Canned fish quality detection system and method based on intelligent algorithm

Abstract

The invention discloses a canned fish quality detection system and method based on an intelligent algorithm, and relates to the technical field of can detection.The method comprises the steps that after abnormal cans are preliminarily screened, sound vibration energy transfer mapping analysis is further executed, and energy values at different resonant frequencies of all collection points under multi-frequency excitation are extracted; and constructing an energy normalization mapping matrix Emap, and outputting a comprehensive freshness score Sfinal by calculating the mapping matrix Emap and combining the total pedigree drift distance Stotal and adopting an exponential function mapping mode. Based on the comprehensive freshness score Sfinal, a freshness secondary evaluation standard is set, intelligent and precise classification of whether the canned fish is fresh or not is achieved, and follow-up treatment measures such as replacement packaging or batch risk marking are executed according to different freshness scores. Through drift trajectory and energy anomaly two-dimensional fusion evaluation, the early recognition precision of tiny qualitative change of the internal tissue of the canned fish is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of canned food detection, and particularly to a quality detection system and method for fish cans based on intelligent algorithms. Background Art

[0002] Fish cans are ready-to-eat canned products made from fresh or frozen fish through processes such as processing, canning, adding seasonings, sealing, and sterilization. According to different processing methods, fish cans are divided into categories such as braised, tomato sauce, deep-fried, steamed, smoked, oil-packed, and water-packed. The fish can container has a cylindrical structure with an opening at the top, which is often sealed with metal materials such as tinplate or aluminum alloy. And the inside of the fish can container is coated with an anti-corrosion coating to ensure the safety and hygiene of the food sealing layer. During the storage of fish cans, it is necessary to ensure the integrity of the seal of the fish can container, and to determine the absence of leakage, swelling, etc. in the fish cans. At the same time, the outer surface of the fish can container should not have rust, and the coating on the inner wall of the container should not peel off, so as to ensure the quality of the fish cans. When conducting quality inspection on fish can products, since fish cans are foods with high protein and high nutritional value and need to have a long storage period, higher requirements are also put forward for monitoring the internal quality changes.

[0003] Food safety detection belongs to the large field of the cross-integration of food engineering and intelligent manufacturing. With the improvement of consumers' requirements for food quality and the continuous growth of the global food trade volume, ensuring the freshness and quality of processed foods, especially canned products that are sealed and stored for a long time, has become a crucial research direction in the food industry. In the field of fish can product detection, traditional methods often rely on opening cans for spot checks, chemical analysis, or sensory evaluation to detect the internal quality of fish cans to ensure the quality of fish cans. However, correspondingly, the above detection methods also lead to low efficiency, and it is impossible to achieve non-destructive and rapid detection and screening. At the same time, the detection accuracy is relatively low, greatly reducing the detection efficiency of fish cans. Therefore, there are generally problems such as strong destructiveness of detection means, long detection cycle, low detection efficiency, and strong subjectivity of evaluation criteria in the freshness detection of fish cans. Opening cans for detection will damage the circulation value of the product and the sample coverage rate is limited. At the same time, analyzing chemical components requires a lot of time and manpower, and it is difficult to apply it on a large scale in real time. And conducting sensory detection on fish cans, such as judging the smell and color, has great human error and environmental dependence. Moreover, even existing non-destructive detection technologies are usually based on static sound analysis or can body appearance detection, limited to detecting large-scale quality changes and unable to capture early tiny physical changes inside the can. Especially in the initial stage of spoilage, when autolysis of fish tissue, changes in viscosity, and liquefaction phenomena have not produced obvious gas or appearance abnormalities, traditional methods are difficult to give effective early warnings, resulting in potential risk products flowing into the market and increasing food safety hazards.

[0004] Therefore, how to intelligently detect the quality of canned fish is a technical problem that technicians need to solve at present. This application proposes a quality detection method for canned fish based on the dynamic analysis of the resonant spectrum of the canned fish body itself and combined with intelligent algorithms, which is a key innovation direction to promote the development of this field. Summary of the Invention

[0005] In view of the deficiencies of the prior art, the present invention provides a quality detection system and method for canned fish based on intelligent algorithms to solve the problems mentioned in the background art.

[0006] To achieve the above object, the first aspect of this application provides a quality detection method for canned fish based on intelligent algorithms, including the following steps:

[0007] S1. Set a plurality of percussion acquisition points on the canned fish to be tested, perform set percussion on each percussion acquisition point according to a set path, and collect the resonant response signals generated by each percussion acquisition point in real time through a sensor group;

[0008] S2. Transmit each of the resonant response signals to a canned fish detection server, extract features of the resonant response signals through the canned fish detection server to obtain a vibration feature set, and preprocess the vibration feature set to obtain a standard vibration feature set;

[0009] S3. Based on the standard vibration feature set, calculate the main frequency change trajectory F of each percussion acquisition point, and then perform summary calculation based on the main frequency change trajectory F to obtain the total spectrum drift amount △Stotal, and preliminarily compare and evaluate the total spectrum drift amount △Stotal with a freshness screening threshold F1 to judge the structural integrity of the canned fish;

[0010] S4. When it is determined that the canned fish to be tested is abnormal in the comparison and evaluation, perform acoustic-vibration energy transfer mapping analysis, and perform matrix modeling on the acoustic-vibration energy transfer after multi-frequency micro-percussion to obtain a mapping matrix Emap;

[0011] S5. Based on the total spectrum drift amount △Stotal and the mapping matrix Emap, calculate and output a comprehensive canned fish freshness score Sfinal, and perform a secondary comparison and evaluation based on the output result of the comprehensive canned fish freshness score Sfinal to determine the freshness of the canned fish to be tested.

