Estimation of Oil and Carotenoid Quantities in Maize Grain by Image Processing Technique

Fatih Kahrıman; Cem Ömer Egesel

Makale detayları

Estimation of Oil and Carotenoid Quantities in Maize Grain by Image Processing Technique

Yazarlar:

Fatih Kahrıman | Cem Ömer Egesel

Yayıncı:

Çanakkale Onsekiz Mart University

Ana Başlıklar:

Tarımsal Biyoteknoloji (Diğer)

Dergi:

BİLİM-TEKNOLOJİ-YENİLİK EKOSİSTEMİ DERGİSİ Vol.5 Issue 2, pp.91-104

Belge Tipi:

Makale

Yayın Tarihi:

12/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001348309

pISSN:

2757-6140

Anahtar kelimeler:

Özet:

 Chemical methods used for the analysis of samples in breeding studies for oil and carotenoid content, components of maize seeds, are laborious and expensive procedures. Therefore, there is a need for different techniques that can be used for sample screening and characterization. Image processing techniques, which have become widespread in almost every field of agriculture, offer important opportunities in this context. This study was conducted to examine the success of estimating oil and carotenoid contents in maize seeds using image processing-based approaches. Digital images were taken from fifteen different genotypes in 3 replicates and oil and carotenoid contents were determined using 10 replicates in the same samples using reference methods. Color value and embryo arearatio were extracted from the digital images with the algorithm designed specifically for this study. Data on HSV, Lab, XYZ spaces were also obtained from RGB space, the basic color space in the images. The results of the reference analysis were assigned as dependent variables for oil and carotenoid ratios separately and prediction models were developed using random forest technique with other variables as predictors. The variation of samples in terms of oil and carotenoids was evaluated by one-way ANOVA. The R2 value of the prediction model for oil ratio was 0.8975 and that of carotenoid was 0.9617. It was seen that other variables except color value could be used for prediction purposes in the models created. The results showed that image processing methods can be used for screening and characterization of oil and carotenoid content of maize samples.

 Chemical methods used for the analysis of samples in breeding studies for oil and carotenoid content, components of maize seeds, are laborious and expensive procedures. Therefore, there is a need for different techniques that can be used for sample screening and characterization. Image processing techniques, which have become widespread in almost every field of agriculture, offer important opportunities in this context. This study was conducted to examine the success of estimating oil and carotenoid contents in maize seeds using image processing-based approaches. Digital images were taken from fifteen different genotypes in 3 replicates and oil and carotenoid contents were determined using 10 replicates in the same samples using reference methods. Color value and embryo arearatio were extracted from the digital images with the algorithm designed specifically for this study. Data on HSV, Lab, XYZ spaces were also obtained from RGB space, the basic color space in the images. The results of the reference analysis were assigned as dependent variables for oil and carotenoid ratios separately and prediction models were developed using random forest technique with other variables as predictors. The variation of samples in terms of oil and carotenoids was evaluated by one-way ANOVA. The R2 value of the prediction model for oil ratio was 0.8975 and that of carotenoid was 0.9617. It was seen that other variables except color value could be used for prediction purposes in the models created. The results showed that image processing methods can be used for screening and characterization of oil and carotenoid content of maize samples.

 Chemical methods used for the analysis of samples in breeding studies for oil and carotenoid content, components of maize seeds, are laborious and expensive procedures. Therefore, there is a need for different techniques that can be used for sample screening and characterization. Image processing techniques, which have become widespread in almost every field of agriculture, offer important opportunities in this context. This study was conducted to examine the success of estimating oil and carotenoid contents in maize seeds using image processing-based approaches. Digital images were taken from fifteen different genotypes in 3 replicates and oil and carotenoid contents were determined using 10 replicates in the same samples using reference methods. Color value and embryo arearatio were extracted from the digital images with the algorithm designed specifically for this study. Data on HSV, Lab, XYZ spaces were also obtained from RGB space, the basic color space in the images. The results of the reference analysis were assigned as dependent variables for oil and carotenoid ratios separately and prediction models were developed using random forest technique with other variables as predictors. The variation of samples in terms of oil and carotenoids was evaluated by one-way ANOVA. The R2 value of the prediction model for oil ratio was 0.8975 and that of carotenoid was 0.9617. It was seen that other variables except color value could be used for prediction purposes in the models created. The results showed that image processing methods can be used for screening and characterization of oil and carotenoid content of maize samples.

