Revista Cienﬁca, FCV-LUZ / Vol. XXXV Recibido: 16/05/2025 Aceptado: 04/07/2025 Publicado: 23/07/2025 hps://doi.org/10.52973/rcfcv-e35650 UNIVERSIDAD DEL ZULIA Serbiluz Sistema de Servicios Bibliotecarios y de Información Biblioteca Digital Repositorio Académico 1 of 7 Revista Cienﬁca, FCV-LUZ / Vol. XXXV hps://doi.org/10.52973/rcfcv-e35702 UNIVERSIDAD DEL ZULIA Serbiluz Sistema de Servicios Bibliotecarios y de Información Biblioteca Digital Repositorio Académico Eﬀects of chilled storage on ﬁsh freshness using computer vision and arﬁcial neural network modeling Efectos del almacenamiento refrigerado en la frescura del pescado ulizando visión por computadora y modelado bajo red neuronal arﬁcial Saliha Lakehal 1 * , Brahim Lakehal 2 ¹ESPA laboratory, University of Batna 1, Instute of Veterinary Science and Agricultural Sciences, Department of Veterinary Sciences. Batna, Algeria. ²University of Batna 2, Instute of Hygiene and Industrial Security. Batna, Algeria. *Corresponding author: saliha.lakehal@univ–batna.dz ABSTRACT The present study invesgates the impact of refrigeraon storage on the freshness and shelf life of European sea bass (Dicentrarchus labrax). This invesgaon ulises computer vision systems and arﬁcial neural networks (ANNs) to analyse the dynamics of the process. A non-destrucve assessment approach was established by analysing the eye colour characteriscs (RGB, Lab*, and HSI values) of ﬁsh stored at +4 °C for 15 days, with sampling occurring every three days. There were considerable changes in the colour range throughout the time, particularly a reduction of brightness (L*), which can be one indicator of the progressive deterioration of the fish›s freshness. The neural network multilayer perceptron was opmised with 20 neurons in the hidden layer and demonstrated a high correlaon coeﬃcient (R² = 0.98) between predicted and experimental shelf life values. The data indicates that the values of rack life, which were inially determined to be cauous, exhibited a high degree of correlaon with the esmated values. The R 2 value was determined to be 0.98. The technique oﬀers a rapid and reliable non-destrucve method for determining the freshness of ﬁsh, with potenal applicaons in relevant areas such as quality control and natural security examinaon for aquaculture products. Key words: color analysis; ﬁsh meat; eye color; quality control; storage me. RESUMEN Este estudio examina el efecto del almacenamiento refrigerado en la frescura y vida úl de la lubina europea (Dicentrarchus ladrax) mediante sistemas de visión arﬁcial y redes neuronales arﬁciales (RNA). Se estableció un enfoque de evaluación no destrucva mediante el análisis de las caracteríscas del color de los ojos (valores RGB, Lab* y HSI) del pescado almacenado a +4 °C durante 15 días, con muestreos cada tres días. Se observaron cambios considerables en la gama de colores a lo largo del empo, en parcular una reducción del brillo (L*), que puede ser un indicador del deterioro progresivo de la frescura del pescado. Se entrenó un perceptrón mulcapa de red neuronal opmizado con 20 neuronas en la capa oculta, con un alto coeﬁciente de correlación (R² = 0,98) entre los valores predichos y experimentales de vida úl. Los valores de vida úl temporalmente prudentes presentaron una alta correlación con los valores esmados (R² = 0,98). Esta técnica ofrece una técnica no destrucva rápida y ﬁable para la determinación de la frescura del pescado, con potencial aplicación en áreas relevantes como el control de calidad y la evaluación de la seguridad natural de productos acuícolas. Palabras clave: Análisis de color; carne de pescado; color del ojo; control de calidad; empo de almacenamiento.