[0012] Preferably, in S1, the device for performing set percussion on each percussion acquisition point according to a set path is a micro electromagnetic impactor, and the sensor group includes a MEMS accelerometer;

[0013] The percussion acquisition points are respectively at the top, middle waist and bottom of the canned fish to be tested.

[0014] Preferably, in the step S2, the fish can detection server extracts the features of the resonance response signal, including:

[0015] Performing a short-time Fourier transform (STFT) on the resonance response signal to obtain an STFT spectrogram, and extracting the frequency corresponding to the position with the maximum energy in the STFT spectrogram as the initial main frequency value \(f\) of the \(i\)-th acquisition point; 0，i ;

[0016] Setting a search bandwidth around the initial main frequency value \(f\) of the \(i\)-th acquisition point, identifying different harmonic peaks within the search bandwidth, and extracting the amplitudes of each peak in the harmonic peaks as the amplitude \(a\); 0，i ; i，n ;

[0017] Based on the STFT spectrogram, extracting the \(n\)-th resonance frequency \(f\) of the \(i\)-th acquisition point, i，n calculating to obtain the angular frequency \(\omega\), i，n specifically, the calculation formula is: \(\omega\) i，n \(= 2\pi f\), i，n where \(\pi\) represents the pi;

[0018] Based on the STFT spectrogram, extracting the \(n\)-th resonance frequency \(f\) of the \(i\)-th acquisition point, i，n extracting a small frequency band of \(\pm0.5\) Hz for integration to obtain the vibration energy \(E(f\) i，n );

[0019] where the vibration feature set includes the initial resonance frequency \(f\), 0，i the amplitude \(a\), i，n the angular frequency \(\omega\), i，n the initial phase \(X_{\omega}\), i，n and the vibration energy \(E(f\) i，n ).

[0020] Preferably, in the step S3, based on the standard vibration feature set, calculating the main frequency change trajectory \(F\) of each percussion acquisition point, including:

[0021] Constructing a trajectory model of the main resonance frequency changing with time;

[0022] Inputting the vibration feature set obtained from each acquisition point into the trajectory model, and calculating and outputting the main frequency change trajectory \(F\) of each percussion acquisition point. The calculation method of the main frequency change trajectory \(F\) is:

[0023]

[0024] where \(f\) 0，i represents the initial resonance frequency, and \(F\) i(t) represents the main frequency change trajectory of the i-th acquisition point at time t, N represents the total number of resonant response signals, sin represents the cosine function, W i，n (t) represents the angular frequency of the n-th resonant component of the i-th acquisition point at time t.

[0025] Preferably, in the step S3, the calculation formula for the total spectral drift amount △Stotal is:

[0026]

[0027] In the formula, M represents the total number of acquisition points, T represents the detection time window, and dt represents the time differential variable.

[0028] Preferably, in the step S3, the preliminary comparison and evaluation of the total spectral drift amount △Stotal with the freshness screening threshold F1 includes:

[0029] When the total spectral drift amount △Stotal < the freshness screening threshold F1, it indicates that the resonant drift of the canned fish to be tested is normal;

[0030] When the total spectral drift amount △Stotal ≥ the freshness screening threshold F1, it indicates that the resonant drift of the canned fish to be tested is abnormal, and the acoustic vibration energy transfer mapping analysis is triggered.

[0031] Preferably, the triggering of the acoustic vibration energy transfer mapping analysis includes:

[0032] Extracting the vibration energy E(f i，n at the n-th resonant frequency f of the i-th acquisition point at different frequencies for each knocking acquisition point i，n );

[0033] Performing matrix modeling on each of the vibration energies E(f i，n ) to obtain a mapping matrix Emap, and the expression of the mapping matrix Emap is:

[0034]

[0035] In the formula, Eref(f i，n ) represents the reference value of the vibration energy at the n-th resonant frequency f of the i-th acquisition point in the standard healthy canned fish, F represents the total number of resonant frequencies, and M represents the total number of acquisition points; i，n ;

[0036] Based on the mapping matrix Emap, analyzing the change ratio of the actual vibration response relative to the health standard at a specific position and specific frequency, and determining the abnormal changes inside the canned fish to be tested.

[0037] Preferably, in S5, the calculation formula for the comprehensive canned food freshness score Sfinal is:

[0038]

[0039] Wherein, exp represents an exponential function, a1 and a2 represent the weight coefficients of the total spectrum drift △Stotal and the mapping matrix Emap, respectively, and a1+a2=1, I represents the unit matrix under the ideal standard state, Represents the Frobenius norm, which is used to analyze the overall difference between the energy mapping matrix Emap and the ideal state.

[0040] Preferably, in S5, a secondary comparative evaluation is performed based on the output result of the canned food freshness comprehensive score Sfinal, including:

[0041] When the comprehensive can freshness score Sfinal ≥ the score threshold, it indicates that the canned fish has structural abnormalities and the food in the container is fresh, a can replacement signal is sent to the user terminal and the can is resealed;

[0042] When the canned food freshness comprehensive score Sfinal is less than the score threshold, it indicates that there is an abnormality in the canned fish being tested, and the freshness of the food in the container is also abnormal. The current batch will be marked as a risky batch, and a signal will be issued to prompt sampling and can opening for verification.

