OBJECTIVE:

Assuring food safety is a major concern, and it is hard to find anyone who has not encountered an unpleasant moment of food-borne illness at least once in their past years. Food-borne illnesses may result from the consumption of food contaminated by microbial pathogens, toxic chemicals or radioactive materials.

Degradation analysis of banana will surely pave the way of terminating food-borne illness and identify freshness or degradation of fruits and foods.

MOTIVATION:

The main purpose of our research project is to implement Machine Learning in the field of Agriculture and increase the accuracy rate of fruit grading system ensuring more effective fruit sorting process.

EXPECTED OUTCOME:

This research has a long term vision and goal to achieve most efficient way to detect not only freshness of banana but also assure quality of other types foods. Currently developed models aim to achieve following goals:

· Ability to detect banana’s degraded areas.

· Identifying age of the fruit in days.

· Detecting the degraded percentage.

· Classifying whether the fruit is fresh or degraded.

Proposed Methodology:

• Convolutional Neural Networks (CNN)

• CONTOUR Analysis

REQUIRED TECHNOLOGIES :

• Python

• Numpy

• Pandas

• Matplotlib

• Bokeh

• SciKit-Learn

• TensorFlow, keras

DEVELOPMENT PHASES :

Methodology 1: CNN (convolutional neural network)

A convolutional neural network (CNN) is a specific type of artificial neural network that uses perceptrons, a machine learning unit algorithm, for supervised learning, to analyze data.

In neural networks, Convolutional neural networkis one of the main categories to do images recognition, images classifications. Objects detections, recognition faces etc. we used CNN to identify the degradation banana

each input image is passed through a series of convolution layers with filters (Kernals), Pooling, fully connected layers (FC) and apply Sigmoidx/ReLu function to classify the freshness with probabilistic values between 0 and 1.

Methodology 2: CONTOUR

We first import all useful libraries and load images to make them convert. After that, we convert the frame from BGR to HSV format so that later we can create a mask. We have defined HSV ranges with lower and upper of a specific color, in this contour model to detect degrade areas we chose gray. After that to find the contours we have used the opencv built in function findContours. Contours contain an array with the coordinates of all the contours of the object.

MORPHE_ELLIPSE:

An elliptic structuring element, where a filled ellipse inscribed into the defined rectangle size Rect(0, 0, esize.width, esize.height). So therefore we have applied another helpful structuring element from active contour model cv2.getStructuringElement(cv2.MORPH_ELLIPSE,(9,9)) to calculate morphological gradients.

ALGORITHMS

1. PCA (Principal Component Analysis)

2. RF (Random Forests)

RESULT:

The trained CNN model has been able to classify freshness of banana with around 99% accuracy. and COUNTER model has around 98.75% accuracy which are very good for the seven days test and training sets on 300 chosen banana fruits of different color, shape and size.

FUTURE WORK:

• Develop the model to identify different types of fruit freshness

• Develope commercial automated System for fruit sorting and classification

• To Enhance accuracy and performance.

CHALLENGES:

• Controlled environment for capturing data images.

• Humidity and temperature.

• Shortage of data.

• Budget crisis.