Eﬀects of chilled storage on ﬁsh using computer vision / Lakehal and Lakehal UNIVERSIDAD DEL ZULIA Serbiluz Sistema de Servicios Bibliotecarios y de Información Biblioteca Digital Repositorio Académico INTRODUCTION The European sea bass (Dicentrarchus labrax) is among the most economically valuable marine ﬁsh species for aquaculture in the Mediterranean region [1]. In Algeria, although aquaculture has yet to reach the producon stage, but it is emerging in the coastal region through some private schemes being developed, parcularly those involving the farming of European sea bass (Dicentrarchus labrax) [2]. As stated in [3], the freshness of ﬁsh is a signiﬁcant parameter. This is due to the fact that it is related not only to the health of the ﬁsh, but also to the assurance of consumer acceptance. However, the quality and freshness of European sea bass can deteriorate at various levels, including processing, transportaon, retail, and domesc storage [3],4]. Denaturaon of proteins, oxidaon of lipids, bacterial growth, and autolyc processes are also among the reasons for such loss [5 , 6]. Quantave analysis of stored ﬁsh quality is typically accomplished through the ulisaon of microbiological and chemical methodologies. These methodologies encompass the measurement of volale basic nitrogen, thiobarbituric acid content, and the total number of bacteria [7 ,8,9]. The researchers have expressed conﬁdence in their ﬁndings, but it should be noted that the study is not without its inherent limitaons. The process is costly, me-consuming, destrucve, and in certain cases, hazardous when the reagents are considered [10 , 11]. These disadvantages are indicave of the urgent need to develop a cost-eﬀecve, rapid, accurate, and environmentally sound method for the eﬀecve evaluaon of ﬁsh quality. Advances in image technologies and computer vision systems are creang new opportunies for the non-destrucve assessment of ﬁsh freshness [1]. A computer vision system provides valuable informaon about color properes and can be ulized in overall ﬁsh quality predicon [12], 13]. Previous studies have already demonstrated the eﬀecveness of computer vision in measuring the freshness of ﬁsh [14 , 15 , 16 , 17 , 18]. Arﬁcial neural networks (ANN) have also become highly potent predicon and modeling tools [19]. Such systems are coded to replicate the funconing of biological neurons, parcularly those within the nervous system and brain [20] In the context of computer applicaons, an arﬁcial neural network is disnguished by its remarkable capacity to learn and to idenfy and represent intricate, non-linear relaonal paerns between the input and output of a system [21]. This study demonstrated that this method, employing computer vision and arﬁcial intelligence, enables non-destrucve and accurate meat quality assessment. Similarly, Lalabadi et al. [22] suggested a non-destrucve method for evaluang ﬁsh freshness by observing the eyes and gills color using arﬁcial neural networks (ANNs). Similar studies on ﬁsh freshness classiﬁcaon (large yellow croaker) have also been conducted with computer vision techniques using arﬁcial neural networks [23]. Among external indicators, the eye is a parcularly praccal choice for imaging analysis, which makes the process quicker and more hygienic. On the other hand, examinaon of gills although informave, requires opening and manipulaon of internal ssues, which is more invasive and not suitable for non- destrucve method. Computer vision technology and predicve modeling studies have not been thoroughly studied to enhance the accuracy of ﬁsh freshness detecon in cold storage. In