[0043] In a second aspect, the present application provides a canned fish quality detection system based on an intelligent algorithm, comprising: a resonance response signal acquisition module, a resonance feature extraction module, a can structure analysis module, an acoustic vibration energy analysis module, and a freshness analysis module;

[0044] The resonance response signal acquisition module sets the tapping collection points according to the set path, taps the fish cans with a standardized tapping device, and sets a sensor group at the collection points to collect the can resonance response signals in real time. At the same time, a fish can detection server is built and the can vibration signals are transmitted to the fish can detection server.

[0045] The resonance feature extraction module extracts features from the resonance response signal in the canned fish detection server to obtain a vibration feature set, and pre-processes the vibration feature set to obtain a standard vibration feature set;

[0046] The can structure analysis module calculates and outputs the main frequency change trajectory F of each tapping collection point based on the standard vibration feature set, and then summarizes and calculates the total spectrum drift ΔStotal based on the main frequency change trajectory F. At the same time, a freshness screening threshold F1 is set and the total spectrum drift ΔStotal is used for preliminary comparative evaluation to determine the structural integrity of the fish can;

[0047] After determining the abnormality of the fish can through preliminary comparison and evaluation, the acoustic-vibration energy analysis module performs acoustic-vibration energy transfer mapping analysis, conducts matrix modeling on the acoustic-vibration energy transfer after multi-frequency micro-tapping, and obtains the mapping matrix Emap;

[0048] The freshness analysis module calculates the comprehensive freshness score Sfinal of the can through combined calculation based on the total spectral drift amount △Stotal and the mapping matrix Emap, and conducts secondary comparison and evaluation based on the output result of the comprehensive freshness score Sfinal of the can to judge the freshness of the fish can.

[0049] The present invention provides a fish can quality detection system and method based on an intelligent algorithm. It has the following beneficial effects:

[0050] (1) By setting standard acquisition points at the top, middle waist and bottom of the fish can, applying a constant impact with a micro electromagnetic impactor, and using a MEMS accelerometer to collect the resonant response signal of the can body after tapping in real time, a fish can detection server is established and real-time data collection and transmission are realized through LoRa wireless transmission. By performing Fourier transform STFT processing on the resonant response signal, extracting the vibration feature set, and using the Z-Score normalization method for unified preprocessing, a normalized vibration feature set is obtained. This process realizes the standardization, efficient collection and dimensionless processing of vibration signal data, provides a high-precision and stable consistency data basis for subsequent intelligent analysis and evaluation, and optimizes the collection quality and detection speed.

[0051] (2) By establishing a main frequency change trajectory model based on the standard vibration feature set, calculating the dynamic change curve of the main frequency of each acquisition point in real time, and further synthesizing the drift amount of each trajectory to output the total spectral drift amount △Stotal. By statistically setting the freshness screening threshold F1 based on the mean value of the drift amount of fresh fish can samples and embedding it in the FPGA module, high-speed preliminary comparison and evaluation of the total spectral drift amount △Stotal and F1 are realized. When the drift amount exceeds the set threshold, the subsequent in-depth detection process is triggered. Through the dynamic modeling of the main frequency micro-drift and the standard threshold screening, this design can not only detect the internal structure abnormality of the fish can at an early stage, but also realize real-time detection on the production line and intelligent pre-screening, significantly improving the detection batch coverage rate and the ability to isolate the initial risk of abnormal cans.

[0052] (3) After the initial screening of abnormal cans, this method further performs acoustic vibration energy transfer mapping analysis, extracts the energy values at different resonant frequencies of each acquisition point under multi-frequency excitation, and constructs an energy normalization mapping matrix Emap. By calculating the overall difference between the mapping matrix Emap and the ideal unit matrix, combined with the total spectrum drift △Stotal, an exponential function mapping method is used to output the comprehensive freshness score Sfinal. Based on the comprehensive freshness score Sfinal, a secondary freshness evaluation standard is set to achieve intelligent and accurate classification of the freshness of canned fish. According to different freshness scores, subsequent processing measures such as replacing the packaging or batch risk marking are implemented respectively. Through the dual-dimensional fusion evaluation of drift trajectory and energy anomaly, the early recognition accuracy of minor quality changes in the internal structure of canned fish is improved, and the efficiency of batch management and quality traceability is optimized. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a schematic diagram of the steps of a canned fish quality detection method based on an intelligent algorithm of the present invention;

[0054] Figure 2 This is a flow chart of a canned fish quality inspection system based on an intelligent algorithm according to the present invention;

[0055] Figure 3 It is the heat map of the energy mapping matrix Emap. DETAILED DESCRIPTION

[0056] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0057] Example 1

[0058] See also Figure 1 and Figure 3 The present invention provides a canned fish quality detection method based on an intelligent algorithm. To achieve the above purpose, the present invention is implemented through the following technical solutions: comprising the following steps:

[0059] S1. Set tapping collection points according to the set path, use a standardized tapping device to tap the fish can, set a sensor group at the collection point to collect the can resonance response signal in real time, build a fish can detection server, and transmit the can vibration signal to the fish can detection server;

[0060] S2. In the fish can detection server, extract the features of the resonance response signal to obtain the vibration feature set, and preprocess the vibration feature set to obtain the standard vibration feature set;

[0061] S3. Based on the standard vibration feature set, calculate and output the main frequency change trajectory F of each tapping acquisition point. Then, based on the main frequency change trajectory F, perform a summary calculation to output the total spectrum drift amount △Stotal. At the same time, set the freshness screening threshold F1 and make a preliminary comparison and evaluation with the total spectrum drift amount △Stotal to judge the structural integrity of the fish can;

[0062] S4. After initially comparing and evaluating and determining that the fish can is abnormal, perform acoustic-vibration energy transfer mapping analysis, matrix model the acoustic-vibration energy transfer after multi-frequency micro-tapping, and obtain the mapping matrix Emap;

[0063] S5. Based on the total spectrum drift amount △Stotal and the mapping matrix Emap, perform a combined calculation to output the comprehensive can freshness score Sfinal, and perform a secondary comparison and evaluation based on the output result of the comprehensive can freshness score Sfinal to judge the freshness of the fish can.