 Mısır tohumlarının bileşenlerinden olan yağ ve karotenoid içeriğine yönelik ıslah çalışmalarında örneklerin analizi için kullanılan kimyasal yöntemler zahmetli ve pahalı işlemlerdir. Bu nedenle örnek tarama ve karakterizasyon amacıyla kullanılabilecek farklı tekniklere ihtiyaç duyulmaktadır. Tarımın hemen her alanında yaygınlaşmaya başlayan görüntü işleme teknikleri bu bağlamda önemli fırsatlar sunmaktadır. Bu çalışma, mısır tohumlarında yağ ve karotenoid içeriklerinin görüntü işlemeye dayalı yaklaşımlar kullanılarak tahminlenmesindeki başarı durumunu incelemek amacıyla yürütülmüştür. On beş farklı genotipten 3 tekrarlamalı dijital görüntüler alınmış, aynı örneklerde referans metotlarla yağ ve karotenoid içerikleri 10 tekrarlamalı olarak belirlenmiştir. Dijital görüntülerden bu çalışmaya özel olarak tasarlanan algoritma ile renk değeri ve embriyo alanıoranına ilişkin sonuçlar çıkarılmıştır. Alınan dijital görüntülerdeki temel renk uzayı olan RGB uzayından HSV, Lab, XYZ uzaylarına ait veriler de elde edilmiştir. Referans analiz sonuçları yağ ve karotenoid için ayrı ayrı bağımlı değişken olarak atanmış ve diğer değişkenler tahminleyici olarak kullanılmak suretiyle rassal orman tekniğinden yararlanılarak tahmin modelleri geliştirilmiştir. Örneklerin yağ ve karotenoid bakımından değişimi tek yönlü varyans analizi ile değerlendirilmiştir. Yağ oranına ilişkin tahmin modelinin R2 değeri 0.8975, karotenoidinki ise 0.9617 olarak tespit edilmiştir. Oluşturulan modellerde renk değeri hariç diğer değişkenlerin tahmin amaçlı kullanılabileceği görülmüştür. Sonuçlar, incelen görüntü işleme yöntemlerinin mısır örneklerinin yağ ve karotenoid içeriğinin taranmasında ve karakterizasyonunda kullanılabileceğini göstermiştir.

Makale detayları

Estimation of Oil and Carotenoid Quantities in Maize Grain by Image Processing Technique

Yazarlar:

Fatih Kahrıman | Cem Ömer Egesel

Yayıncı:

Çanakkale Onsekiz Mart University

Ana Başlıklar:

Tarımsal Biyoteknoloji (Diğer)

Dergi:

BİLİM-TEKNOLOJİ-YENİLİK EKOSİSTEMİ DERGİSİ Vol.5 Issue 2, pp.91-104

Belge Tipi:

Makale

Yayın Tarihi:

12/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001348309

pISSN:

2757-6140

Regression model

color space

random forestRegresyon modeli

renk uzayı

rassal orman

Sayı

Makale detayları

Estimation of Oil and Carotenoid Quantities in Maize Grain by Image Processing Technique

Yazarlar: Fatih Kahrıman | Cem Ömer Egesel

Yayıncı: Çanakkale Onsekiz Mart University

Ana Başlıklar: Tarımsal Biyoteknoloji (Diğer)

Dergi: BİLİM-TEKNOLOJİ-YENİLİK EKOSİSTEMİ DERGİSİ Vol.5 Issue 2, pp.91-104

Belge Tipi: Makale

Yayın Tarihi: 12/2024

Belge Dili: Türkçe

Document ID (Org No.): 2001348309

pISSN: 2757-6140

Regression model

color space

random forestRegresyon modeli

renk uzayı

rassal orman

Sayı

Yazarlar:

Fatih Kahrıman | Cem Ömer Egesel

Yayıncı:

Çanakkale Onsekiz Mart University

Ana Başlıklar:

Tarımsal Biyoteknoloji (Diğer)

Dergi:

BİLİM-TEKNOLOJİ-YENİLİK EKOSİSTEMİ DERGİSİ Vol.5 Issue 2, pp.91-104

Belge Tipi:

Makale

Yayın Tarihi:

12/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001348309

pISSN:

2757-6140