the present study, the cornea of the eye of Dicentrarchus labrax was selected as the research subject, with measurement undertaken using colour parameters L*, a*, b*, R, G, B, and H, S, I. The objecve is to develop a reliable predicon model of ﬁsh spoilage by employing an arﬁcial neural network (ANN) to analyse eye colour. MATERIALS AND METHODS Sea bass samples Ten whole, non-eviscerated European sea bass (Dicentrarchus labrax) were purchased from a local ﬁsh market in Batna, Algeria. The ﬁsh were placed in a polystyrene bag ﬁlled with ice blocks and transported to the laboratory within 24 hours (h). All ﬁsh were stored in a refrigerator (CRF–NT64GF40, Condor, Algeria) at a stable temperature of +4 ± 0.2 °C. Samples were taken for imaging every three days (d). Computer vision system The image acquision setup comprised a custom-designed enclosure equipped with two adjustable lamps posioned 50 cm above the samples at a 45° angle to achieve consistent illuminaon. A digital camera (Canon DS126621, China) was ﬁxed vercally at a distance of 30 cm from the sample surface. To reduce external light interference, the interior of the enclosure was covered with light-absorbing black fabric [24]. Digital analysis of colorimetric parameters, including Lab, HSI, and RGB, was performed using Adobe Photoshop CS3 soﬅware (FIG. 1). FIGURE 1. Computer Vision System Modelling with ANN The development and integraon of arﬁcial neural networks (ANNs) are a strict selecon of some architectural features. Among the numerous ANN architectures available, a mullayer perceptron (MLP) was employed in this research. The MLP architecture has three major layers: an input layer, one or mulple hidden layers, and an output layer [25]. The neurons of the input layer portray three color aributes like Lab, HSI, and RGB of frozen/thawed ﬁsh, while the output layer depicts the shelf life. The hidden layer(s) contain adjustable neurons, and the correct number of nodes is determined empirically because there is no rigid rule to determine the size of the hidden layer. The design and opmizaon were conducted via the MATLAB interface. 2 of 7

Revista Cienﬁca, FCV-LUZ / Vol. XXXV UNIVERSIDAD DEL ZULIA Serbiluz Sistema de Servicios Bibliotecarios y de Información Biblioteca Digital Repositorio Académico For performance evaluaon of the network, standard stascal measures were employed, i.e., correlaon coeﬃcient (R 2 ), mean squared error (MSE), and mean absolute error (MAE). R 2 value reﬂected goodness-of-ﬁt of the model through esmaon of how closely predicted points coincided with actual points. MSE provided a well-established predicve accuracy measure, parcularly for external validaon sets. In addion, MAE was used as a measure of performance in predicve models, with smaller values represenng greater accuracy [21]. Stascal analysis Data analysis was carried out using analysis of variance (ANOVA) through SPSS soﬅware version 22 (IBM SPSS Stascs v22). Tukey’s post hoc test was applied to compare the means. Stascal signiﬁcance was deﬁned at a threshold of P < 0.05, while P values equal to or greater than 0.05 were considered not signiﬁcant. RESULTS AND DISCUSSION Analysis of ﬁsh freshness through eye observaon Over me, the ﬁsh eye undergoes a sequence of changes, which mirror the increasing loss of freshness. Inially, in FIG. 2, the fresh eye (0 d) of the sea bream was clear, shiny, and slightly protruding, which indicated maximum freshness. With longer storage periods, the eye became duller and brownish in color (FIG. 2), indicang advanced decomposion [26]. Fish eyes, which contain high levels of lipids, are suscepble