A method and system for detecting the maturity of agricultural products

Abstract

The present invention relates to the field of maturity detection technology, and more particularly to a method and system for detecting the maturity of agricultural products. The method comprises the following steps: scanning pears at multiple angles using an ultrasonic detector to obtain frequency domain data of reflected acoustic waves and perform bandwidth variation analysis to obtain regional acoustic bandwidth variation data; then, integrating the high- and low-frequency energy ratios to obtain data on the structural differences in high-frequency energy attenuation between regions; then, performing a simulation assessment of the incremental viscosity of the fruit pulp, and further assessing the softness and toughness of the fruit pulp; and finally, assessing the maturity of the pears based on the results of the flesh softness and toughness assessment. The present invention further improves maturity detection technology by optimizing it.

Description

Technical Field

[0001] The present invention relates to the technical field of maturity detection, and in particular to a method and system for detecting the maturity of agricultural products. Background Art

[0002] Ultrasonic testing technology, as a nondestructive physical testing method, offers advantages such as high sensitivity, ease of operation, and immunity to environmental influences. Ultrasonic testing of the internal structure of agricultural products can reveal physical properties such as tissue density, hardness, and elasticity, thereby inferring their maturity. In particular, analysis of ultrasonic wave velocity, attenuation, and reflection data can accurately reveal the maturity of fruits or agricultural products. Furthermore, frequency analysis based on sound waves can provide valuable information on the internal distribution and structural changes of fruit pulp, further optimizing the accuracy and timeliness of maturity testing. However, a traditional method for detecting the maturity of agricultural products suffers from low accuracy in analyzing sound wave attenuation, resulting in large errors in detecting the softness and toughness of the flesh and low maturity accuracy. Summary of the Invention

[0003] Based on this, it is necessary to provide a method and system for detecting the maturity of agricultural products to solve at least one of the above technical problems.

[0004] To achieve the above object, a method for detecting the maturity of agricultural products is provided, the method comprising the following steps:

[0005] Step S1: Scanning the pear at multiple angles using an ultrasonic detector, then performing frequency domain analysis on the regional division to generate regional frequency domain data of reflected sound waves; performing sound wave bandwidth variation analysis on the regional frequency domain data of reflected sound waves to obtain regional sound wave bandwidth variation data;

[0006] Step S2: performing high-low frequency energy ratio integration processing between different regions on the regional acoustic wave bandwidth change data to obtain regional high-low frequency energy ratio integration data; performing high-frequency energy attenuation structure analysis between different regions based on the regional high-low frequency energy ratio integration data to obtain attenuation structure localization difference data;

[0007] Step S3: performing a simulation evaluation of the pulp viscosity increment based on the attenuation structure localization difference data to obtain pulp viscosity increment data; performing a pulp softness and toughness evaluation based on the pulp viscosity increment data to obtain pulp softness and toughness evaluation data;

[0008] Step S4: evaluating the maturity of the pears according to the pulp softness and toughness evaluation data, thereby obtaining the pear maturity evaluation data.

[0009] Preferably, step S1 includes the following steps:

[0010] Step S11: Scan the pear at multiple angles using an ultrasonic detector to obtain multi-angle reflected sound waves from the pear;

[0011] Step S12: performing noise filtering on the pear multi-angle reflected sound waves to obtain the pear multi-angle reflected filtered sound waves;

[0012] Step S13: performing frequency domain analysis on the pear multi-angle reflected filtered sound waves to generate frequency domain data of the reflected sound waves;

[0013] Step S14: performing acoustic wave bandwidth change analysis on the regional frequency domain data of the reflected acoustic wave to obtain regional acoustic wave bandwidth change data.

[0014] Preferably, step S2 includes the following steps:

[0015] Step S21: calculating the sound wave frequency change rate ratios between different regions based on the regional sound wave bandwidth change data to obtain the regional sound wave frequency change rate ratios;

[0016] Step S22: performing high-low frequency energy ratio integration processing on the regional sound wave bandwidth change data between different regions according to the regional sound wave frequency change rate ratio to obtain regional high-low frequency energy ratio integration data;

[0017] Step S23: performing high-frequency energy attenuation structure analysis between different regions based on the regional high-low frequency energy ratio integral data to obtain regional high-frequency energy attenuation structure data;

[0018] Step S24: performing multi-scale localized difference analysis between different regions on the regional high-frequency energy attenuation structure data to obtain attenuation structure localized difference data.

[0019] Preferably, step S22 includes the following steps:

[0020] Step S221: performing high-low frequency change morphology identification of sound waves between different regions based on the regional sound wave frequency change rate ratio and the regional sound wave bandwidth change data to obtain high-low frequency change morphology data of sound waves between different regions;

[0021] Step S222: performing high-low frequency discrete entropy regression analysis on the high-low frequency variation morphological data of the sound waves between different regions to obtain high-low frequency discrete entropy regression data;

[0022] Step S223: performing nonlinear fluctuation polynomial interpolation on the high-low frequency variation morphological data of the sound waves between different regions based on the high-low frequency discrete entropy value regression data to obtain high-low frequency nonlinear fluctuation interpolation data;

[0023] Step S224: calculating the variance of the segmented fluctuation slope growth rate on the high-low frequency nonlinear fluctuation interpolation data to obtain the variance of the high-low frequency segmented fluctuation slope growth rate;

[0024] Step S225: performing high-low frequency energy ratio integration processing on the variance of the high-low frequency segmented fluctuation slope growth rate between different regions according to the Romberg integration method to obtain regional high-low frequency energy ratio integration data.

[0025] Preferably, step S23 includes the following steps:

[0026] Step S231: identifying high-frequency energy attenuation fluctuation frequency bands between different regions based on regional high-low frequency energy ratio integral data, and obtaining high-frequency energy attenuation fluctuation frequency bands between different regions;

[0027] Step S232: performing attenuation concentration density analysis on the high-frequency energy attenuation fluctuations between different regions to obtain the high-frequency energy attenuation concentration density between different regions;

[0028] Step S233: performing non-uniform complex wavenumber gradient differential processing on the high-frequency energy attenuation fluctuation frequency band based on the high-frequency energy attenuation concentration density to obtain high-frequency attenuation complex wavenumber gradient differential data;

[0029] Step S234: calculating the attenuation acceleration variance between different regions of the high-frequency energy attenuation fluctuation frequency band according to the high-frequency attenuation complex wave number gradient differential data to obtain the attenuation acceleration variance;

[0030] Step S235: performing high-frequency energy attenuation structure analysis between different regions based on the high-frequency attenuation complex wavenumber gradient differential data and the attenuation acceleration variance to obtain regional high-frequency energy attenuation structure data.

[0031] Preferably, step S233 includes the following steps:

[0032] Based on the high-frequency energy attenuation concentration density, the frequency decay node non-uniformity analysis of the concentrated center waveform of the high-frequency energy attenuation fluctuation frequency band is performed to obtain the frequency decay node non-uniformity data;

[0033] According to the non-uniform data of the frequency decay node, the high-frequency energy attenuation fluctuation frequency band is processed by complex spectrum construction to obtain a non-uniformly distributed complex spectrum matrix;

[0034] Perform first-order complex wave number differentiation on the non-uniformly distributed complex spectrum matrix to obtain complex wave number gradient change data;

[0035] The complex domain phase singular spectrum entropy is calculated for the complex wave number gradient change data to obtain the complex domain phase singular spectrum entropy;

[0036] The high-frequency attenuated complex wave number gradient differential data are obtained by performing non-uniform complex wave gradient differential processing according to the complex domain phase singular spectrum entropy.

