Automatic Mango Variety Classification and Grading using Deep learnig

The mango is known as a flavorful fruit and king of the fruit all over the world due to its nutritional qualities. Consumption of mangoes had increased in the world by increased population, and also usage of mango as an ingredient in cooked food, jellies/jams, drinks, and salad. At the result, the d

Project Title

Automatic Mango Variety Classification and Grading using Deep learnig

Project Area of Specialization

Computer Science

Project Summary

The mango is known as a flavorful fruit and king of the fruit all over the world due to its nutritional qualities. Consumption of mangoes had increased in the world by increased population, and also usage of mango as an ingredient in cooked food, jellies/jams, drinks, and salad. At the result, the demand for mangoes is also increased and expected to continue increasing, especially in markets such as the United States, Canada, the European Union, China and other Asian markets.  In order to compliment this fatly increasing demand for mango, there is a need to replace traditional method of manually classifying the various varieties of mangos and grading by adopting automation technologies in Pakistan. Manual classification and grading is inefficient and tiresome task with high risk of human error; in result farmer is fail to satisfy the criteria of export quality of mango Automatic grading of mangoes has been done with the different methods including fuzzy logic and neural network. In this paper we present an approach to automatic classification and grading of mangoes based on deep learning techniques convolutional neural network (CNN). so that the good quality mangoes can reach the market by the mango traders on time, and to enable farmer to gain maximum profit from mango  crop and  improving the agriculture economy of country through exporting large amount of mango.

Project Objectives

The Proposed study will achieve following objectives:

• To automatically predict mango yield in a specific mango orchard.
• To automatically classify mango varieties.
• To automatically grading mango quality.

 To strengthen export and agriculture economy and achieving appropriate profit from mango crop in Pakistan. To enhance the export of mango in Pakistan we propose an automatic system for mangoes detection, classification, grading, yield estimation to enable farmer taking informed and monitor their orchards in real-time about the total amount of mango and what variety of mango in what quantity he can export at what time.

Project Implementation Method

There have been several techniques in the literature for classifying various fruits such as mangoes, date fruits, olives, and grapes. Classification and grading of mangoes are done on the base of quality features. To find regions of interest classifying and learning approaches are used at several stages of the process. Classifiers are comprised of segmentation criteria based on simple image processing techniques or of some complex approaches associating with machine learning methods such as neural networks. Classifiers may integrate mango fruit’s features such as color, shape or texture in order to perform classification of various mango varieties. Classification is usually referred to as a supervised learning approach and these approaches are mainly used to map data into concerned cluster or groups. Classification process involved on two steps, in the first step a classifier model is developed for detailed description of pre-determined set of groups. This step of classification referred to as Training step or learning phase of model. In this step classifier are developed by algorithms through learning from labeled data. In the second step, the classifier that was generated in learning phase, is applied on testing data set in order to classifying different mangoes verities into specific class, the validity predicted results also checked after classification.

Benefits of the Project

Manual classification and grading of the various variety of mango is labor intensive, time consuming, costly, inefficient for farmers, and there is a high chance of error. As a result farmer fails to satisfy the criteria of providing export quality of mango.

All above mentioned factors are hurdles for farmer to strengthen export and agriculture economy and achieving appropriate profit from mango crop in Pakistan. To enhance the export of mango in Pakistan we propose an automatic system for mangoes detection, classification, grading, yield estimation to enable farmer taking informed and monitor their orchards in real-time about the total amount of mango and what variety of mango in what quantity he can export at what time.

                       

Technical Details of Final Deliverable

For classification data set will be created using the digital camera by fixed horizontally and mounted on the stand. For each variety  of mango 100 image will be captured with different angle.  Total 700 Images will be captured for classification. The model will be trained on 400 randomly selected images and 200 images will be used for testing the model. The most famous eight varieties of mango 1. Chaunsa (S.B), 2. Chaunsa (Black), 3. Chaunsa (White), 4. Dashehari, 5. Langra,  6. Sindhri,  7. Anwar Ratool, and 8. Fajri. grown in Multan Punjab will be selected for classification.. Classification comprises of two steps, the first of classification generate a classifier model to describe predefined set of classes, actually this step referred as training step,  classification algorithm develop a classifier through learning from data of their predefined class labels. In the second step of classification, the developed model  in first step will be used for classify various objects on the base of  features extracted from the images. There are several classification models including, artificial neural network (ANN), and principle component analysis (PCA) that will be used for classification support vector machine (SVM). SVM is powerful tool in order to perform nonlinear regression and non-linear classification, as it based on the statistical theory. An advantage of SVM over other algorithms is that it does not require large number of datasets to train model, also presence of outlier cannot affected it .

Final Deliverable of the Project

HW/SW integrated system

Core Industry

Food

Other Industries

IT , Agriculture

Core Technology

Artificial Intelligence(AI)

Other Technologies

Internet of Things (IoT)

Sustainable Development Goals

Zero Hunger, Industry, Innovation and Infrastructure

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com