to oxidaon when stored due to the acon of endogenous enzymes and microbial metabolism [22]. Oxidaon of lipids results in the formaon of peroxides, which subsequently deteriorate into low-molecular-weight alcohols and carbonyl compounds. These compounds have been shown to aﬀect the color, odor, texture, and ﬂavor of the ﬁsh [27]. Furthermore, the high water content of the eyeball leads to delayed moisture loss, which can result in the collapse of the eye over me. This phenomenon has been parcularly observed in lapia [28]. These visual and chemical alteraons are indicave of the progressive loss of ﬁsh freshness. 0 day 3 days 6 days 9 days 12 days 15 days FIGURE 2. Change in the original image of ﬁsh eyes during storage Analysis of ﬁsh freshness through colors parameters The FIG. 3 a shows that with the increase in storage me, there was a steady decline in R, G, and B values, with a noteworthy drop aﬅer the sixth d (P < 0.05). The simultaneous decrease in chromac components suggests a loss in eye brightness, likely due to changes in biochemical and structural composion over me. The stabilizaon of the values from the ninth day onwards indicates that degradaon is where pigments responsible for coloring are extensively degraded or oxidized. Such physical changes can be ulized as indicators of ﬁsh freshness, in addion to inﬂuencing its aracveness to consumers [29 ,[30]. According to FIG 3b, ﬁsh freshness was predominantly determined through the measurement of brightness (L*), which is one of the key parameters disnguishing fresh from spoiled ﬁsh. As me went on, the value of (L*) signiﬁcantly decreased, from 88.24 to approximately 33.06 (P < 0.05), reﬂecng a gradual darkening of the eyes. This decrease in brightness can be aributed to one or more factors combined, including air exposure, drying out, and chemical and biochemical breakdown, such as pigment oxidaon and cell structure deterioraon, compounded by prolonged storage at low temperature [31]. The (b*) value also decreased signiﬁcantly (P < 0.05) up to day six of storage, indicang loss of yellow pigment component, before increasing exponenally on day nine (P < 0.05), which could be an indicaon of an accumulaon of secondary pigments related to biochemical breakdown. As a consequence of this increase, (b*) held steady unl the ﬁnal d of the storage me. Regarding the (a*) value, it remained relavely stable during the inial days of storage, then increased sharply around the ninth day, and subsequently remained stable unl the ﬁnal d of the storage period. The HSI colour model, which is closely related to the physiology of the human eye, characterised by three components (H, S and I) that operate relavely independently. In the HSI colour model, the H component deﬁnes the main hue of a colour and is measured as an angle between 0° and 360° in the visible spectrum, where 0° corresponds to red, 60° to yellow, 120° to green and 180° to cyan [32]. During storage (FIG 3c), the H value progressively decreased before stabilising around the twelﬅh and ﬁﬅeenth days of storage (≥ 0.05), rendering it ineﬀecve in reﬂecng changes in freshness over me. The S component represents the degree of colour purity, also known as saturaon, based on the spectral distribuon of light. Although saturaon can ﬂuctuate under diﬀerent lighng condions, the S value generally remains stable, ensuring consistency in colour assessment [33]. The S-value of the eye of Dicentrarchus labrax gradually increased during storage, indicang a progressive darkening of its original colour. This change intensiﬁed the hue and improved its depth, leading to a general darkening of the eye as storage me increased. The I component of the HSI colour space is a subjecve parameter that exclusively represents the luminosity of the image; unlike the other components, it contains no colour informaon and has no inﬂuence on colour percepon. Its role is therefore limited to evaluang light intensity independently of colour variaons [34]. Throughout storage, the I value of the eyes of gilthead sea bream (Sparus aurata) decreased signiﬁcantly, from 139.55 to 73.87. This decrease reﬂects a progressive reducon in luminosity. 