[0064] In this embodiment, the method sets top, middle waist, and bottom collection points on the surface of the fish can according to a set path, uses a standardized tapping device to tap, combines the configured MEMS accelerometer sensor group to collect the resonant response signal of the can in real time, and transmits the data to the fish can detection server through the LoRa wireless network. In the detection server, Fourier transform feature extraction is performed on the collected resonant response signal to obtain a vibration feature set including the initial main frequency, the amplitudes of each harmonic, the angular frequency, the initial phase, and the corresponding vibration energy, and standardization processing is performed by the Z-Score method to form a unified standard vibration feature set. Based on the standard vibration feature set, the main frequency change trajectory F of each collection point is constructed, the total pedigree drift amount △Stotal is calculated and summarized and output, and through a preliminary comparison and evaluation with the set freshness screening threshold F1, the intelligent screening of the structural integrity of the fish can is realized. After the preliminary evaluation determines an abnormality, further perform acoustic-vibration energy transfer mapping analysis, extract the energy distribution of each collection point under multi-frequency tapping, construct an energy normalization mapping matrix Emap, and quantify the local energy abnormal distribution characteristics. Finally, combining the total pedigree drift amount △Stotal and the energy mapping matrix Emap, a comprehensive freshness score Sfinal of the fish can is calculated and output by using a non-linear exponential function, and a secondary comparison and evaluation is performed based on the scoring result to accurately judge the freshness of the fish can and the processing decision. Through the above implementation method, the present invention achieves the purpose of intelligently evaluating the freshness of fish cans in real time, in batches, and non-destructively without damaging the appearance and sealing integrity of the cans. Compared with the prior art, the present invention can, in the early stage of microscopic qualitative change inside the fish can, through the multi-dimensional feature extraction and fusion modeling of the micro-drift of the main frequency trajectory and the abnormal change of the energy distribution pattern, discover potential spoilage risks in advance, improve the screening sensitivity and accuracy, and at the same time avoid the problems of long detection cycle, high destructiveness, and low sample coverage rate in the traditional method.

[0065] Embodiment 2

[0066] Please refer to Figure 1 , specifically: S1 includes S11 and S12;

[0067] S11. Set collection points around the fish can, and use the micro electromagnetic impactor of the standardized tapping device to apply a constant impact to the fish can, tap the fish can, set a sensor group near the collection points, and record the resonant response signal after tapping in real time;

[0068] The sensor group includes MEMS accelerometers;

[0069] The collection points are set at the top, middle waist, and bottom of the fish can;

[0070] S12. Build a canned fish detection server and set up a LoRa wireless transmission network. Wirelessly connect the communication module of the sensor group to the canned fish detection server, and transmit the real-time resonance response signal to the canned fish detection server.

[0071] In this embodiment, the method arranges three standardized knocking collection points at the top, middle waist and bottom of the fish can according to a set path on the surface of the fish can, and uses a miniature electromagnetic impactor as a standardized tapping device to apply constant impact energy at each collection point position to perform standard knocking excitation on the fish can. A sensor group composed of MEMS accelerometers is synchronously fixed near each collection point to collect and record the resonant response signal after knocking in real time. In order to ensure efficient and stable data acquisition and transmission, a fish can detection server is further constructed, and the communication module of the sensor group is wirelessly connected to the detection server using the LoRa wireless transmission network, thereby realizing real-time, delay-free remote transmission and centralized management of the vibration response signal. Through the above-mentioned specific implementation methods, the present invention achieves the technical purpose of applying excitation in a standardized manner, recording the resonant response data in real time and efficiently transmitting it to the detection server for unified processing without destroying or unsealing the fish can. Compared with existing detection technologies, this invention significantly improves the consistency, stability, and scalability of vibration signal acquisition through a combination of standard excitation, multi-point deployment, and wireless transmission. This effectively avoids the instability, complex wiring, and significant signal interference associated with traditional manual tapping, enhancing the overall reliability, real-time performance, and automation of detection. Furthermore, this technology provides a high-quality, efficient data foundation for subsequent analysis of internal structural changes and intelligent freshness assessment of canned fish, comprehensively optimizing the quality inspection and grading capabilities of canned fish in large-scale, assembly-line environments.

[0072] Example 3

[0073] See also Figure 1 , specifically: S2 includes S21 and S22;

[0074] S21. Receiving the resonance response signal of each collection point in real time in the canned fish detection server, and performing feature extraction on the resonance response signal of each collection point in the canned fish detection server to obtain a vibration feature set;

[0075] The vibration feature set includes the initial resonant frequency f of the i-th acquisition point 0，i , the amplitude a of the nth resonance component at the i-th acquisition point i，n , the angular frequency W of the nth resonant component of the i-th acquisition point i，n , the initial phase Xw of the nth resonant component of the i-th acquisition point i，n and the nth resonant frequency f at the ith acquisition point i，n The vibration energy E(f i，n );

[0076] The initial resonance frequency f of the i-th acquisition point 0，i By performing the short-time Fourier transform (STFT) on the resonance response signal, obtaining the STFT spectrogram, extracting the frequency corresponding to the position with the maximum energy in the STFT spectrogram, and setting it as the initial main frequency value f of the i-th acquisition point 0，i ;