[0037] Preferably, step S3 includes the following steps:

[0038] Step S31: identifying the acoustic wave scattering enhancement gradients between different regions based on the attenuation structure localization difference data to obtain the acoustic wave scattering enhancement gradients between different regions;

[0039] Step S32: performing a simulation evaluation of the pear intercellular expansion based on the acoustic wave scattering enhancement gradients between different regions to obtain intercellular expansion data;

[0040] Step S33: performing a simulation evaluation of the pulp viscosity increment based on the attenuation structure localization difference data to obtain pulp viscosity increment data;

[0041] Step S34: Evaluate the softness and toughness of the pulp based on the intercellular space expansion data and the pulp viscosity increment data to obtain pulp softness and toughness evaluation data.

[0042] Preferably, step S33 includes the following steps:

[0043] Step S331: performing pulp-related absorption attenuation measurement mapping based on the attenuation structure localization difference data to obtain pulp-related absorption attenuation measurement data;

[0044] Step S332: Deducing the regional pulp density difference interval based on the pulp-related absorption attenuation measurement data to obtain the regional pulp density difference interval;

[0045] Step S333: performing pulp shear viscosity increment fitting based on the pulp-associated absorption attenuation measurement data and the regional pulp density difference interval to obtain pulp shear viscosity increment fitting data;

[0046] Step S334: performing a pulp volume viscosity ratio analysis based on the pulp-associated absorption attenuation measurement data, the pulp shear viscosity increment fitting data, and the pulp density difference interval to obtain the pulp volume viscosity ratio;

[0047] Step S335: performing a simulation evaluation of the pulp viscosity increment based on the pulp shear viscosity increment fitting data and the pulp volume viscosity ratio to obtain pulp viscosity increment data.

[0048] Preferably, the present invention further provides an agricultural product maturity detection system for executing the agricultural product maturity detection method described above, the agricultural product maturity detection system comprising:

[0049] The acoustic wave bandwidth variation analysis module is used to scan the pears at multiple angles using an ultrasonic detector, and then perform frequency domain analysis on the regional division to generate regional frequency domain data of the reflected acoustic wave; the acoustic wave bandwidth variation analysis is performed on the regional frequency domain data of the reflected acoustic wave to obtain regional acoustic wave bandwidth variation data;

[0050] The high-frequency energy attenuation structure analysis module is used to integrate the high-to-low-frequency energy ratio between different regions based on the regional acoustic wave bandwidth change data to obtain regional high-to-low-frequency energy ratio integral data; based on the regional high-to-low-frequency energy ratio integral data, the high-frequency energy attenuation structure analysis between different regions is performed to obtain localized difference data of the attenuation structure;

[0051] The pulp softness and toughness evaluation module is used to simulate and evaluate the pulp viscosity increment based on the attenuation structure localization difference data to obtain the pulp viscosity increment data; and to evaluate the pulp softness and toughness based on the pulp viscosity increment data to obtain the pulp softness and toughness evaluation data;

[0052] The maturity evaluation module is used to evaluate the maturity of pears based on the flesh softness and toughness evaluation data, thereby obtaining the pear maturity evaluation data.

[0053] The present invention has the beneficial effect of using an ultrasonic detector to perform multi-angle scanning on pears, generate regional frequency domain data of reflected sound waves, and perform sound wave bandwidth variation analysis. The beneficial effect of this step is that through the multi-angle data collection of ultrasonic scanning, a comprehensive understanding of the internal physical structure and density distribution of the pear can be obtained, and detailed frequency domain data can be generated to reflect the sound wave propagation characteristics of each region of the fruit. The analysis of bandwidth variation further reveals the changes in the internal distribution of the fruit, can accurately reflect the tissue characteristics of the pear during the ripening process, and provides a high-precision non-destructive testing method. The regional sound wave bandwidth variation data is processed by high-low frequency energy ratio integration to obtain regional high-low frequency energy ratio integration data, and based on this data, high-frequency energy attenuation structure analysis is performed. This processing can reveal the differences in sound wave attenuation between different regions, thereby reflecting the structural differences of the fruit. The attenuation of high-frequency energy is often closely related to the maturity, density and physical changes of the flesh. Through this localized difference analysis of the attenuation structure, it is helpful to accurately judge the changes in different regions during the fruit ripening process, further improving the accuracy and reliability of the detection. Based on the localized difference data of the attenuation structure, an incremental simulation evaluation of the pulp viscosity is performed to obtain incremental pulp viscosity data, and the pulp softness and toughness are further evaluated. The beneficial effect of this step is that the incremental viscosity simulation can carefully capture the viscosity changes of the pulp tissue and infer its softness and toughness characteristics. The softness and toughness of the pulp directly affects the taste and edible value of the fruit. Therefore, an accurate assessment of softness and toughness can provide a more quantitative and objective basis for judging the maturity of the fruit. Based on the pulp softness and toughness assessment data, the pear maturity is evaluated and the pear maturity assessment data is obtained. The beneficial effect of this step is that through the precise early assessment of the pulp softness and toughness, combined with a comprehensive judgment of maturity, the true maturity state of the pear can be determined. This method is not only non-destructive and efficient, but also avoids the errors introduced by traditional methods, making the maturity assessment results more accurate and reliable. Therefore, the present invention is an optimization treatment of a traditional method for detecting the maturity of agricultural products, which solves the problem that the traditional method for detecting the maturity of agricultural products has low accuracy of sound wave attenuation analysis, resulting in large errors in detecting the softness and toughness inside the pulp and low accuracy in maturity judgment. It improves the accuracy of sound wave attenuation analysis, reduces the error in detecting the softness and toughness inside the pulp, and improves the accuracy of maturity judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 The figure is a flowchart of a method for detecting the maturity of agricultural products;

[0055] Figure 2 for Figure 1 Detailed implementation steps of step S3 in FIG.

[0056] Figure 3 for Figure 2Detailed implementation steps of step S33 are shown in the flowchart. DETAILED DESCRIPTION

[0057] See also Figures 1 to 3 , a method for detecting the maturity of agricultural products, the method comprising the following steps:

[0058] Step S1: Scanning the pear at multiple angles using an ultrasonic detector, then performing frequency domain analysis on the regional division to generate regional frequency domain data of reflected sound waves; performing sound wave bandwidth variation analysis on the regional frequency domain data of reflected sound waves to obtain regional sound wave bandwidth variation data;

[0059] Step S2: performing high-low frequency energy ratio integration processing between different regions on the regional acoustic wave bandwidth change data to obtain regional high-low frequency energy ratio integration data; performing high-frequency energy attenuation structure analysis between different regions based on the regional high-low frequency energy ratio integration data to obtain attenuation structure localization difference data;

[0060] Step S3: performing a simulation evaluation of the pulp viscosity increment based on the attenuation structure localization difference data to obtain pulp viscosity increment data; performing a pulp softness and toughness evaluation based on the pulp viscosity increment data to obtain pulp softness and toughness evaluation data;

[0061] Step S4: evaluating the maturity of the pears according to the pulp softness and toughness evaluation data, thereby obtaining the pear maturity evaluation data.

