عنوان پایان‌نامه
بررسی و تشخیص چند بیماری مهم گیاهی با استفاده از تکنیک آنالیز تصویر
- رشته تحصیلی
- مهندسی کشاورزی - مکانیک ماشینهای کشاورزی
- مقطع تحصیلی
- کارشناسی ارشد
- محل دفاع
- کتابخانه مرکزی پردیس کشاورزی و منابع طبیعی شماره ثبت: 5511;کتابخانه مرکزی -تالار اطلاع رسانی شماره ثبت: 59522
- تاریخ دفاع
- ۲۰ شهریور ۱۳۹۲
- دانشجو
- الهام عمرانی
- استاد راهنما
- سیدسعید محتسبی
- چکیده
- بیماریهای گیاهی میتوانند باعث کاهش کیفیت و کمیت محصولات کشاورزی شوند. معمولاً کارشناسان گیاهپزشکی بیماریهای گیاهی را مستقیماً با چشم تشخیص میدهند که این کار پرهزینه بوده و به زمان زیادی نیاز دارد. تشخیص خودکار بیماریهای گیاهی پژوهشی ضروری است که میتواند در نظارت بر مزارع و باغات بزرگ کاربرد زیادی داشته باشد. در بعضی از کشورهای در حال توسعه، کشاورزان زمان قابل توجهی برای مشاوره با گیاه‌پزشکان صرف می‌کنند، در حالیکه زمان عامل مهم در کنترل بیماری می‌باشد؛ بنابراین ارائه روشی آسان، سریع، ارزان و دقیق برای تشخیص بیماریهای گیاهی لازم به نظر میرسد. در این تحقیق، با استفاده از روش پردازش تصویر، سه اختلال درخت سیب (بیماریهای لکه سیاه سیب، آلترناریا و آفت مینوز) تشخیص داده می‌شوند. پس از جمعآوری برگهای بیمار و انتقال آن‌ها به آزمایشگاه، تصاویر برگها تحت شرایط نور کنترل شده تهیه شده و سپس به کمک الگوریتم طراحی شده در نرمافزار MATLAB، ابتدا نواحی بیماری روی برگها با استفاده از دو روش خوشه‌بندی k- میانگین کلاسیک و c- میانگین فازی تشخیص و جداسازی شدند و سپس ویژگی‌های مربوط به رنگ و بافت تصویر نواحی بیماری استخراج شدند. در ادامه چهار مدل توسعه داده شد که مدل اول شامل ویژگی‌های حاصل از ماتریس هم‌رویدادی، مدل دوم شامل ویژگی‌های رنگی، مدل سوم شامل ویژگی‌های استخراج شده از تبدیل‌های موجک و فوریه و مدل چهارم شامل همه‌ی ویژگی‌ها بود و سپس با استفاده از شبکه عصبی مصنوعی(ANN ) و سیستم استنتاج عصبی-فازی تطبیقی (ANFIS) بیماریهای گیاهی طبقه‌بندی شدند. نتایج نشان داد که شبکه عصبی مصنوعی به طور موفقیت‌آمیزی توانست لکه‌های بیماری مشخص شده با دو روش خوشه‌بند k- میانگین کلاسیک و خوشه‌بند c- میانگین فازی را با دقت 100 % طبقه‌بندی کند و سیستم استنتاج عصبی-فازی برگهای بیمار مشخص شده با خوشه‌بند k- میانگین کلاسیک و خوشه‌بند c- میانگین فازی را به ترتیب با دقت‌های 100و 3386/84 % طبقه‌بندی کرد.
- Abstract
- Plant diseases cause significant reduction in quality and quantity of agricultural products. Plant pathologists usually detect the diseases directly by their eye, however, this action requires continuous monitoring of the pathologists which might be time consuming and also very expensive in large farms. Moreover, in some developing countries, farmers may have to spend considerable time to see the experts, whereas time is an important factor in controlling the diseases. Therefore; looking for a fast, automatic, inexpensive and accurate method to detect plant disease is important and it can provide significant benefits in monitoring large fields of crops, and thus it can provide detecting the symptoms of diseases automatically as soon as disease symptoms appear on plant leaves. In this study, three different apple diseases that appear on leaves (Alternaria, VentoriaInaequalis and Phyllonrycterturanica disasters) were chosen to be investigated via image processing technique. After sampling, infected leaves were transferred to the laboratory, then the images of leaves were captured under controlled light, and after that the algorithms were designed in MATLAB. For diagnosis, at first, disease region on leaves with using K-means clustering and fuzzy c-mean clustering were detected and then color and texture features were extracted. Four models were developed, which include co-occurrence matrix features, color features, extracted features of wavelet transform and Fourier transform, and the fourth model includes all the features. Artificial Neural network (ANN) and Adaptive Neuro-Fuzzy Inference System (ANFIS) were used for diseases classification.The results showed that the ANN was able to successfully Segment leaves with classical k-means clustering method and fuzzy c-mean clustering methodwith 100% accuracyand ANFISclassified leaves with k-mean clustering and fuzzy c-mean clustering method with 100 and 84.3386% respectively. Plant diseases cause significant reduction in quality and quantity of agricultural products. Plant pathologists usually detect the diseases directly by their eye, however, this action requires continuous monitoring of the pathologists which might be time consuming and also very expensive in large farms. Moreover, in some developing countries, farmers may have to spend considerable time to see the experts, whereas time is an important factor in controlling the diseases. Therefore; looking for a fast, automatic, inexpensive and accurate method to detect plant disease is important and it can provide significant benefits in monitoring large fields of crops, and thus it can provide detecting the symptoms of diseases automatically as soon as disease symptoms appear on plant leaves. In this study, three different apple diseases that appear on leaves (Alternaria, VentoriaInaequalis and Phyllonrycterturanica disasters) were chosen to be investigated via image processing technique. After sampling, infected leaves were transferred to the laboratory, then the images of leaves were captured under controlled light, and after that the algorithms were designed in MATLAB. For diagnosis, at first, disease region on leaves with using K-means clustering and fuzzy c-mean clustering were detected and then color and texture features were extracted. Four models were developed, which include co-occurrence matrix features, color features, extracted features of wavelet transform and Fourier transform, and the fourth model includes all the features. Artificial Neural network (ANN) and Adaptive Neuro-Fuzzy Inference System (ANFIS) were used for diseases classification.The results showed that the ANN was able to successfully Segment leaves with classical k-means clustering method and fuzzy c-mean clustering methodwith 100% accuracyand ANFISclassified leaves with k-mean clustering and fuzzy c-mean clustering method with 100 and 84.3386% respectively.
