INTELLIGENT COCOA FRUIT DISEASES DETECTION SYSTEM BASED ON CONVOLUTIONAL NEURAL NETWORK (CNN)

ABSTRACTNowadays sustenance crop reduction is the major problem in nutriment security because of different cocoa diseases. Early identification and accurate diagnosis of the health status of the cocoa is a critical job to limit the spread of different cocoa diseases and it should be in a technological manner rather than … See more

ABSTRACTNowadays sustenance crop reduction is the major problem in nutriment security because of different cocoa diseases. Early identification and accurate diagnosis of the health status of the cocoa is a critical job to limit the spread of different cocoa diseases and it should be in a technological manner rather than by the labor force. Traditional observation methods by farmers or domain area experts perhaps time-consuming, expensive and sometimes inaccurate. Deep learning approaches are the most accurate models for the detection of cocoa disease. Convolutional Neural network (CNN) is one of the popular approaches that allows computational models that are composed of multiple processing layers to learn representations of image data with multiple levels of abstraction. These models have dramatically improved the state-of-the-art in visual object recognition and image classification that makes it a good way for cocoa disease classification problems. For this purpose, we can appropriate CNN based model for identifying and classifying the most critical diseases of cocoa fruit such as cocoa bacterial wilt disease, cocoa Leaf spot disease, and Root mealybug disease of the cocoa. A total of 379 images are used for conducting experiments including augmented images with four different categories; diseased and a healthy class obtained from the different agricultural sectors stationed at Kantrol 2 Bono region Sunyani Ghana, these images are provided as input to the proposed model. Under the 10-fold cross-validation strategy, the experimental results show that the proposed model can effectively detect and classify classes of cocoa diseases condition with the best classification accuracy of 99.53% which is higher than compared to other classical deep learning models such as MobileNet and Inception v3 deep learning models.Keywords: Convolution neural networks; Cocoa disease detection; Detection of cocoa diseases, bacterial wilt; leafspot; root mealybug.TABLE OF CONTENT DECLARATIONDEDICATIONACKNOWLEDGMENTSTABLE OF CONTENTLIST OF FIGURESLIST OF TABLESABBREVIATIONS AND ACRONYMSABSTRACTCHAPTER ONEINTRODUCTION:1.1 Background1.2 Statement of the problem1.3 Objective of the research1.4 Scope and Limitations of the research1.5 Significance of the research1.6 Organization of the thesisCHAPTER TWOLITERATURE REVIEW2.1 The History of Cocoa and its production in Ghana2.2 Introduction of cocoa to Ghana2.3 World cocoa Production2.4 Cocoa production in Ghana2.5 Cocoa Flowers2.6 Varieties of Cocoa Trees2.7 Establishment of the cocoa farm2.8 Cocoa Farm Maintenance and Crop Husbandry2.9 Best Known Practices of Cocoa Cultivation2.10 West African Cocoa Farming2.11 Packaging and storage2.12 Human welfare, health and safety of cocoa producers2.13 Farm record keeping2.14 Cocoa Economics2.15 Products and Uses of Cocoa2.16 Cacao’s Fruit Pest and Disease Identification2.17 Deep Learning Implementation2.18 Convolutional Neural Network2.18.1Activation Functions2.19 Related Works2.20 SummaryCHAPTER THREERESEARCH METHODOLOGIES3.1 Dataset collection and preparation3.2 Data pre-processing3.3 Data augmentation3.4 Dataset Partitioning3.5 Tools used to implement and test the proposed model3.6 Evaluation3.7 Model Design and Experiment3.7.1. Experimental Setup3.7.2 Hyper Parameters configuration3.7.3 Input Layer Setup3.7.4 Feature Extraction Layers Setup3.7.5 Convolution layer Setup3.7.6 Pooling layer Setup3.7.7 Flattening Layer Setup3.7.8 Classification Layers Setup3.7.9 Fully Connected Layer Setup3.8 Training of Proposed Model3.9 Experiments with other pre-trained Models3.10 SummaryCHAPTER FOURRESULTS AND DISCUSSIONS4.1 Experimental results4.2 Experimental Results of the proposed Model4.3 Experiment 1: Training using 10-fold Cross-validation strategy4.4 Experiment 2: Training using train test split strategy4.5 Experiment 3: Training using train test split strategy4.6 Experiment 4: Training using train test split strategy4.7 Comparison with other Pre-Trained Models4.8 Experiment 1: Experimental results of InceptionV3 Model4.9 Experiment 2: Experimental results of MobileNet Model4.10 Results and Evaluation of MobileNet Model4.11 SummaryCHAPTER FIVECONCLUSION, CONTRIBUTION, AND FUTURE WORK5.1 Conclusion5.2 Contributions5.3 Future workSYSTEM SCREENSHOTREFERENCEAPPENDIX A: INPLIMENTATION OF PROPOSED MODEL

See less