3 of 7

Eﬀects of chilled storage on ﬁsh using computer vision / Lakehal and Lakehal UNIVERSIDAD DEL ZULIA Serbiluz Sistema de Servicios Bibliotecarios y de Información Biblioteca Digital Repositorio Académico FIGURE 3. Change of color parameters (R, G, B, L*, a*, b*, H, S, I) in ﬁsh eyes during storage a b c Arﬁcial neural network Designing an arﬁcial neural network (ANN) model requires extra cauon in deﬁning its topology, especially in terms of how many neurons there are in the hidden layer [35]. Diﬀerent architectures were aempted for this work and aﬅer a series of comparave experiments, the best conﬁguraon selected was a mullayer perceptron (MLP) having 20 neurons in the hidden layer. The symbolic diagram of the ﬁnal ANN model is given in FIG.4. FIGURE 4. Illustraon of a typical mul-layer neural network 4 of 7 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

Revista Cienﬁca, FCV-LUZ / Vol. XXXV UNIVERSIDAD DEL ZULIA Serbiluz Sistema de Servicios Bibliotecarios y de Información Biblioteca Digital Repositorio Académico The number of neurons that provided opmal funconality was determined through rigorous analysis. This analysis was focused on maximising the coeﬃcient of determinaon (R²) while minimising the mean squared error (MSE). The results of this analysis are summarised in TABLE I. It was observed that the R² and MSE values exhibited signiﬁcant variaon up to the addion of 18 neurons. The incorporaon of an addional 19 neurons, followed by a further 20 neurons, resulted in the stabilisaon of R² at a high value of 0.98. Concurrently, the MSE values exhibited ﬂuctuaons that were not conngent on the number of neurons, with the lowest values being aained with 18 and 19 neurons. Conversely, the use of a single neuron in the hidden layer yielded the highest MSE value (0.190), underscoring the signiﬁcance of suﬃcient neuron count to ensure model eﬃcacy (TABLE I). Despite the main objecve of maximizing R² and minimizing MSE, the most balanced results were obtained with 18 neurons, this conﬁguraon oﬀering a slight improvement in MSE (MSE=0.028825) compared to the 20-neuron conﬁguraon. A high correlaon coeﬃcient (R² = 0.98) conﬁrms the robustness of the model and the relevance of the variables used in its design (FIGS. 5a; 5b). This performance highlights the potenal of ANN modeling for praccal implementaon in freshness predicon systems using non-destrucve colorimetric input. These observaons are consistent with results reported in the literature for example, Liu et al. [36] showed that the inclusion of six neurons in the hidden layer led to the minimum MSE in an ANN model for rainbow trout ﬁllet quality evoluon predicon (Oncorhynchus mykiss). In a similar applicaon, Rezende-de-Souza et al. [37] developed an ANN model for ﬁsh quality evaluaon based on colorimetric parameters (CIELab) and total volale basic nitrogen (TVB-N) and concluded that its performance was best achieved with only three neurons. TABLE I The results of the arﬁcial neural network model assessment Number of the neurons MSE R 2 MAE 1 0.195242 0.896824 2.1036 2 