Melastoma Malabathricum L. Extracts-Based Indicator for Monitoring Shrimp Freshness Integrated With Classification Technology Using Nearest Neighbours Algorithm

Abstract

As a maritime country, shrimp commodity production in Indonesia is very high and continues to increase. However, because shrimp is a perishable food, we need a detection device. This is because conventional methods that are widely used by the community in detecting freshness of shrimp are only based on the smell. Of course, this is a problem when shrimp are packed in closed containers. In this paper, a method for detecting shrimp is proposed using the Melastoma malabathricum L. - based label indicator. The high content of flavonoids in the extracts allows the changing the colour of the label from red to grey due to the interaction between the label with the OH- group that arises from the shrimp spoilage process. The colour that appears on the label indicator will correlate with the level of shrimp freshness. By increasing detection effectiveness, the classification is performed using the nearest-neighbours algorithm, which is equipped with an image processing mechanism in the form of colour quantization. There are four classifications used to express the quality of shrimp, namely "acceptable," "just acceptable," "unacceptable," and "more unacceptable." The accuracy of applying this method is 71.9%, with the majority of detection errors occurring in the "acceptable" class. Based on these results, it can be stated that the label indicators prepared in this study are very promising to be developed into intelligent packaging components.

Performance analysis of various types of surface crack detection based on image processing

“…Nearest neighbour is a flexible technique and the simplest machine learning algorithm because it can classify test data into label classes by searching for train data which is relatively the same as test data [15]. Furthermore, although it is a simple algorithm, it still provides good performance results [16]. The output performance was the accuracy result (%) of how the system can accurately classify the input images based on respected classes.…”