At least 10 million tons of seafood products are spoiled or damaged during transportation or storage every year worldwide. Economic loss due to fish spoilage exceeds 2 billion euros every year. This is due to unplanned consumer purchases and the inability to effectively determine the shelf life of seafood, which results in the spoilage of many seafood products, leading to wastage [1]. Traditional freshness monitoring methods are not useful in many situations and require professional knowledge support, which does not allow consumers to accurately judge the remaining shelf-life of food in real-time [2]. The pH-responsive smart packaging method is a convenient and fast non-destructive testing method, which has been widely used in the smart packaging of fresh food [3,4]. Anthocyanins are natural pigments; therefore, they are more suitable for food smart packaging than chemical dyes [5]. Phenolic acids and plant extracts can form non-covalent complexes with anthocyanins, causing bathochromic shifts and hyperchromic effects and enhancing the coloration of anthocyanins [6]. Mechanisms include intermolecular stacking of anthocyanins through specific alignment of aromatic acyl residues with the flavylium cation and intramolecular interactions between anthocyanins and co-pigments leading to π-π stacking [7]. In addition, anthocyanins can form ionic complexes with anionic polysaccharides, such as chondroitin sulfate. Charge-charge interactions between the strong negative groups of sulfated polysaccharides and flavonoid cations, as well as intermolecular stacking and hydrophobic interactions, are responsible for the formation of the complexes [8,9]. And like sulfated polysaccharides, positively charged flavonoid cations of anthocyanins can form ionic complexes directly with carboxyl groups having polysaccharides, including alginates [10]. These results provide new ideas and directions for intelligent packaging.

Polysaccharide hydrogels are a class of materials with a unique three-dimensional network structure. They exhibit excellent biocompatibility and controllable drug-release ability [11]. Hydrogels have high water content and their responsiveness is not easily affected by the humidity level in the environment due to their structural characteristics [12]. Traditional hydrogels have limitations due to their brittleness. Multiple cross-linking systems in hydrogels can impart excellent mechanical properties [13]. trans-2-Hexenal is a green leaf volatile that has been widely used as a food additive. Some studies have shown that it has good antibacterial activity [14]. Polyvinyl alcohol can alter the properties of polymers, exhibit good biocompatibility, and is widely used in food materials. It can undergo an aldol condensation reaction and form chemical cross-linking under the action of acidic or alkaline catalysts [15]. The chemical cross-linking network between citric acid and polyvinyl alcohol can be formed by esterification reaction [16]. Machine learning can provide more possibilities for smart packaging, which can more accurately recognize the pheromones of smart packaging and feedback to the user. Among them, the back propagation model (BP) is a neural network that adjusts the weights and biases of the network through a back-propagation algorithm, and the radial basis function model (RBF) is a neural network based on radial basis functions, both of which can deal with complex nonlinear problems [17]. The Genetic Algorithm-BP model (GA-BP) solves the problem that BP models tend to fall into local optimization by using genetic algorithm [18].The extreme learning machine model (ELM) does not need to adjust the weights constantly, which means that it has an extremely fast learning capability [19]. The superiority of these four models has been well proven in the engineering field [20]. In addition, multichannel data further extends the potential applications of machine learning. Machine learning models can provide a more accurate understanding of the condition of an environment or system by considering multiple parameters simultaneously, providing a more comprehensive basis for intelligent decision making [21].

In this study, we enhanced the chromogenic response of blueberry and bilberry anthocyanins to biogenic amines by the co-pigmentation effect of phenolic acids, enhanced anthocyanin stability by ionic complexation generated between sodium alginate and blueberry and bilberry anthocyanins, and created a dual cross-linking network of the hydrogel using trans-2-hexenal, polyvinyl alcohol, and citric acid, to construct a hydrogel with volatile antimicrobial properties and excellent mechanical properties. The cross-linked structures of the hydrogels were analyzed using Fourier-transform infrared spectroscopy (FTIR) and X-ray diffraction (XRD), and their freeze–thaw resistance, broad-spectrum antimicrobial activity, and mechanical properties were characterized. Additionally, we established a two-channel monitoring system based on hydrogel color characteristics, trained four machine learning models, and constructed a real-time prediction platform that integrates data analysis, machine learning, and interactive user interface using optimal machine learning algorithms. By simply uploading a photo of the smart hydrogel to the platform, users can determine the freshness level of rainbow trout in real time based on color characteristics (Fig. 1).