0.159423 0.916622 1.7513 3 0.132795 0.931068 1.5681 4 0.114666 0.940776 2.2658 5 0.114502 0.940864 1.8482 6 0.104563 0.946143 1.8103 7 0.09907 0.949049 1.8176 8 0.085285 0.956301 1.7547 9 0.081388 0.958341 2.3661 10 0.075843 0.961236 1.7370 11 0.077278 0.960488 1.9206 12 0.074028 0.962182 1.9699 13 0.061524 0.968674 2.1000 14 0.050462 0.974381 1.9285 15 0.052752 0.973202 2.3769 16 0.043977 0.97771 2.0734 17 0.029439 0.985135 2.3978 18 0.028825 0.985447 4.2540 19 0.034111 0.982755 3.4642 20 0.029932 0.984884 2.7007 MSE: mean squared error R 2 : correlaon coeﬃcient MAE: mean absolute error FIGURE 5. The network MSE and R2 values vs the number of neurons in the hidden layer a b 5 of 7 0.25 0.2 0.15 0.1 0.05 0 1 0.98 0.96 0.94 0.92 0.9 0.88

Eﬀects of chilled storage on ﬁsh using computer vision / Lakehal and Lakehal UNIVERSIDAD DEL ZULIA Serbiluz Sistema de Servicios Bibliotecarios y de Información Biblioteca Digital Repositorio Académico CONCLUSION The ﬁndings of this study demonstrate that a computer vision system, ulising ﬁsheye analysis, can serve as an eﬀecve tool for assessing the freshness of European seabass without resorng to destrucve techniques. The approach via arﬁcial neural networks yielded excellent classiﬁcaon precision on the basis of storage days. By using arﬁcial neural networks (ANN), high predicon accuracy was achieved in classifying ﬁsh according to storage me, conﬁrming the strong relaonship between ocular colour parameters and spoilage progression. Furthermore, it was evident that variaons in ocular colouraon during storage periods could serve as a uniform criterion for disnguishing between freshness levels. This non-invasive approach eliminates the need for tradional destrucve tesng, reduces analysis me, and allows for real-me applicaon in industrial sengs. This would not only enhance eﬃciency but also improve the quality control measures employed within the industry. The research ﬁndings of this study can be used to develop an automac and eﬃcient ﬁsh sorng system based on freshness. Conﬂict of interests The authors declare no conﬂict of interest regarding the publicaon of this manuscript. BIBLIOGRAPHIC REFERENCES [1] Erdağ M, Ayvaz Z. The use of color to determine ﬁsh freshness: European seabass (Dicentrarchus labrax). J. Aquat. Food Prod. Technol. [Internet]. 2021; 30(7):847– 867. doi: hps://doi.org/pwvr [2] Mokrani D, Oumouna M, Cuesta A. Fish farming condions aﬀect to European sea bass (Dicentrarchus labrax L.) quality and shelf life during storage in ice. Aquaculture. [Internet]. 2018; 490:120–124. doi: hps:// doi.org/gdcqwq [3] Wu X, Zhang Q, Wang Z, Wang Z, Yan H, Zhu L, Chang J. Nondestrucve freshness predicon of large yellow croaker (Pseudosciaena crocea) using computer vision and machine learning techniques based on pupil color. J. Food Sci. [Internet]. 2024; 89(12):9392–9406. doi: hps://doi.org/pwvt [4] Yi Z, Xie J. Assessment of spoilage potenal and amino acids deaminaon & decarboxylaon acvies of Shewanella putrefaciens in bigeye tuna (Thunnus obesus). LWT. [Internet]. 2022; 156:113016. doi: hps:// doi.org/gq2wg7 [5] Cheng JH, Sun DW. Rapid and non-invasive detecon of ﬁsh microbial spoilage by visible and near infrared hyperspectral imaging and mulvariate analysis. LWT Food Sci. Technol. [Internet]. 2015; 62(2):1060–1068. doi: hps://doi.org/gq33j4 [6] Zhang Y, Qin N, Luo Y, Shen H. Eﬀects of diﬀerent concentraons of salt and sugar on biogenic amines and quality changes of carp (Cyprinus carpio) during chilled storage. J. Sci. Food Agric. [Internet]. 