[0062] In the embodiment of the present invention, reference Figure 1 The above is a schematic flow chart of the steps of a method for detecting the maturity of agricultural products according to the present invention. In this example, the method for detecting the maturity of agricultural products includes the following steps:

[0063] Step S1: Scanning the pear at multiple angles using an ultrasonic detector, then performing frequency domain analysis on the regional division to generate regional frequency domain data of reflected sound waves; performing sound wave bandwidth variation analysis on the regional frequency domain data of reflected sound waves to obtain regional sound wave bandwidth variation data;

[0064] In this embodiment of the present invention, multiple ultrasonic detection heads are evenly fixed to the inner wall of the detection cavity. The ultrasonic transmission frequency is set to 2.25 MHz, the transmission power is 10 W, and the transmission angle is set to ±45°, with a step-by-step scanning interval of 5°. This ensures that reflected sound waves from multiple spatial angles on the surface and inside of the pear are captured. The collected raw echo signals are recorded at a sampling rate of 500 MS / s. The echo signals are processed by time window interception according to the angle number through timing synchronization control. The raw echo signals are filtered using a Butterworth bandpass filter with a passband range of 1.5 MHz to 3.0 MHz to eliminate interference from environmental noise and equipment stray frequencies. In the filtered sound wave data, the reflected waveform of each angle data is Fourier transformed to obtain the frequency domain signal spectrum of each area. The pear was divided into 20 spherical regions according to its geometric spherical coordinate system. The reflected frequency domain energy density of each region was then divided into four frequency bands: 1.5–1.8 MHz, 1.8–2.2 MHz, 2.2–2.5 MHz, and 2.5–3.0 MHz. The bandwidth of the energy amplitude in each frequency band was then calculated. By calculating the change in peak energy density within each bandwidth, the regional acoustic bandwidth variation characteristics were extracted. Ultimately, a matrix was formed, recording the energy fluctuation data corresponding to the center frequency of each frequency band in each region, forming the "regional acoustic bandwidth variation data."

[0065] In another embodiment, an ultrasonic detector performs multi-angle scanning on a pear. First, a pulsed ultrasonic transmitter with a frequency of 1 MHz is used, scanning 360 degrees across the pear surface with a step angle of 5 degrees to ensure that the acoustic signal covers all directions of the pear. The collected raw reflected acoustic signal is processed through a bandpass filter to remove noise components with frequencies below 0.5 MHz and above 2 MHz, thereby improving the signal-to-noise ratio. Subsequently, a short-time Fourier transform is used to perform frequency domain analysis on the filtered reflected acoustic signal. The pear surface is divided into several fixed-size regions (e.g., each region covering 10 square millimeters). The acoustic signal within each region is spectrally calculated to obtain frequency domain data for the regional reflected acoustic wave. For each region's spectral data, a bandwidth measurement algorithm is used to calculate the 3dB bandwidth variation of the acoustic signal. Specifically, the bandwidth variation data for each region is obtained by measuring the frequency range corresponding to a 3dB drop in energy in the spectrum. This process was implemented by writing a signal processing program. Parameter settings included a 5MHz sampling rate, a Hanning window function with a window length of 1024 points and a 50% overlap ratio to ensure balanced time and frequency resolution in the frequency domain analysis. These steps enabled the regional division of the frequency domain analysis and the extraction of bandwidth variations for the multi-angle reflected sound waves from the pear.

[0066] Step S2: performing high-low frequency energy ratio integration processing between different regions on the regional acoustic wave bandwidth change data to obtain regional high-low frequency energy ratio integration data; performing high-frequency energy attenuation structure analysis between different regions based on the regional high-low frequency energy ratio integration data to obtain attenuation structure localization difference data;

[0067] In this embodiment of the present invention, the "regional acoustic bandwidth variation data" obtained above is used to calculate inter-regional ratios. The integration method is used to calculate the ratio of the total energy between the high-frequency band (2.2–3.0 MHz) and the low-frequency band (1.5–2.2 MHz) for each region. The integration is performed using the numerical trapezoidal method, dividing the frequency axis into 100 equally spaced sampling points. After integration, the ratio is calculated for each segment to form "regional high-to-low-frequency energy ratio integral data." Furthermore, to determine the energy distribution differences within the pear's internal tissue structure, a high-frequency energy attenuation structure analysis method is employed. By calculating the gradient of the high-frequency energy integral difference between adjacent regions, an inter-regional energy difference gradient field is constructed. The spatial gradient map is then used to perform convolution enhancement on the gradient intensity in each direction using a 3×3 Sobel kernel. The directional features with the most pronounced energy attenuation are extracted. Combined with the regional spatial distance function, normalized mapping is performed on the different gradient variation directions to produce "attenuation structure localization difference data." This data can be specifically described as a 20-dimensional vector, with each dimension representing the high-frequency attenuation gradient value relative to the reference region.

[0068] In another embodiment, for the regional acoustic wave bandwidth change data obtained in step S1, the acoustic wave frequency change rate ratio between different regions is first calculated. The specific method is to perform differential processing on the bandwidth change data of adjacent regions to obtain the frequency change rate, and then calculate the rate ratio. Based on the rate ratio, a high-low frequency energy ratio integral processing is adopted. Specifically, the bandwidth change data is divided into a high frequency band (1.2 MHz to 2 MHz) and a low frequency band (0.5 MHz to 1.2 MHz), and the energy integral in each frequency band is calculated respectively. The integration adopts the Romberg integration method, the integration step is set to 0.01 MHz, and the integration interval is strictly limited to ensure the integration accuracy. The integration result is used to construct the regional high-low frequency energy ratio integral data. Subsequently, based on the data, a high-frequency energy attenuation structure analysis is performed. A multi-scale localization analysis method is adopted. Specifically, high-frequency energy attenuation characteristics at different scales are extracted by wavelet transform. The scale range is set to 1 to 5 and the resolution is 0.1. The local attenuation fluctuation frequency band is extracted, and the attenuation clustering density analysis is performed on the extracted frequency band. The spatial distribution density of the attenuation fluctuation is calculated using the kernel density estimation method. The kernel function uses a Gaussian kernel and the bandwidth parameter is set to 0.05. Combined with the non-uniform complex wavenumber gradient differential processing, complex spectrum construction and first-order complex wavenumber differential are adopted, and the complex domain phase singular spectrum entropy is calculated by the phase singular spectrum entropy algorithm. The parameter settings include a phase window length of 256 points and a step size of 64 points. Finally, the attenuation acceleration variance is calculated to obtain the localized difference data of the attenuation structure between different regions.

