Melon Ripeness Determination Using K-nearest Neighbor Algorithm

This paper presents a method for determining the ripeness of Cantaloupe using a K-Nearest Neighbors (KNN) Algorithm on a Raspberry PI. One of the most common problems is determining fruit ripeness purely by visual inspection and traditional methods, such as relying on touch, which is challenging to implement. The Color Segmentation Algorithm used in the study operates in the HSV color space. The Canny Edge detection technique utilizes a region-growing approach, region merging, and initial seed selection. Following the segmentation process, the ripeness of the Cantaloupe is determined using the K-Nearest Neighbors (KNN) Algorithm based on its features, where accuracy reports from the dataset determine the best value of K. The proposed Color Segmentation Algorithm successfully segments the captured Cantaloupe images without any errors and determines their ripeness in most cases based on the KNN Algorithm. However, there are instances where the KNN algorithm incorrectly predicts ripeness from uneven lighting and objects detected in the image, resulting in an accuracy of 80 percent. In general, the system's accuracy based on the Confusion Matrix testing dataset is 95 percent, and as for actual testing, it's 80 percent, as stated before.

I. Introduction

When a person decides to pick or buy a melon, choosing the ones that are not ripe is discouraging since most people favor melons that they can quickly consume and put on display to mature for a week or two [1]–[4]. Fruits in the gourd family, like watermelon and squash, tend to give out clues such as tendrils being directly opposite to the point where the gourd is attached to the vine or the vine becoming hard and woody and even with the texture or smell.[1] [5]. Many of us depend on feeling the melon by touch to see if it is ripe, but it can be discouraging because it often results in the melon not being ripe at all and further damaging the fruit [5]. Since melons like the Cantaloupes do not produce any tendrils like the other fruits in the gourd family, the other clue in indicating when the melon is ripe is by its visible features such as the color, the spots/dots caused by rotting, as well as the net development of the melon's skin [1] [6] [7]. Today's technology, including image processing, makes it possible to use the markers above to determine the ripeness of a melon in a way convenient for the user [8]–[10]. This study aims to use the color segmentation algorithm in image processing using a Raspberry Pi device to determine the ripeness of the cantaloupe fruit. The said device has proven to be capable of the same task as any computer today, which was proven by research done by [11], where a Raspberry Pi device was used to classify red blood cells according to their shapes. The procedure of color segmentation can be described as dividing a picture into useful areas depending on color properties [12]. Fruit with raw skin and ripe skin can have drastically differing skin tones since various color spaces can be used to determine whether a fruit is ripe or just beginning to ripen [13]. The fruit undergoes a variety of changes as it ripens, such as a change in peel color and a change in texture (net development of the skin), which is used to assess the fruit's health and measure its level of maturity without causing damage [14]. Using a technique called color segmentation, a colored image is divided into numerous color-based clusters, and with this, crop maturity can be assessed using a variety of color segmentation approaches, including edge detection, region growth, histograms, and machine learning methods like Support Vector Machine (SVM) and OpenCV and more [15]–[18].