Design of Smart System to Detect Ripeness of Tomato and Chili with New Approach in Data Acquisition

2nd Annual Applied Science and Engineering Conference

工业技术;自然科学

Taofik, A.^1 ; Ismail, N.^2 ; Gerhana, Y.A.^3 ; Komarujaman, K.^2 ; Ramdhani, M.A.^3

Agrotech Department, Faculty of Science and Technology, UIN Sunan Gunung Djati, Jalan A.H Nasution 105, Cibiru - Bandung

40614, Indonesia^1

Electrical Engineering Department, Faculty of Science and Technology, UIN Sunan Gunung Djati, Jalan A.H Nasution 105, Cibiru - Bandung

40614, Indonesia^2

Informatics Engineering Department, Faculty of Science and Technology, UIN Sunan Gunung Djati, Jalan A.H Nasution 105, Cibiru - Bandung

40614, Indonesia^3

关键词: Classification process;    Color characteristics;    Fruit ripeness;    Harvesting periods;    Integrated systems;    K-means clustering method;    Overall accuracies;    Testing systems;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/288/1/012018/pdf
DOI  :  10.1088/1757-899X/288/1/012018

来源: IOP

【 摘 要 】

Manual laxity of fruit ripeness classification is highly influenced by operator subjectivity, thus there is inconsistency for some periods in the classification process. Information Technology development allows fruit identification based on color characteristic by computer aids. A developed system was designed to work on a mobile device with the ability to detect four levels of ripeness of tomato and chili fruits. The acquisition of training data is done with a new approach. Training data came from objective observation of the same fruit of tomato and chili, captured since one month before harvesting until harvesting period. Image segmentation uses K-Means Clustering Method while ripeness detection uses fuzzy logic. The system output consists of types and level of ripeness grouped into four categories: unripe 1, unripe 2, medium, and ripe. This article explains preliminary results of the testing system in static and partial condition using a personal computer before being applied into a mobile-based integrated system. The results showed the level of success for fruit segmentation was 80% for tomato and 100% for chili. The fault is due to the similarity of fruit sample size. The level of success for detecting fruit ripeness is 80% for tomato and 90% for chili. By 10 training data of each, it is shown that the good result with an overall accuracy level of average ripeness detection is 85%.