[0069] Step S3: performing a simulation evaluation of the pulp viscosity increment based on the attenuation structure localization difference data to obtain pulp viscosity increment data; performing a pulp softness and toughness evaluation based on the pulp viscosity increment data to obtain pulp softness and toughness evaluation data;

[0070] In this embodiment, a physical simulation method based on quantified structural differences is used to assess the viscosity trend of fruit pulp. Using "localized attenuation structure difference data" as input, a comparison table of existing experimental data is used to establish a correspondence between the damping coefficient and the intra-tissue slip coefficient of pear pulp under specific acoustic wave attenuation structure differences. The damping coefficient represents the shear response delay characteristic of the fruit pulp. The viscosity increment of the fruit pulp is derived using a linear fit weighted difference, expressed in Pa·s. In actual testing, if the high-frequency energy attenuation gradient of a specific test area exceeds a threshold of 0.35 (normalized units), the tissue softening trend in that area is determined to be significant. The viscosity increment data for each area are averaged and combined with the inter-area variance to derive the viscosity increment data for the entire fruit pulp. Based on a pre-defined softness-toughness mapping table, the softness-toughness level is divided into five levels, corresponding to extremely hard, relatively hard, moderate, relatively soft, and extremely soft, respectively. The softness-toughness evaluation is output as a two-dimensional matrix, including the softness-toughness level and the mean square error amplitude, forming the "fruit softness-toughness evaluation data."

[0071] In another embodiment, based on the localized attenuation structure difference data obtained in step S2, acoustic scattering enhancement gradient identification is first performed. A gradient operator is used to calculate the spatial gradient of the localized attenuation difference data using a Sobel operator with a window size of 3×3. The acoustic scattering enhancement gradients between different regions are calculated. Based on this gradient, the expansion of pear intercellular spaces is simulated. Specifically, a model for intercellular space expansion is established. Assuming an initial intercellular space width of 5 microns, the gradient value is used as the expansion factor to calculate the change in intercellular space width after expansion. Subsequently, based on the localized attenuation structure difference data, a simulation and assessment of the pulp viscosity increment is performed. The pulp absorption attenuation coefficient is calculated using a correlation absorption attenuation metrological mapping method. The coefficient range is set between 0.1 and 0.5. Combined with the regional pulp density difference interval deduction, the density range is set between 0.8 and 1.2 g / cm3. A shear viscosity increment fitting method is used, using the least squares method to fit the pulp shear viscosity increment. The fitting parameters include the shear modulus and the viscosity coefficient. Combined with the analysis of the pulp volume viscosity ratio, the proportion of viscous components in the pulp volume was calculated, with the ratio range set between 0.3 and 0.7. Finally, the pulp viscosity increment data was calculated based on the shear viscosity increment fitting data and the volume viscosity ratio. Based on this pulp viscosity increment data, the pulp softness and toughness were assessed using a weighted average method, with the weight coefficient determined based on the viscosity increment and intercellular space expansion data. This yielded the pulp softness and toughness assessment data.

[0072] Step S4: evaluating the maturity of the pears according to the pulp softness and toughness evaluation data, thereby obtaining the pear maturity evaluation data.

[0073] In an embodiment of the present invention, the "fruit flesh softness and toughness evaluation data" is subjected to maturity archiving and comparison processing. This processing is based on a standard pear maturity data table, which covers the softness and toughness distribution ranges under various maturity levels. During the detection implementation, the softness and toughness mean value and its standard deviation of the current sample are compared with the standard table, and the closest maturity level is determined by the minimum distance matching principle. The "pear maturity evaluation data" is output based on the judgment condition that the softness and toughness mean value falls into the maturity level distribution range and satisfies the standard deviation of no more than 0.1, and finally the maturity level of the pear at the current detection moment is obtained.

[0074] In another embodiment, pear maturity is assessed based on the flesh toughness assessment data obtained in step S3. The toughness assessment data is first mapped to maturity grade intervals, which are divided into unripe (toughness value less than 0.4), mature (toughness value between 0.4 and 0.7), and overripe (toughness value greater than 0.7). A threshold determination method is used to determine the maturity grade of the pear based on the numerical value of the toughness assessment data. Specifically, the toughness assessment data is normalized within a range of 0 to 1 to ensure comparability between different samples. The maturity assessment result is output as a numerical value and grade, with the numerical precision retained to three decimal places, ultimately obtaining the pear maturity assessment data.

[0075] Step S1 includes the following steps:

[0076] Step S11: Scan the pear at multiple angles using an ultrasonic detector to obtain multi-angle reflected sound waves from the pear;

[0077] Step S12: performing noise filtering on the pear multi-angle reflected sound waves to obtain the pear multi-angle reflected filtered sound waves;

[0078] Step S13: performing frequency domain analysis on the pear multi-angle reflected filtered sound waves to generate frequency domain data of the reflected sound waves;

[0079] Step S14: performing acoustic wave bandwidth change analysis on the regional frequency domain data of the reflected acoustic wave to obtain regional acoustic wave bandwidth change data.

[0080] In an embodiment of the present invention, an ultrasonic detector with a center frequency of 2.25 MHz and a transmitting pulse width of 0.3 microseconds is selected to collect multi-angle reflected sound waves from pears. An ultrasonic transmitting-receiving integrated transducer is arranged inside the detection cavity and evenly distributed spherically around the outside of the pear. The transmitting transducer sequentially transmits ultrasonic pulses in a progressive manner of 15° polar angle and azimuth angle. The receiving transducer synchronously receives the sound wave signals reflected from the inside or surface of the pear. A channel synchronization trigger is used to accurately control the working timing of each transducer to ensure that the reflected signals at all angles cover the entire three-dimensional structural area of ​​the pear. An analog-to-digital converter is used to collect the received signals at a sampling rate of 500 MS / s, and the collected data numbers are marked with their corresponding transmitting angles and receiving angle numbers, thereby obtaining complete multi-angle reflected sound wave data of the pear. The noise filtering process was performed on the multi-angle reflected acoustic wave data obtained from pears. First, a fourth-order Butterworth bandpass filter was used to perform preliminary frequency domain screening on all reflected acoustic wave data. The filter cutoff frequencies were set to 1.8 MHz and 2.7 MHz, respectively, to remove low-frequency environmental background noise and high-frequency electromagnetic interference signals outside the sensor operating frequency. After preliminary filtering, wavelet denoising technology was further used. The Daubechies 4 wavelet function was selected for 5-layer decomposition. The soft threshold compression strategy was used to compress the noise of the high-frequency component coefficients, and the filtered reflected acoustic wave signal was reconstructed. The processed signal retained the main reflection waveform structure while effectively reducing background interference, making the subsequent frequency domain analysis more accurate. The frequency domain analysis of the multi-angle reflected filtered acoustic wave data of pears was performed by regional division. First, the pear surface was divided into 20 equal-area sub-regions according to the spherical coordinate system. The division method was to set the polar angle and azimuth angle to five equal parts to form a spherical grid mapping. The mapping result was used to group each reflected signal according to the region corresponding to the incident and receiving paths. Then, the filtered reflected acoustic wave signal in each region was converted into the frequency domain using fast Fourier transform (FFT). The frequency domain transformation length was set to 2048 points, and the window function used a Hamming window to reduce spectrum leakage. The amplitude spectrum of the transformed frequency domain signal was extracted to form a spectrum data set corresponding to each sub-region. The main frequency peak, energy density distribution range and frequency energy concentration parameters were marked in each spectrum, thereby constructing the frequency domain data of the reflected acoustic wave region, which was used to characterize the frequency domain distribution differences of the acoustic wave reflection characteristics in different spatial regions.Based on the frequency domain data of the reflected sound wave region, the sound wave bandwidth changes in each region are analyzed. First, in each regional spectrum, the part with energy amplitude greater than 80% of the maximum value is selected as the main energy bandwidth interval, and the difference between the starting frequency and the ending frequency of this interval is calculated as the effective bandwidth value. Further, by statistically analyzing the changes in the effective bandwidth size of different regions, its distribution trend in the overall structure is analyzed. In order to eliminate the interference of small fluctuation errors on the judgment of the overall bandwidth change, the 5-point sliding average method is used to smooth the bandwidth data of each region. At the same time, the bandwidth change gradient matrix is ​​constructed by calculating the bandwidth difference gradient between regions. This matrix serves as the key input data for characterizing structural heterogeneity and sound wave propagation heterogeneity, forming the final regional sound wave bandwidth change data.

