Pengembangan Sistem Pencitra Berbasis Smartphone dan Machine Learning untuk Deteksi Kesegaran serta Prediksi Sisa Masa Simpan Daging Sapi Menggunakan Sensor Halokromik Antosianin-Nanoselulosa

Rabbani, Muhammad Amir Ma'ruf (2026) Pengembangan Sistem Pencitra Berbasis Smartphone dan Machine Learning untuk Deteksi Kesegaran serta Prediksi Sisa Masa Simpan Daging Sapi Menggunakan Sensor Halokromik Antosianin-Nanoselulosa. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5009221023-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (7MB) | Request a copy

Abstract

Penilaian kesegaran daging sapi secara visual masih bersifat subjektif dan kurang mampu memberikan informasi kuantitatif mengenai tingkat kesegaran maupun sisa masa simpan. Penelitian ini bertujuan mengembangkan sistem pencitra berbasis smartphone untuk akuisisi citra sensor halokromik antosianin-nanoselulosa, serta mengintegrasikannya dengan machine learning untuk mendeteksi tingkat kesegaran dan memprediksi sisa masa simpan daging sapi sirloin dan tenderloin. Sensor halokromik dibuat dari nanoselulosa, gliserin, dan antosianin, kemudian diuji responsnya terhadap uap amonia sebagai senyawa model untuk merepresentasikan senyawa basa volatil yang terbentuk selama pembusukan. Sistem pencitra dirancang menggunakan smartphone, LED light, macrolens, dan mounting bracket agar proses akuisisi citra lebih terkendali. Citra sensor diambil selama 72 jam pada suhu ruang dengan interval 1 jam, kemudian dikalibrasi menggunakan ColorChecker, disegmentasi untuk menetapkan region of interest (ROI), dan diekstraksi menjadi fitur warna RGB, HSV, dan CIE Lab. Nilai TVB-N digunakan sebagai acuan kimia untuk menentukan tingkat kesegaran daging. Hasil penelitian menunjukkan bahwa sistem pencitra yang dikembangkan mampu menghasilkan citra sensor yang dapat digunakan sebagai masukan model machine learning. Sensor antosianin-nanoselulosa juga menunjukkan respons warna terhadap amonia, sedangkan nilai TVB-N meningkat selama penyimpanan dari 21,67 menjadi 389,28 mg N/100 g pada sirloin dan dari 21,42 menjadi 360,80 mg N/100 g pada tenderloin. Pada klasifikasi biner, data 72 jam memberikan performa terbaik dengan Gradient Boosting sebagai model tertinggi dan Random Forest sebagai model paling stabil. Pada klasifikasi multi-kelas, Extra Trees menjadi model paling representatif, sedangkan pada regresi sisa masa simpan, Extra Trees dan KNN-R menunjukkan performa kuat dengan nilai R² hingga 0,95. Analisis SHAP menunjukkan bahwa fitur G, H, S, R, L, dan a berkontribusi dominan terhadap prediksi. Dengan demikian, sistem pencitra berbasis smartphone yang terintegrasi dengan machine learning menunjukkan potensi sebagai pendekatan objektif untuk mendeteksi kesegaran dan memprediksi sisa masa simpan daging sapi dalam kondisi pengujian yang digunakan pada penelitian iniuntuk deteksi kesegaran serta prediksi sisa masa simpan daging sapi berbasis sensor halokromik.
==================================================================================================================================
Visual assessment of beef freshness remains subjective and provides limited quantitative information regarding freshness level and remaining shelf life. This study aimed to develop a smartphone-based imaging system for acquiring images of an anthocyanin nanocellulose halochromic sensor and to integrate the system with machine learning for detecting freshness levels and predicting the remaining shelf life of beef sirloin and tenderloin. The halochromic sensor was fabricated from nanocellulose, glycerol, and anthocyanin, and its response to ammonia vapor was evaluated using ammonia as a model compound representing the volatile basic compounds generated during spoilage. The imaging system consisted of a smartphone, LED illumination, a macro lens, and a mounting bracket to ensure controlled and consistent image acquisition. Sensor images were captured over 72 h at room temperature at 1 h intervals. The images were subsequently calibrated using a ColorChecker, segmented to define the region of interest (ROI), and processed to extract RGB, HSV, and CIE L*a*b* color features. Total volatile basic nitrogen (TVB-N) values were used as a chemical reference for determining beef freshness levels. The results demonstrated that the developed imaging system produced sensor images suitable as inputs for machine-learning models. The anthocyanin nanocellulose sensor exhibited a distinct color response to ammonia, while TVB-N values increased during storage from 21.67 to 389.28 mg N/100 g in sirloin and from 21.42 to 360.80 mg N/100 g in tenderloin. For binary classification, the 72 h dataset yielded the best overall performance, with Gradient Boosting achieving the highest predictive performance and Random Forest demonstrating the greatest stability. For multiclass classification, Extra Trees was identified as the most representative model. In remaining shelf-life regression, Extra Trees and KNN regression exhibited strong predictive performance, with R² values of up to 0.95. SHAP analysis revealed that the G, H, S, R, L*, and a* features contributed most strongly to the model predictions. Overall, the integration of smartphone-based imaging, machine learning, and a halochromic sensor shows considerable potential as an objective approach for detecting beef freshness and predicting its remaining shelf life under the experimental conditions applied in this study.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Smart Packaging, Sensor Halokromik, Antosianin-Nanoselulosa, Machine Learning, Kesegaran Daging Sapi, Sustainable Development Goals (SDGs)
Subjects:	T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T57.8 Nonlinear programming. Support vector machine. Wavelets. Hidden Markov models. T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques. Image analysis--Data processing. T Technology > T Technology (General) > TA404 Materials--Biodegradation
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Physics Engineering > 30201-(S1) Undergraduate Thesis
Depositing User:	Muhammad Amir Ma'ruf Rabbani
Date Deposited:	04 Aug 2026 01:09
Last Modified:	04 Aug 2026 01:09
URI:	http://repository.its.ac.id/id/eprint/142631

Actions (login required)

View Item