[0077] The amplitude a of the i-th acquisition point at the n-th resonance component i，n By relying on the initial main frequency value f of the i-th acquisition point 0，i Setting the search bandwidth, identifying different harmonic peaks within the search bandwidth, such as the main frequency peak, harmonic peak, and sub-harmonic peak, and extracting the amplitudes of each peak in the harmonic peaks to obtain;

[0078] The angular frequency ω of the i-th acquisition point at the n-th resonance component i，n Extracting the n-th resonance frequency f of the i-th acquisition point through the STFT spectrogram i，n , and performing calculation and extraction. The specific calculation formula is: ω i，n = 2πf i，n , where π represents the circumference ratio, with a value of 3.14;

[0079] The n-th resonance frequency f of the i-th acquisition point i，n The vibration energy E(f i，n ) at is extracted by extracting the n-th resonance frequency f of the i-th acquisition point through the STFT spectrogram i，n , and performing integration on a small frequency band of ±0.5 Hz for extraction;

[0080] S22. Based on the obtained vibration feature set, using the Z-Score normalization method, perform normalization processing to eliminate the dimensions of all parameters in the vibration feature set, and convert them into unified dimensionless standard numbers to obtain the standard vibration feature set.

[0081] In this embodiment, the method receives in real time the resonance response signals transmitted from each acquisition point through the canned fish detection server, and performs feature extraction processing on each acquisition point based on the received signals. Specifically, a vibration feature set is extracted and formed. After the vibration feature extraction is completed, based on the Z-Score normalization method, all parameters in the vibration feature set are normalized to eliminate the differences in the original dimensions and transformed into a unified dimensionless standard vibration feature set for subsequent unified modeling and evaluation. Through the above specific implementation manners, the present invention achieves the technical purpose of efficiently and accurately extracting the key information of the resonance response at different detection points of the canned fish and uniformly normalizing the complex physical quantities. Compared with the traditional detection methods based on a single frequency value or low-dimensional features, the present invention can comprehensively capture the multi-order resonance feature changes inside the canned fish and enhance the multi-dimensional characterization ability of the internal tissue state changes. Through the normalization processing, the problem of model training distortion caused by different feature dimensions and magnitude differences is effectively solved, and the consistency, comparability and computer processing friendliness of the feature data are improved. Furthermore, the data quality of the freshness intelligent evaluation in the modeling stage is greatly improved, ensuring the accuracy and stability of the subsequent judgment results, and providing a solid data foundation for realizing the highly reliable intelligent screening of the freshness of canned fish.

[0082] Embodiment 4

[0083] Please refer to Figure 1 , specifically: S3 includes S31, constructing a trajectory model of the main resonance frequency changing with time, and extracting the obtained vibration feature set of each acquisition point, inputting it into the trajectory model, and calculating to output the main frequency change trajectory F of each tapping acquisition point to analyze the actual main frequency change process during the retapping of the canned fish;

[0084] The main frequency change trajectory F is calculated and output through the following trajectory model;

[0085]

[0086] In the formula, F i (t) represents the main frequency change trajectory of the i-th acquisition point at time t, N represents the total number of resonance response signals, sin represents the cosine function, and W i，n (t) represents the angular frequency of the n-th resonance component of the i-th acquisition point at time t;

[0087] f 0，i is a reflection of the basic structure;

[0088] For each order, a i，n ·sin(W i，n (t)+Xw i，n ) is the manifestation of the small-range vibration or energy dispersion of the local tissue;

[0089] The significance of the formula lies in that the main frequency changes slightly over time, with multiple orders of weak resonance superimposed. The subtle changes in the curve reveal the real-time changes in the internal tissue elasticity, density, and rheological properties of the canned food, and can capture the internal qualitative changes at an extremely early stage when there is no obvious gas leakage or appearance deformation.

[0090] S3 also includes S32, traversing the main frequency change trajectories F of all acquisition points, calculating the unified offset amount for the trajectories of all acquisition points, and outputting the total spectrum drift amount △Stotal to measure the comprehensive degree of variation of the overall canned fish from local to whole;

[0091] The total spectrum drift amount △Stotal is calculated and output through the following algorithm formula;

[0092]

[0093] In the formula, M represents the total number of acquisition points, T represents the detection time window, and dt represents the time differential variable.

[0094] S3 also includes S33, calculating the total spectrum drift amount △Stotal of each sample based on fresh canned fish samples, statistically obtaining the mean of these samples and setting it as the freshness screening threshold F1. At the same time, an FPGA module is embedded in the canned fish detection server, and a fast comparison program is set to conduct a preliminary comparison and evaluation of the freshness screening threshold F1 and the total spectrum drift amount △Stotal to judge the stability of the internal structure of the canned fish. The specific evaluation content is as follows;

[0095] When the total spectrum drift amount △Stotal < freshness screening threshold F1, it indicates that the resonance drift is normal, and at this time, it is put into storage and flows;

[0096] When the total spectrum drift amount △Stotal ≥ freshness screening threshold F1, it indicates that the resonance drift is abnormal, and at this time, the acoustic vibration energy transfer mapping analysis is triggered.

