed for Septoria tritici blotch assessment, and the HSB colour space threshold adjustment (HCSTA) was tested in ImageJ software, and the results were compared. In determining the number of pustules and measuring the areas of pustules on leaves as a result of HCSTA severity measurements, the Spearman correlation coefficient (Rs) respectively was determined as Rs: 0.77 and Rs: 0.72. Moreover, the pustule areas were measured smaller than the actual areas. In the measurements made with the ZymoMacro_Scanner_v2.1.1 macro, the default macro settings, the edited macro settings and the YR pustule area measurements of the edited macro on the other leaves were respectively determined as Rs: 0.69, Rs: 0.92 and Rs: 0.85. Moreover, the pustule areas were measured much larger than the actual areas. The conditional severity measurement methods for YR pustules showed that it was Significant for HSB adjustments but not sufficient for disease assessment. In the ZymoMacro_Scanner_v2.1.1 results, the edited macro results were more successful than the default macro and HCSTA. However, when the edited macro was tried on other images, it was seen that the success rate decreased. The prediction success of the YR detection model and the promising correlations obtained in disease assessment methods indicate that this study is suitable for further development. (Word Count: 11166)
Özet (Çeviri)
Özet çevirisi mevcut değil.
Benzer Tezler
- Yüz ifadesi tanıma başarımı için ilgili bölge belirleme ve çekişmeli üretici ağlar kullanılarak örnek artırım yöntemlerinin geliştirilmesi
Development of a region of interest detection and a generative adversarial network based image augmentation approach for improving facial expression recognition performance
ÖMER FARUK SÖYLEMEZ
Doktora
Türkçe
2021
Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve KontrolFırat ÜniversitesiBilgisayar Mühendisliği Ana Bilim Dalı
PROF. DR. BURHAN ERGEN
- Uzamsal görüntü filtreleme ile yapay sinir ağı tabanlı retina damar bölütlenmesinin iyileştirilmesi
Improving artificial neural network based retinal vessel segmentation with spatial image filtering
CEM YAKUT
Yüksek Lisans
Türkçe
2021
Mühendislik BilimleriTrakya ÜniversitesiHesaplamalı Bilimler Ana Bilim Dalı
DR. ÖĞR. ÜYESİ SEZER ULUKAYA
DR. ÖĞR. ÜYESİ İLKAY ÖKSÜZ
- Fusion of dynamic and static features in signature verification
İmza doğrulamada dinamik ve statik özelliklerin birleştirilmesi
MUSTAFA SEMİH SADAK
Doktora
İngilizce
2022
Bilim ve TeknolojiYıldız Teknik ÜniversitesiElektronik ve Haberleşme Mühendisliği Ana Bilim Dalı
DOÇ. DR. NİHAN KAHRAMAN
DR. UMUT ULUDAĞ
- An approach for the chest tube detection in chest radiographs using convolutional neural network
Katmanlı sinir ağları kullanılarak göğüs radyografilerinde göğüs tüpü tespiti için bir yaklaşım
CEM AHMET MERCAN
Doktora
İngilizce
2015
Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve Kontrolİstanbul Teknik ÜniversitesiHesaplamalı Bilimler ve Mühendislik Ana Bilim Dalı
PROF. DR. MUSTAFA SERDAR ÇELEBİ
- Developing a deep learning approach for detecting brain cancer from mri images
Mrg görüntülerinden beyin kanserini saptamada derin öğrenme yaklaşimi geliştirmek
AHMAD TANASH
Yüksek Lisans
İngilizce
2020
Elektrik ve Elektronik MühendisliğiBahçeşehir ÜniversitesiBilgisayar Bilimleri Ana Bilim Dalı
DR. ÖĞR. ÜYESİ SERKAN AYVAZ