[0081] Step S2 includes the following steps:

[0082] Step S21: calculating the sound wave frequency change rate ratios between different regions based on the regional sound wave bandwidth change data to obtain the regional sound wave frequency change rate ratios;

[0083] Step S22: performing high-low frequency energy ratio integration processing on the regional sound wave bandwidth change data between different regions according to the regional sound wave frequency change rate ratio to obtain regional high-low frequency energy ratio integration data;

[0084] Step S23: performing high-frequency energy attenuation structure analysis between different regions based on the regional high-low frequency energy ratio integral data to obtain regional high-frequency energy attenuation structure data;

[0085] Step S24: performing multi-scale localized difference analysis between different regions on the regional high-frequency energy attenuation structure data to obtain attenuation structure localized difference data.

[0086] In the embodiment of the present invention, based on the regional sound wave bandwidth change data obtained in step S14, the frequency bandwidth edge point change trend between different regions is quantitatively analyzed and processed. First, the bandwidth value of each region is arranged in order according to the regional number, and the center value of the bandwidth upper boundary frequency and the lower boundary frequency of each region is extracted to form a regional representative frequency sequence. For the representative frequency data in adjacent numbered regions, the frequency change rate sequence is formed by dividing the representative frequency of the previous region by the difference, which is recorded as , and then construct a frequency change rate ratio matrix. The calculation method is that the ratio of the frequency change rates of each pair of adjacent regions constitutes a two-dimensional matrix. The diagonal elements in the matrix represent that the frequency change rate of each region itself is 1, and the non-diagonal elements are used to characterize the relative difference in the frequency change rates between regions. The rate ratio matrix is ​​further normalized by the mean to map it between 0 and 1, forming the regional sound wave frequency change rate ratio data. The regional acoustic wave frequency change rate ratio data obtained in step S21 is used to perform high- and low-frequency energy integration processing on the regional acoustic wave bandwidth change data. First, in the bandwidth change data, the frequency range is clearly divided into a low frequency band of 1.5 MHz to 2.2 MHz and a high frequency band of 2.2 MHz to 3.0 MHz. The spectrum density function curves of each region within the two frequency bands are numerically integrated respectively. The integration method adopts the Simpson integration method. By dividing the frequency band into 50 small intervals, the spectrum amplitude in each small interval is weighted and accumulated to obtain the high frequency band energy E_high and the low frequency band energy E_low respectively. Then, E_high is divided by E_low to form the high- and low frequency energy ratio of the region. The relative weight of the ratio between different regions is adjusted in combination with the rate ratio data. After weighted processing of the energy ratios of all regions, a unified format regional high- and low-frequency energy ratio integral data matrix is ​​formed. The matrix has each region as a row and the high- and low-frequency energy ratios as columns. The comparison data is used for subsequent high-frequency energy structure analysis.

[0087] Based on the regional high- and low-frequency energy ratio integral data obtained in step S22, the high-frequency energy attenuation structure between different regions is analyzed and processed. First, the regions whose energy ratios deviate significantly from the regional mean are selected and marked, and the deviation threshold is set to ±15% of the regional mean. The high-frequency energy integral gradient values ​​of these significant regions are extracted. The calculation method is to divide the difference between the ratio of the region and the ratio of its adjacent regions by the square of the distance between the regional centers to obtain the energy attenuation gradient vector. Then, the gradient vectors of all regions are spatially interpolated, and a three-dimensional energy attenuation field is constructed using the inverse distance weighted method. The attenuation field is then used to extract the features of the main gradient direction of the attenuation direction. The method is to obtain the main direction change information through the eigenvalue decomposition of the Hessian matrix, perform local fitting analysis on the energy attenuation trend of the main direction, and use the cubic spline function to fit the gradient change curve. Finally, the high-frequency energy attenuation amplitude and change trend of each region in the main direction are output to form a regional high-frequency energy attenuation structure data set. Based on the regional high-frequency energy attenuation structure data in step S23, the multi-scale localized differences between different regions are analyzed and processed. First, the regional data is constructed into a three-dimensional coordinate array according to the physical spatial position. The three-scale difference method is used to calculate the mean difference of high-frequency energy attenuation in each region within 10 neighborhoods in the large scale, 6 neighborhoods in the medium scale, and 3 neighborhoods in the small scale. At the same time, a two-dimensional Gaussian kernel function is used to perform weighted smoothing on the local difference value of each scale. Then, the localized difference values ​​at the three scales are merged to construct a three-channel feature map. The feature map is subjected to principal component analysis (PCA) to extract the main difference feature dimensions. The variance analysis method is then used to perform a significance test on the difference distribution at each scale, and the localized areas with significant changes are extracted. The difference values ​​of these areas are Z-score normalized to finally form a unified format of attenuation structure localized difference data matrix, which records the energy structure heterogeneity parameters of each region at different scales and is used for subsequent pulp structure correlation analysis.

[0088] Step S22 includes the following steps:

[0089] Step S221: performing high-low frequency change morphology identification of sound waves between different regions based on the regional sound wave frequency change rate ratio and the regional sound wave bandwidth change data to obtain high-low frequency change morphology data of sound waves between different regions;

[0090] Step S222: performing high-low frequency discrete entropy regression analysis on the high-low frequency variation morphological data of the sound waves between different regions to obtain high-low frequency discrete entropy regression data;

[0091] Step S223: performing nonlinear fluctuation polynomial interpolation on the high-low frequency variation morphological data of the sound waves between different regions based on the high-low frequency discrete entropy value regression data to obtain high-low frequency nonlinear fluctuation interpolation data;

[0092] Step S224: calculating the variance of the segmented fluctuation slope growth rate on the high-low frequency nonlinear fluctuation interpolation data to obtain the variance of the high-low frequency segmented fluctuation slope growth rate;

[0093] Step S225: performing high-low frequency energy ratio integration processing on the variance of the high-low frequency segmented fluctuation slope growth rate between different regions according to the Romberg integration method to obtain regional high-low frequency energy ratio integration data.