[0097] In this embodiment, the method constructs a trajectory model of the main resonant frequency changing over time, and inputs the vibration characteristic parameters of each collection point into the trajectory model for calculation, and outputs the main frequency change trajectory F of each knocking collection point. The trajectory model truly reflects the main frequency micro-drift characteristics of the canned fish during the knocking process in the form of small dynamic changes through the superposition of multi-order resonant components, wherein each order resonant component represents the change in the vibration characteristics of the local tissue in a small range. Through this modeling process, it is possible to capture the slight changes in the elasticity, density and rheological properties of the canned fish in the early stages, and timely reveal the internal qualitative changes that have not yet been manifested, such as cell fluid exudation and local liquefaction. Subsequently, the main frequency change trajectory F of all collection points is traversed, and a unified drift integral calculation is performed on each trajectory, and the total spectrum drift △Stotal is output to comprehensively quantify the degree of internal tissue variation of the canned fish from local to overall. Furthermore, based on the total spectrum drift △Stotal of multiple fresh canned fish samples,

[0098] Stotal data, statistically average and set freshness screening threshold F1, embed an FPGA module in the canned fish detection server and set a fast comparison program. Through the above specific implementation methods, the present invention achieves the technical purpose of using main frequency dynamic trajectory modeling and multi-point unified drift calculation to achieve early and non-destructive identification of microscopic quality changes inside canned fish, and realize intelligent and rapid preliminary screening based on set thresholds. Compared with traditional static detection or empirical judgment methods, the present invention can accurately reflect subtle changes in internal structure through micro-drift trajectory modeling, and through standardized drift integral evaluation, it greatly improves the consistency, sensitivity and automation level of detection.

[0099] Example 5

[0100] See also Figure 1 and Figure 3 Specifically: S4 includes S41, after preliminary comparative evaluation and triggering of acoustic vibration energy transfer mapping analysis, extracting the nth resonance frequency f of each acquisition point at different frequencies at the i-th acquisition point i，n The vibration energy E(f i，n ), and perform matrix modeling on the vibration energy E, obtain the mapping matrix Emap, and analyze the change ratio of the actual vibration response relative to the health standard at a specific position and specific frequency;

[0101] The mapping matrix Emap is calculated and output by the following algorithm formula;

[0102]

[0103] Where, Eref(f i，n ) represents the i-th sampling point at the n-th resonant frequency f of the standard healthy canned fish i，n The reference value of vibration energy at , F represents the total number of resonant frequencies;

[0104] The physical meaning of the formula is that the entire mapping matrix Emap describes the energy distribution pattern and local abnormal response. The energy distribution pattern represents the two-dimensional characteristics of space and frequency, and the local abnormal response represents the local energy decrease or abnormal enhancement. If there are local liquefaction inside the canned food, the high-frequency response will decrease; for stratification, some points will be abnormally enhanced or weakened in the low-frequency band; for loose structure, the overall energy pattern will drift. These abnormal phenomena can be accurately detected through the analysis of the mapping matrix Emap.

[0105] In this embodiment, after initially comparing and evaluating the total lineage drift amount △Stotal to determine that there are potential structural abnormalities in the canned fish, the method further performs acoustic vibration energy transfer mapping analysis to establish an energy normalization mapping matrix Emap. Through the analysis of this mapping matrix Emap, at specific detection points and specific frequencies, the change ratio of the actual vibration response energy relative to the standard healthy state is obtained, so as to identify the characteristics of abnormal local energy distribution. The mapping matrix Emap can not only reveal the changes in the overall energy distribution pattern, but also accurately identify local abnormal responses, such as the decrease in high-frequency response caused by local liquefaction, the abnormal enhancement or weakening of energy in the low-frequency band caused by stratification, and the drift of the overall energy pattern caused by loose structure, etc., so as to achieve refined detection and quantitative analysis of early abnormal changes in the internal tissues of canned fish. Through the above specific implementation manners, the present invention achieves the technical purpose of further accurately identifying local qualitative change regions and internal structure deterioration modes through the spatial frequency distribution characteristics of acoustic vibration energy after the preliminary abnormal screening of the internal structure of canned fish. Compared with the traditional single-frequency point detection or overall sound evaluation method, the present invention models through an energy mapping matrix, integrating the dual dimensions of spatial position and frequency response, greatly improving the detection sensitivity and accuracy of local tiny abnormal changes. Furthermore, the present invention realizes a deeper, non-destructive and quantitative analysis of potential qualitative changes inside canned fish, provides higher-resolution data support for intelligent freshness evaluation, effectively improves the timeliness and reliability of discovering risk batches, and overall improves the level of quality control and safety guarantee of canned fish.

[0106] Example 6

[0107] Please refer to Figure 1 , specifically: S5 includes S51, which performs combined calculation based on the obtained total lineage drift amount △Stotal and mapping matrix Emap, and outputs the comprehensive canned food freshness score Sfinal;

[0108] The comprehensive canned food freshness score Sfinal is calculated and output through the following algorithm formula;

[0109]

[0110] Where exp represents the exponential function, a1 and a2 represent the weight coefficients of the total spectrum drift △Stotal and the mapping matrix Emap, respectively. The specific values are set by the user, and a1+a2=1. I represents the unit matrix, which indicates the ideal standard state. Represents the Frobenius norm, which is used to analyze the overall difference between the energy mapping matrix Emap and the ideal state.

[0111] S5 also includes S52, performing a secondary comparative evaluation based on the output result of the canned food freshness comprehensive score Sfinal to determine the freshness of the canned fish;

[0112] When the canned food freshness comprehensive score Sfinal ≥ the score threshold, it means that the can has structural abnormalities but the freshness is normal. In this case, the canned fish detection server sends a message to the user end to replace the can and reseal it;

[0113] When the comprehensive can freshness score Sfinal is less than the score threshold, it means that there are abnormalities in the can structure and freshness. At this time, the current batch will be marked as a risky batch, and a sampling can opening verification prompt will be given.

