Article
Teknologi Computer Vision dan Chemometric untuk Deteksi dan Kontrol Mutu Produk Perikanan: Prinsip, Aplikasi, dan Tantangan: Computer Vision and Chemometric Technologies for Quality Detection and Control of Fishery Products: Principles, Applications, and Challenges
Jurnal Fish Protech, Vol. 9 No. 2 (2026), hlm. 267–280
Abstrak
Produk perikanan segar seperti ikan, udang, dan kekerangan merupakan komoditas pangan bernilai gizi tinggi yang rentan mengalami kemunduran mutu akibat aktivitas enzim endogen dan mikroorganisme selama penanganan, penyimpanan, dan distribusi dalam rantai pasok. Metode deteksi mutu konvensional, seperti quality index method (QIM) yang bersifat subjektif serta kromatografi gas dan spektrometri massa yang destruktif dan memerlukan waktu analisis lama, dinilai kurang sesuai untuk pengawasan mutu secara cepat dan berkelanjutan. Perkembangan teknologi spektroskopi, sensor sensorik elektronik, dan computer vision yang dipadukan dengan metode chemometric menawarkan alternatif deteksi non-destruktif yang objektif dan berpotensi diterapkan secara daring. Artikel ini disusun melalui pendekatan narrative literature review untuk mengkaji prinsip kerja, perkembangan, dan aplikasi teknologi near-infrared spectroscopy, hyperspectral imaging, Raman spectroscopy, spektroskopi fluoresensi, electronic nose, electronic tongue, serta machine/computer vision yang dipadukan dengan chemometric seperti principal component analysis (PCA), partial least squares regression (PLSR), dan artificial neural network (ANN) dalam evaluasi kesegaran produk perikanan. Hasil kajian menunjukkan integrasi computer vision dengan chemometric mampu memprediksi indikator mutu seperti nilai K, total volatile basic nitrogen (TVB-N), histamin, dan pH dengan akurasi tinggi (R² pada kisaran 0,80–0,99), sekaligus memungkinkan visualisasi spasial-temporal distribusi mutu pada sampel utuh, sebagaimana ditunjukkan pada studi pencitraan fluoresensi multidimensi udang beku. Namun, keterbatasan generalisasi model antarspesies, kebutuhan perangkat keras yang kompleks, serta kesenjangan performa antara pengujian laboratorium dan aplikasi industri masih menjadi tantangan utama. Riset ke depan diarahkan pada pengembangan sensor portabel, fusi data multisensor, dan pemanfaatan deep learning guna mendukung sistem deteksi mutu perikanan yang real-time dan dapat diterapkan secara komersial. ABSTRACT Fresh fishery products such as fish, shrimp, and shellfish are highly nutritious food commodities that are particularly susceptible to quality deterioration caused by endogenous enzymatic activity and microbial spoilage during handling, storage, and distribution along the supply chain. Conventional quality assessment methods, including the subjective quality index method (QIM) and destructive, time-consuming gas chromatography and mass spectrometry, are considered inadequate for rapid and continuous quality monitoring. Advances in spectroscopic technology, electronic sensory sensors, and computer vision combined with chemometric methods offer non-destructive, objective, and potentially online detection alternatives. This article was developed through a narrative literature review approach to examine the principles, developments, and applications of near-infrared spectroscopy, hyperspectral imaging, Raman spectroscopy, fluorescence spectroscopy, nuclear magnetic resonance, electronic nose, electronic tongue, and machine/computer vision integrated with chemometric tools such as principal component analysis (PCA), partial least squares regression (PLSR), and artificial neural networks (ANN) for evaluating fishery product freshness. The review shows that integrating computer vision with chemometrics enables high-accuracy prediction (R2 in the range of 0.80– 0.99) of quality indices such as K-value, total volatile basic nitrogen (TVB-N), histamine, and pH, while also enabling spatial-temporal visualization of quality distribution across whole samples, as demonstrated by multidimensional fluorescence imaging studies on frozen shrimp. Nevertheless, limited model generalizability across species, complex instrumentation requirements, and the performance gap between laboratory testing and industrial application remain major challenges. Future research should focus on portable sensor development, multi-sensor data fusion, and deep learning applications to support real-time, commercially applicable fishery quality detection systems.