[0094] In an embodiment of the present invention, based on the regional acoustic wave frequency change rate ratio data and the regional acoustic wave bandwidth change data obtained in step S21, a two-dimensional feature comparison map is constructed to identify the high and low frequency change morphologies of acoustic waves between different regions. First, the bandwidth change data of each region is divided into two parts on the spectrum: a low frequency band of 1.5 MHz to 2.2 MHz and a high frequency band of 2.2 MHz to 3.0 MHz. The energy density change gradients of the frequency change curves in the two frequency bands are extracted respectively, and the change trend weights of the high and low frequency parts of each region are adjusted using the frequency change rate ratio. Then, the second-order differential curvatures of the morphological curves of the two frequency bands are calculated respectively. The curvature change patterns of the high and low frequency bands of different regions are compared with the position offsets of the energy concentration areas. Based on whether the offsets are concentrated at the two ends or the center of the spectrum, the change morphology types are identified, including five types: low frequency type, high frequency increasing type, symmetrical type, asymmetric high frequency contraction type, and multi-peak type. Finally, the change morphology corresponding to each region is recorded in matrix form to form a sound wave high and low frequency change morphology dataset, which is used for further entropy analysis of high and low frequency structural differences. For the high and low frequency change morphological data of the sound waves obtained in step S221, high and low frequency discrete entropy value regression analysis is performed. First, each regional type in the morphological data is numerically encoded, and the frequency energy peak positions of the high and low frequency parts of each region are discretized into 20 frequency nodes according to the frequency gradient distribution density map. The energy value at each node is normalized and a frequency distribution probability vector is constructed. The Shannon entropy definition is used to calculate the discrete entropy values ​​of the high and low frequency bands of each region. Then, a linear regression fitting analysis is performed on the entropy value ratio of the high and low frequency bands between different regions. The fitting variables are the entropy value ratio and the frequency change rate ratio. The regression parameters are solved by the least squares method and the fitting residual distribution is evaluated. The high and low frequency discrete entropy value regression data finally output includes the regression residual, fitting slope and discrete trend index of the entropy value ratio in the high and low frequency bands of each region.

[0095] Based on the high- and low-frequency discrete entropy regression data obtained in step S222, the above-mentioned morphological data is subjected to nonlinear fluctuation polynomial interpolation processing. First, the energy values ​​at 20 frequency points in the high-frequency and low-frequency spectrum intervals of each region are respectively extracted to form a sampling vector. The entropy regression residual is used as a disturbance factor to fit the original data into a fifth-order nonlinear polynomial function. The function coefficients are obtained by the minimum norm approximation method. After the polynomial interpolation is completed, the first-order derivative and second-order derivative sequence of the interpolation function in the entire spectrum range are calculated to construct a fluctuation change trend graph. The interpolated spectrum curve is then compared with the original morphological data for point-to-point error. After confirming that the interpolation error does not exceed 0.03 of the normalized amplitude, high- and low-frequency nonlinear fluctuation interpolation data are formed. This data is used to extract the growth rate trend of frequency change and the slope information of regional energy change. Based on the high- and low-frequency nonlinear fluctuation interpolation data obtained in step S223, the interpolation function is subjected to spectrum segmentation processing. The interpolation frequency band is divided into 10 segments using a fixed bandwidth of 0.15 MHz. The first-order derivative value of each curve segment is extracted as the slope growth rate indicator of the segment. Then, the standard deviation is calculated from the segment slope growth rate values ​​of each segment in each region and the position of the maximum change segment is recorded. The sliding variance window method is used to calculate the growth rate variance of the three consecutive segment slopes. The window size is set to 3 segments. After smoothing the variance sequence of the slope growth rate within the full spectrum range, the high and low frequency fluctuation slope change characteristics of each region are obtained. Finally, a high and low frequency segment fluctuation slope growth rate variance data matrix is ​​formed with the region number as the row and the fluctuation slope growth rate variance of each frequency band as the column for the next energy ratio integral analysis. Based on the high- and low-frequency segmented fluctuation slope growth rate variance data obtained in step S224, the Lumberg integration method is used to integrate the high-frequency and low-frequency slope growth rate variance curves of each region. The Lumberg integration method constructs an integration sequence in an adaptive encrypted node manner. The initial integration spacing is set to 0.05MHz, and the integration accuracy error is controlled within 1e-5. The high-frequency and low-frequency slope growth rate functions are integrated for each region respectively, and the high-frequency integral value is divided by the low-frequency integral value to obtain the energy structure change ratio. This ratio constitutes the core indicator of the regional high- and low-frequency energy ratio integral data. The ratios of all regions are then summarized into a matrix form. At the same time, the weight and error correction parameters of each frequency interval in the integration process are recorded to constitute the final output regional high- and low-frequency energy ratio integral data.

[0096] Step S23 includes the following steps:

[0097] Step S231: identifying high-frequency energy attenuation fluctuation frequency bands between different regions based on regional high-low frequency energy ratio integral data, and obtaining high-frequency energy attenuation fluctuation frequency bands between different regions;

[0098] Step S232: performing attenuation concentration density analysis on the high-frequency energy attenuation fluctuations between different regions to obtain the high-frequency energy attenuation concentration density between different regions;

[0099] Step S233: performing non-uniform complex wavenumber gradient differential processing on the high-frequency energy attenuation fluctuation frequency band based on the high-frequency energy attenuation concentration density to obtain high-frequency attenuation complex wavenumber gradient differential data;

[0100] Step S234: calculating the attenuation acceleration variance between different regions of the high-frequency energy attenuation fluctuation frequency band according to the high-frequency attenuation complex wave number gradient differential data to obtain the attenuation acceleration variance;

[0101] Step S235: performing high-frequency energy attenuation structure analysis between different regions based on the high-frequency attenuation complex wavenumber gradient differential data and the attenuation acceleration variance to obtain regional high-frequency energy attenuation structure data.

[0102] In an embodiment of the present invention, based on the regional high- and low-frequency energy ratio integral data obtained in step S225, a focused analysis is performed on the high-frequency bands in the spectral data of all regions. The high-frequency band is set to 2.2 MHz to 3.0 MHz, and the band is divided into 40 equally spaced sub-bands, each with a width of 0.02 MHz. The spectral energy density value of each region within the band is extracted to form a band energy sequence. The sequence is smoothed using a five-point sliding average to eliminate local minimal disturbances. Then, the energy drop rate between adjacent bands is calculated, and the frequency band intervals with a drop rate exceeding 25% are marked as candidate attenuation fluctuation bands. The candidate bands are further subjected to a continuous drop segment determination logic, and three or more consecutive frequency bands that meet the drop rate threshold and have a negative local energy gradient are set as valid attenuation fluctuation bands. By performing the above logical recognition operation on the spectral data of each region, the high-frequency energy attenuation fluctuation bands of each region are finally extracted and recorded in a matrix form with the region number as the row and the frequency band start and end boundaries as the column, forming a high-frequency energy attenuation fluctuation band data set between different regions. Based on the high-frequency energy attenuation fluctuation frequency band data of different regions obtained in step S231, a clustering density analysis operation is performed on the attenuation frequency band in each region. First, an energy decrease gradient sequence is constructed for the energy density values ​​within the attenuation frequency band of each region. The first-order difference of the sequence is calculated and the number and length of negative segments are counted. The local clustering degree is defined as the number of negative segments multiplied by the average length. All local clustering degrees are normalized to unify the data scale. Then, in each region, the peak search algorithm is used to find the local strongest attenuation point as the clustering center. Boundary points are searched for on both sides of the clustering center in the directions of increasing and decreasing frequency. The expansion stop condition is that the energy change rate of two consecutive points is less than 2%. The clustering boundary range is obtained, and the total energy density drop value within the boundary range is divided by the frequency band width to form the high-frequency energy attenuation clustering density value. Finally, the clustering density values ​​of all regions are stored in a three-dimensional array, with the dimensions corresponding to the region number, frequency band number and clustering density value, respectively, to form a high-frequency energy attenuation clustering density data set between different regions.