[0114] In this embodiment, the method first uses a fusion computational model to perform a unified quantitative assessment of the two types of feature information, based on the total spectrum drift ΔStotal and the acoustic vibration energy mapping matrix Emap, to output a comprehensive freshness score Sfinal for the canned fish. After obtaining the comprehensive freshness score Sfinal, a scoring threshold of 0.8 is set as a secondary freshness comparison benchmark. This secondary determination strategy dynamically implements differentiated post-processing measures based on the comprehensive score results, enabling intelligent grading and traceability control of the canned fish's quality status. Through the above-described specific implementation, the present invention achieves the technical objectives of integrating the main frequency drift characteristics with the energy response deviation characteristics to output a quantifiable comprehensive score, enabling automated freshness grade assessment and disposal decisions. Unlike traditional single-threshold judgment or qualitative identification methods, the present invention employs an exponentially responsive scoring mechanism to amplify the impact of subtle quality anomalies on the score, significantly improving detection sensitivity and classification accuracy. Furthermore, through intelligent secondary determination, a risk level management mechanism is established, providing technical support for batch-level quality traceability, risk warning, and production decision-making for canned fish. Ultimately, this effectively improves the overall quality control, risk isolation, logistics diversion, and intelligent management of canned fish.

[0115] Example 7

[0116] See also Figure 2 , a canned fish quality detection system based on intelligent algorithm, including resonance response signal acquisition module, resonance feature extraction module, can structure analysis module, acoustic vibration energy analysis module and freshness analysis module;

[0117] The resonant response signal acquisition module sets the tapping acquisition points according to the set path, uses a standardized tapping device to tap the canned fish, and sets a sensor group at the acquisition points to collect the resonant response signal of the canned food in real time. At the same time, a canned fish detection server is constructed, and the vibration signal of the canned food is transmitted to the canned fish detection server;

[0118] The resonance feature extraction module extracts features from the resonance response signal in the canned fish detection server to obtain a vibration feature set, and preprocesses the vibration feature set to obtain a standard vibration feature set;

[0119] The canned food structure analysis module calculates and outputs the main frequency change trajectory F of each tapping acquisition point based on the standard vibration feature set, and calculates and outputs the total spectral drift amount △Stotal based on the main frequency change trajectory F. At the same time, a freshness screening threshold F1 is set and compared with the total spectral drift amount △Stotal for preliminary comparative evaluation to judge the structural integrity of the canned fish;

[0120] After the acoustic vibration energy analysis module determines that the canned fish is abnormal through preliminary comparative evaluation, it performs acoustic vibration energy transfer mapping analysis, models the acoustic vibration energy transfer after multi-frequency micro-tapping to obtain a mapping matrix Emap;

[0121] The freshness analysis module calculates and outputs the comprehensive freshness score Sfinal of the canned food based on the total spectral drift amount △Stotal and the mapping matrix Emap, and performs a secondary comparative evaluation based on the output result of the comprehensive freshness score Sfinal of the canned food to judge the freshness of the canned fish.

[0122] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention.

Claims

1. A method for detecting the quality of canned fish based on intelligent algorithms, characterized in that, It includes the following steps: S1. Set multiple tapping acquisition points on the canned fish to be tested respectively. Tap each tapping acquisition point according to a set path, and collect the resonant response signals generated by each tapping acquisition point in real time through a sensor group; S2. Transmit each of the resonant response signals to the canned fish detection server respectively. Extract the features of the resonant response signals through the canned fish detection server to obtain a vibration feature set, and preprocess the vibration feature set to obtain a standard vibration feature set; S3. Based on the standard vibration feature set, calculate the main frequency change trajectory F of each tapping acquisition point, and then perform a summary calculation based on the main frequency change trajectory F to obtain the total spectral drift amount △Stotal. Compare and evaluate the total spectral drift amount △Stotal with the freshness screening threshold F1 preliminarily to judge the structural integrity of the canned fish; S4. When it is determined that the canned fish to be tested is abnormal in the comparison and evaluation, perform acoustic-vibration energy transfer mapping analysis, and perform matrix modeling on the acoustic-vibration energy transfer after multi-frequency micro-tapping to obtain a mapping matrix Emap; S5. Based on the total spectral drift amount △Stotal and the mapping matrix Emap, calculate and output the comprehensive canned fish freshness score Sfinal, and perform a secondary comparison and evaluation based on the output result of the comprehensive canned fish freshness score Sfinal to determine the freshness of the canned fish to be tested.

2. The method for detecting the quality of canned fish based on an intelligent algorithm according to claim 1, characterized in that, In S1, the device for tapping each tapping acquisition point according to a set path is a micro electromagnetic impactor, and the sensor group includes a MEMS accelerometer; The tapping acquisition points are respectively at the top, middle waist and bottom of the canned fish to be tested.

3. The quality inspection method of canned fish based on intelligent algorithm according to claim 1, characterized in that, In S2, extracting the features of the resonant response signals through the canned fish detection server includes: The STFT spectrogram is obtained by performing Fourier transform STFT on the resonant response signal, and the frequency corresponding to the position with the maximum energy in the STFT spectrogram is extracted as the initial main frequency value f of the i-th acquisition point 0，i ; The initial main frequency value f at the i-th acquisition point 0，i Set a search bandwidth thereon, identify different harmonic peaks within the search bandwidth, and take the amplitude of each peak in the harmonic peaks as the amplitude a i，n ; Based on the STFT spectrogram, extract the nth resonant frequency f of the ith acquisition point i，n Perform calculations to obtain the angular frequency W i，n , and the specific calculation formula is: W i，n = 2πf i，n , where π represents the pi Based on the STFT spectrogram, extract the nth resonant frequency f of the ith acquisition point i，n , extract a small frequency band of ±0.5 Hz for integration to obtain the vibration energy E(f i，n ); Among them, the vibration feature set includes the initial resonance frequency f 0，i , amplitude a i，n , angular frequency W i，n , initial phase Xw i，n and vibration energy E(f i，n ).