2015; 95(6):1157– 1162. doi: hps://doi.org/gj89h7 [7] Chun HN, Kim B, Shin HS. Evaluaon of a freshness indicator for quality of ﬁsh products during storage. Food Sci. Biotechnol. [Internet]. 2014; 23:1719–1725. doi: hps://doi.org/f6nkz3 [8] Liao Q, Wei C, Li Y, Gou L, Ouyang H. Developing a machine vision system equipped with UV light to predict ﬁsh freshness based on ﬁsh-surface color. Food Nutr. Sci. [Internet]. 2021; 12(3):239–248. doi: hps://doi.org/ gpbc4m [9] Sigurgislador S, Hafsteinsson H, Jonsson A, Lie Ø, Nortvedt R, Thomassen M, Torrissen O. Textural properes of raw salmon ﬁllets as related to sampling method. J. Food Sci. [Internet]. 1999; 64(1):99–104. doi: hps://doi.org/dmntwv [10] Prabhakar PK, Vatsa S, Srivastav PP, Pathak SS. A comprehensive review on freshness of ﬁsh and assessment: Analycal methods and recent innovaons. Food Res. Int. [Internet]. 2020; 133:109157. doi: hps:// doi.org/mpdk [11] Wu L, Pu H, Sun DW. Novel techniques for evaluang freshness quality aributes of ﬁsh: A review of recent developments. Trends Food Sci. Technol. [Internet]. 2019; 83:259–273. doi: hps://doi.org/gjvdn3 [12] Gümüş B, Gümüş E, Balaban MO. Color of rainbow trout (Oncorhynchus mykiss) ﬁllets by image and sensory analysis, and correlaon with SalmoFan numbers. J. Food Sci. [Internet]. 2023; 88(1):430–446. doi: hps://doi.org/ pwv2 [13] Korel F, Luzuriaga DA, Balaban MÖ. Objecve quality assessment of raw lapia (Oreochromis nilocus) ﬁllets using electronic nose and machine vision. J. Food Sci. [Internet]. 2001; 66(7):1018–1024. doi: hps://doi.org/ cv9zq7 [14] Cengizler C. Fish spoilage classiﬁcaon based on color distribuon analysis of eye images. Mar. Sci. Technol. Bull. [Internet]. 2023; 12(1):63–69. doi: hps://doi.org/ pwv4 [15] Yasin ET, Ozkan IA, Koklu M. Detecon of ﬁsh freshness using arﬁcial intelligence methods. Eur. Food Res. Technol. [Internet]. 2023; 249(8):1979–1990. doi: hps://doi.org/pwv5 [16] Huang X, Xu H, Wu L, Dai H, Yao L, Han F. A data fusion detecon method for ﬁsh freshness based on computer vision and near-infrared spectroscopy. Anal. Methods. [Internet]. 2016; 8(14):2929–2935. doi: hps://doi.org/ gjvdpb [17] Gümüş B, Balaban MO. Predicon of the weight of aquacultured rainbow trout (Oncorhynchus mykiss) by image analysis. J. Aquat. Food Prod. Technol. [Internet]. 2010; 19(3–4):227–237. doi: hps://doi.org/cgs3pv [18] Misimi E, Erikson U, Digre H, Skavhaug A, Mathiassen JR. Computer vision-based evaluaon of pre- and postrigor changes in size and shape of Atlanc cod (Gadus morhua) and Atlanc salmon (Salmo salar) ﬁllets during rigor mors and ice storage: eﬀects of perimortem handling stress. J. Food Sci. [Internet]. 2008; 73(2):E57–E68. doi: hps://doi.org/b48bbd [19] Huang Y, Kangas LJ, Rasco BA. Applicaons of arﬁcial neural networks (ANNs) in food science. Crit. Rev. Food Sci. Nutr. [Internet]. 2007; 47(2):113–126. doi: hps:// doi.org/bﬀx3k 6 of 7

Revista Cienﬁca, FCV-LUZ / Vol. XXXV UNIVERSIDAD DEL ZULIA Serbiluz Sistema de Servicios Bibliotecarios y de Información Biblioteca Digital Repositorio Académico [20] Gonzalez-Fernandez I, Iglesias-Otero MA, Esteki M, Moldes OA, Mejuto JC, Simal-Gandara J. A crical review on the use of arﬁcial neural networks in olive oil producon, characterizaon and authencaon. Crit. Rev. Food Sci. Nutr. [Internet]. 2019; 59(12):1913–1926. doi: hps://doi.org/ggb42s [21] Lakehal S, Lakehal B, Chadi H, Bennoune O, Ayachi A. Eﬀects on beef microstructure using fractal dimension and ANN modelling. J. Hellenic. Vet. Med. Soc. [Internet]. 2024; 75(4):8281–8290. doi: hps://doi.org/pwv6 [22] Lalabadi HM, Sadeghi M, Mireei SA. Fish freshness categorizaon from eyes and gills color features using mul-class arﬁcial neural network and support vector machines. Aquacult. Eng. [Internet]. 