[0103] Based on the high-frequency energy attenuation concentration density data obtained in step S232, the frequency band is subjected to non-uniform complex wavenumber gradient differential processing. First, the energy density spectrum in the high-frequency attenuation band of each region is converted into a complex spectrum representation. The real part is the actual energy density value, and the imaginary part is obtained by Hilbert transform to obtain the corresponding phase response function. After obtaining the complex spectrum, the five-point central difference method is used to calculate its first-order complex wavenumber gradient value. The variable step size difference method is further used to calculate the rate of change of the gradient in the spectral direction to form a complex wavenumber second-order differential matrix. The differential matrix is ​​subjected to non-uniformity analysis. The judgment standard is that the region where the difference between the local maximum and minimum values ​​exceeds twice the standard deviation of the mean is defined as a non-uniform high-variance interval. The data slices of this interval are then extracted to construct non-uniform complex wavenumber gradient differential data. All processing is performed under the condition of frequency accuracy of 0.005 MHz. Finally, the complex wavenumber gradient differential three-dimensional tensor data indexed by the region number and the frequency band number is output.

[0104] Based on the high-frequency attenuation complex wavenumber gradient differential data in step S233, the attenuation acceleration variance of each region within the attenuation frequency band is calculated. Acceleration is defined as the rate of change of the complex wavenumber gradient, that is, the second-order derivative value. The complex wavenumber gradient differential value in the spectrum sequence of each region is extracted, and its local acceleration, that is, the slope change rate sequence, is calculated. The sequence is segmented according to each 0.05 MHz bandwidth. The statistical variance of each acceleration sequence is calculated and stored as the segmented acceleration variance value. At the same time, the frequency center value of the segment is recorded for subsequent structure positioning. The acceleration variance of all frequency bands is then Z-score normalized, and the local maximum extraction algorithm is used to extract the significant variation segment. This part of the data forms the attenuation acceleration variance matrix. The frequency band segment variance and significant change index corresponding to each region are recorded for subsequent energy structure spectrum reconstruction. The high-frequency attenuation complex wavenumber gradient differential data in step S233 and the attenuation acceleration variance data in step S234 are combined to perform high-frequency energy attenuation structure analysis between different regions. First, the complex wavenumber gradient differential matrix and the acceleration variance matrix are fused in corresponding dimensions, and convolution sliding accumulation processing is performed on them according to the interval number on the frequency axis. The convolution kernel width is set to 3 frequency band intervals, and the step size is 1. The average gradient direction change value and the variance average value in the convolution window are extracted to form the local structure change comparison quantity. Principal component analysis is performed on all sliding window data to extract the dominant structure change direction, and then the high-frequency energy attenuation structure spectrum of each region is constructed through two-dimensional tensor mapping. The depth of color in the spectrum represents the local change intensity, and the color gradient direction represents the main direction of the complex wavenumber gradient. Finally, the structure spectrum matrix is ​​output and combined with the regional physical coordinate number to form the regional high-frequency energy attenuation structure data.

[0105] Step S233 includes the following steps:

[0106] Based on the high-frequency energy attenuation concentration density, the frequency decay node non-uniformity analysis of the concentrated center waveform of the high-frequency energy attenuation fluctuation frequency band is performed to obtain the frequency decay node non-uniformity data;

[0107] According to the non-uniform data of the frequency decay node, the high-frequency energy attenuation fluctuation frequency band is processed by complex spectrum construction to obtain a non-uniformly distributed complex spectrum matrix;

[0108] Perform first-order complex wave number differentiation on the non-uniformly distributed complex spectrum matrix to obtain complex wave number gradient change data;

[0109] The complex domain phase singular spectrum entropy is calculated for the complex wave number gradient change data to obtain the complex domain phase singular spectrum entropy;

[0110] The high-frequency attenuated complex wave number gradient differential data are obtained by performing non-uniform complex wave gradient differential processing according to the complex domain phase singular spectrum entropy.

[0111] In this embodiment of the present invention, the peak point of the frequency concentration energy drop in each region is extracted from the high-frequency energy attenuation concentration density data obtained in step S232 as the initial concentration center, and the frequency range is set to 2.35 MHz to 2.85 MHz. Within this range, an expansion window with 5 sub-bands on the left and right sides is constructed with the concentration center as the base point, and the sub-band width is 0.02 MHz. The spectrum within the window is analyzed, and the frequency decay node is determined by detecting whether the energy density change rate between two consecutive frequency bands exceeds a 15% threshold. The energy drop rate, frequency spacing, and second-order derivative values ​​of the spectrum morphology at all nodes are combined into a feature vector matrix. The degree of change between these nodes is evaluated using a normalized index constructed using the standard deviation and the mean, and then the degree of non-uniformity between the nodes is calculated. The mean of the product of the absolute value of the local energy density gradient change and the change frequency difference within the non-uniform fluctuation range is used as the non-uniformity index of the frequency decay node in each region, and finally a frequency decay node non-uniformity data matrix is ​​generated. When constructing the complex spectrum of the high-frequency energy attenuation fluctuation band based on the non-uniform data of the frequency decay nodes, each frequency point in the attenuation band is first reconstructed in the complex domain. The energy density value of the actual spectrum data is used as the real part, and the Hilbert transform is used to calculate the instantaneous phase response corresponding to each frequency point as the imaginary part to construct a complex spectrum signal. This complex spectrum signal is sampled in the frequency range of 2.35 MHz to 2.85 MHz with a resolution of 0.005 MHz to construct a two-dimensional complex spectrum matrix. The horizontal axis of the matrix is ​​the frequency point, and the vertical axis is the real and imaginary components of the complex value. Complex signal normalization is then performed on each column in the matrix to limit the modulus of all complex frequency vectors to between 0 and 1, ensuring that subsequent differential calculations are not distorted due to nonlinear amplitude amplification. The non-uniformly distributed complex spectrum matrix obtained by this processing accurately characterizes the coupling relationship between phase and amplitude during the frequency decay process.