4. The method for detecting the quality of canned fish based on an intelligent algorithm according to claim 3, wherein, In S3, calculating the main frequency change trajectory F of each tapping acquisition point based on the standard vibration feature set includes: Construct a trajectory model of the main resonant frequency changing with time; Input the vibration feature set obtained from each acquisition point into the trajectory model, and calculate and output the main frequency change trajectory F of each tapping acquisition point. The calculation method of the main frequency change trajectory F is: In the formula, the f 0，i represents the initial resonance frequency, F i (t) represents the main frequency change trajectory of the i-th acquisition point at time t, N represents the total number of resonance response signals, sin represents the cosine function, W i，n (t) represents the angular frequency of the n-th resonance component of the i-th acquisition point at time t.

5. The quality inspection method of canned fish based on intelligent algorithm according to claim 4, characterized in that, In S3, the calculation formula of the total spectral drift amount △Stotal is: In the formula, M represents the total number of acquisition points, T represents the detection time window, and dt represents the time differential variable.

6. The quality inspection method for canned fish based on intelligent algorithm according to claim 1, characterized in that, In S3, comparing and evaluating the total spectral drift amount △Stotal with the freshness screening threshold F1 preliminarily includes: When the total spectral drift amount △Stotal < the freshness screening threshold F1, it means that the resonant drift of the canned fish to be tested is normal; When the total spectral drift amount △Stotal ≥ the freshness screening threshold F1, it means that the resonant drift of the canned fish to be tested is abnormal, and trigger acoustic-vibration energy transfer mapping analysis.

7. A method for detecting the quality of canned fish based on an intelligent algorithm according to claim 6, characterized in that: The triggering of the acoustic-vibration energy transfer mapping analysis includes: Extract the nth resonant frequency f of the ith acquisition point at different frequencies for each tapping acquisition point i，n of the vibration energy E(f i，n ) at that location; For each of the vibration energies E(f i，n ), a matrix model is established to obtain a mapping matrix Emap, and the expression of the mapping matrix Emap is as follows: wherein, the Eref(f i，n ) represents the vibration energy reference value of the i-th acquisition point at the n-th resonance frequency f i，n of the standard healthy fish can, the F represents the total number of resonance frequencies, and the M represents the total number of acquisition points; Based on the mapping matrix Emap, analyze the change ratio of the actual vibration response relative to the healthy standard at a specific position and specific frequency to determine the abnormal change inside the canned fish to be tested.

8. A method for detecting the quality of canned fish based on an intelligent algorithm according to claim 1, characterized in that, In the step S5, the calculation formula of the comprehensive score Sfinal of the can freshness is: Wherein, exp represents the exponential function, a1 and a2 respectively represent the weight coefficients of the total pedigree drift amount △Stotal and the mapping matrix Emap, and a1 + a2 = 1, I represents the identity matrix under the ideal standard state, and represents the Frobenius norm, which is used to analyze the overall difference between the energy mapping matrix Emap and the ideal state.

9. A method for detecting the quality of canned fish based on an intelligent algorithm according to claim 1, characterized in that, In the step S5, a secondary comparison and evaluation is performed based on the output result of the comprehensive score Sfinal of the can freshness, including: When the comprehensive score Sfinal of the can freshness ≥ the score threshold, it indicates that the fish can to be tested has a structural abnormality and the food freshness in the container is normal. Then, a signal for replacing the can is sent to the user terminal and resealing is performed. When the comprehensive score Sfinal of the can freshness < the score threshold, it indicates that the fish can to be tested has an abnormality and the food freshness in the container is also abnormal. Then, the current production batch is marked as a risk batch and a signal for prompting sampling and opening the can for verification is sent.

10. A quality inspection system for canned fish based on intelligent algorithms, which is used to execute the method described in any one of claims 1-9, characterized in that, Including: A resonant response signal acquisition module, a resonant feature extraction module, a can structure analysis module, an acoustic vibration energy analysis module, and a freshness analysis module; The resonant response signal acquisition module sets the tapping acquisition points according to a set path, uses a standardized light tapping device to tap the fish can, sets a sensor group at the acquisition points to collect the can resonant response signal in real time, constructs a fish can detection server, and transmits the can vibration signal to the fish can detection server. The resonant feature extraction module extracts features from the resonant response signal in the fish can detection server to obtain a vibration feature set, and preprocesses the vibration feature set to obtain a standard vibration feature set. The can structure analysis module calculates and outputs the main frequency change trajectory F of each tapping acquisition point based on the standard vibration feature set, then performs a summary calculation based on the main frequency change trajectory F to output the total spectrum drift amount △Stotal, and sets a freshness screening threshold F1 to perform a preliminary comparison and evaluation with the total spectrum drift amount △Stotal to judge the structural integrity of the fish can. After it is determined that the fish can is abnormal in the preliminary comparison and evaluation, the acoustic vibration energy analysis module performs an acoustic vibration energy transfer mapping analysis, performs matrix modeling on the acoustic vibration energy transfer after multi-frequency micro-tapping, and obtains a mapping matrix Emap. The freshness analysis module calculates and outputs the comprehensive score Sfinal of the can freshness by combining the total spectrum drift amount △Stotal and the mapping matrix Emap, and performs a secondary comparison and evaluation based on the output result of the comprehensive score Sfinal of the can freshness to judge the freshness of the fish can.

Images