2020; 90:102076. doi: hps://doi.org/kgdk [23] Zheng Y, Zhang Q, Wang X, Guo Q. Classifying the freshness of large yellow croaker (Larimichthys crocea) at 12-and 24-hour intervals using computer vision technique and convoluonal neural network. Smart Agric. Technol. [Internet]. 2025; 10:100767. doi: hps:// doi.org/pwv7 [24] Lakehal S, Lakehal B. Storage me predicon of frozen meat using arﬁcial neural network modeling with color values. Rev. Cient. FCVLUZ. [Internet]. 2023; 33(2):1–6. doi: hps://doi.org/pwv8 [25] Lakehal B, Dibi Z, Lakhdar N, Dendouga A. Electrical equivalent model of intermediate band solar cell using PSpice. Sadhana. [Internet]. 2015; 40:1473–1479. doi: hps://doi.org/kgdn [26] Dowla M, Mohtasebi SS, Omid M, Razavi SH, Jamzad M, De La Guardia M. Freshness assessment of gilthead sea bream (Sparus aurata) by machine vision based on gill and eye color changes. J. Food Eng. [Internet]. 2013; 119(2):277–287. doi: hps://doi.org/pwv9 [27] Masniyom P. Deterioraon and shelf-life extension of ﬁsh and ﬁshery products by modiﬁed atmosphere packaging. Songklanakarin J. Sci. Technol. [Internet]. 2011 [cited Feb 25 2025]; 33(2):181–192. Available in: hps://goo.su/ XjfL [28] Stoknes IS, Økland HM, Falch E, Synnes M. Fay acid and lipid class composion in eyes and brain from teleosts and elasmobranchs. Comp. Biochem. Physiol. B. Biochem. Mol. Biol. [Internet]. 2004; 138(2):183–191. doi: hps:// doi.org/bcb45h [29] Jia Z, Li M, Shi C, Zhang J, Yang X. Determinaon of salmon freshness by computer vision based on eye color. Food Packag. Shelf Life. [Internet]. 2022; 34:100984. doi: hps://doi.org/pwwb [30] Chmiel M, Słowiński M, Dasiewicz K, Florowski T. Use of computer vision system (CVS) for detecon of PSE pork meat obtained from m. semimembranosus. LWT Food Sci. Technol. [Internet]. 2016; 65:532–536. doi: hps:// doi.org/pwwc [31] Shi C, Qian J, Han S, Fan B, Yang X, Wu X. Developing a machine vision system for simultaneous predicon of freshness indicators based on lapia (Oreochromis nilocus) pupil and gill color during storage at 4°C. Food Chem. [Internet]. 2018; 243:134–140. doi: hps://doi. org/gjvdpk [32] Zhou B, Elazab A, Bort J, Vergara O, Serret MD, Araus JL. Low-cost assessment of wheat resistance to yellow rust through convenonal RGB images. Comput. Electron. Agric. [Internet]. 2015; 116:20–29. doi: hps://doi.org/ f7pgr6 [33] Cheng HD, Jiang XH, Sun Y, Wang J. Color image segmentaon: Advances and prospects. Paern Recognit. [Internet]. 2001; 34(12):2259–2281. doi: hps://doi.org/ dgw93m [34] Dhandra BV, Hegadi R, Hangarge M, Malemath VS. Analysis of abnormality in endoscopic images using combined HSI color space and watershed segmentaon. In: 18th Internaonal Conference on Paern Recognion (ICPR’06). Vol. 4. China: IEEE; 2006. p. 695–698. doi: hps://doi.org/cxkh5f [35] Aggabou LK, Lakehal B, Mouda M. An Arﬁcial Neural Network Approach for Construcon Project Risk Management. Int. J. Saf. Secur. Eng. [Internet]. 2024; 14(2):553–561. doi: hps://doi.org/pwwd [36] Liu X, Jiang Y, Shen S, Luo Y, Gao L. Comparison of Arrhenius model and arﬁcial neuronal network for the quality predicon of rainbow trout (Oncorhynchus mykiss) ﬁllets during storage at diﬀerent temperatures. LWT Food Sci. Technol. [Internet]. 2015; 60(1):142–147. doi: hps://doi.org/gsrqzm [37] Rezende-de-Souza JH, de Moraes-Neto VF, Cassol GZ, dos Santos Camelo MC, Savay-da-Silva LK. Use of colorimetric data and arﬁcial neural networks for the determinaon of freshness in ﬁsh. Food Chem. Adv. [Internet]. 2022; 1:100129. doi: hps://doi.org/pwwf 7 of 7