[0112] When performing first-order complex wavenumber differentiation on the non-uniformly distributed complex spectrum matrix, the five-point central difference method is used to calculate the complex derivative of each frequency point of the complex spectrum matrix. For each complex value in the frequency point sequence, the first-order difference of its real and imaginary parts in the frequency direction is calculated with the two adjacent points before and after as the base points, and the difference is synthesized into a complex derivative value, i.e., the complex wavenumber gradient change data. This differential calculation is strictly limited to a frequency step of 0.005 MHz. After being executed for the entire frequency range, a one-dimensional complex wavenumber gradient vector is formed, which is stored in matrix form according to the region number to form a complete complex wavenumber gradient change data set. When calculating the complex domain phase singular spectrum entropy of complex wavenumber gradient change data, the instantaneous phase change value is first calculated for each frequency point in the complex wavenumber gradient vector, and the imaginary and real parts of the complex number are converted into a phase angle sequence using the inverse tangent function. Then, a sliding window of 5 frequency points is divided over the entire frequency band. The distribution probability of the phase angle difference is statistically analyzed in each window. Based on this distribution, the entropy value within the window is calculated and its change trend is recorded. Finally, a complex domain phase singular spectrum entropy sequence is constructed. The entropy peak point, slope mutation point and average entropy value of the sequence are normalized and compared, and a local singularity index is constructed to identify the structural complexity of phase information in different frequency bands. This calculation process is performed independently in each region, and the data of all regions are aggregated to form the complex domain phase singular spectrum entropy matrix. When performing non-uniform complex wave gradient differential processing based on the complex domain phase singular spectrum entropy, the significant mutation points in the singular spectrum entropy matrix are used as non-uniform high-variable frequency reference points. Symmetrical frequency band windows are constructed around these points, and the corresponding complex wavenumber gradient change data are re-extracted within the window range. The local second-order derivative is further calculated using the central difference method to construct the complex wavenumber second-order gradient change sequence. The degree of non-uniformity of the sequence is analyzed, and the difference between the maximum slope change value and the minimum slope change value is used as the non-uniformity index. The singular spectrum entropy change slope and the second-order gradient change value are combined to perform weighted normalization processing to output the final high-frequency attenuation complex wavenumber gradient differential data.

[0113] Step S3 includes the following steps:

[0114] Step S31: identifying the acoustic wave scattering enhancement gradients between different regions based on the attenuation structure localization difference data to obtain the acoustic wave scattering enhancement gradients between different regions;

[0115] Step S32: performing a simulation evaluation of the pear intercellular expansion based on the acoustic wave scattering enhancement gradients between different regions to obtain intercellular expansion data;

[0116] Step S33: performing a simulation evaluation of the pulp viscosity increment based on the attenuation structure localization difference data to obtain pulp viscosity increment data;

[0117] Step S34: Evaluate the softness and toughness of the pulp based on the intercellular space expansion data and the pulp viscosity increment data to obtain pulp softness and toughness evaluation data.

[0118] As an example of the present invention, refer to Figure 2 As shown, in this example, step S3 includes:

[0119] Step S31: identifying the acoustic wave scattering enhancement gradients between different regions based on the attenuation structure localization difference data to obtain the acoustic wave scattering enhancement gradients between different regions;

[0120] In an embodiment of the present invention, based on the localized difference data of the attenuation structure obtained in step S24, the energy change trend in the acoustic wave reflection path of each region is analyzed. First, the frequency analysis range is set to 2.2 MHz to 3.0 MHz, and the frequency band is divided into 40 frequency sub-bands of equal width. The complex wave number gradient value and the attenuation acceleration variance value in each sub-band are extracted respectively. By jointly fitting the complex wave directionality change and the energy decrease trend of each sub-band, the local amplitude mutation point of the acoustic wave scattering is extracted using the spatial derivative enhancement algorithm. The scattering enhancement coefficient is calculated for each region as a unit. The coefficient is the product of the energy change rate per unit frequency and the spatial direction gradient. The scattering enhancement coefficients of all regions are further constructed into a two-dimensional matrix, where the matrix rows represent the spatial scanning angles and the columns represent the frequency sub-band numbers. The local maximum retrieval algorithm is used to identify the scattering enhancement gradient map formed by the enhancement mutation band, and finally the acoustic wave scattering enhancement gradient data between different regions are output.

[0121] Step S32: performing a simulation evaluation of the pear intercellular expansion based on the acoustic wave scattering enhancement gradients between different regions to obtain intercellular expansion data;

[0122] In an embodiment of the present invention, according to the acoustic scattering enhancement gradient data obtained in step S31, the range of significantly enhanced local scattering intensity in each region is identified in the frequency space domain. The enhanced region is strongly correlated with the acoustic wave interface reflection behavior of the cell microstructure. The reflectivity improvement interval within 1.5 mm of each enhanced region is calculated, and the morphological change of the reflection source is calculated in combination with the change amplitude of the scanning angle. The local acoustic wave phase delay regression analysis algorithm is used to reversely reconstruct the acoustic wave propagation delay change trend in different regions. The expansion trend of the cell gap is simulated by constructing a propagation path length change model. Using parameters related to the sound velocity and cell tissue density in known materials, the sound velocity is set to 1540 m / s, the time difference is converted into a spatial scale difference, and the intercellular structure displacement is calculated by a morphological regression model based on multi-angle data fusion. A cell gap expansion simulation curve is formed in each region, and a three-dimensional data body is constructed using the region number and output as cell gap expansion data.

[0123] Step S33: performing a simulation evaluation of the pulp viscosity increment based on the attenuation structure localization difference data to obtain pulp viscosity increment data;

[0124] In an embodiment of the present invention, based on the localized difference data of the attenuation structure obtained in step S24, the frequency attenuation density map and the reflection structure density map of each region are fitted and integrated, and the change value of the absorbed power density per unit volume in the high-frequency energy concentrated attenuation band of each region is calculated. The change value is correlated with the local complex wavenumber gradient change rate, and the difference value of the absorption capacity per unit volume is extracted as the main basis for determining the change in the viscosity characteristics inside the pulp. The Laplace inverse transformation is further used to construct a local impedance mapping of the energy attenuation distribution to the pulp structure. The known mechanical parameters of the pulp tissue, such as the density of 1020 kg / m³ and the viscosity coefficient of 0.45 Pa·s, are used to invert the change in the shear deformation resistance of the pulp area according to the energy attenuation rate, and the result is normalized with the total amount of reflected energy in the region to obtain the pulp viscosity increment coefficient. Finally, the viscosity increments of all regions are tensor-encoded to form a pulp viscosity increment data set for subsequent structural mechanical characteristic evaluation.

[0125] Step S34: Evaluate the softness and toughness of the pulp based on the intercellular space expansion data and the pulp viscosity increment data to obtain pulp softness and toughness evaluation data.

[0126] In an embodiment of the present invention, the cell gap expansion data obtained in step S32 and the pulp viscosity increment data obtained in step S33 are integrated, and the two data sets are dimensionally aligned. The region number and the frequency sub-segment number are uniformly mapped to lattice coordinates in three-dimensional space. Polynomial interpolation is used to construct a surface of the local structural volume change of the pulp. The shear modulus of each regional unit volume on the surface is reconstructed, and the relationship between its stress change rate and the local gap increment is calculated. The static compression modulus data of the pear pulp sample in the actual measurement is used as a reference, and the reference elastic modulus is set to 1.6 MPa. The relative ratio of the stress response of each region is calculated, and then the comprehensive soft-toughness score is calculated based on the deformation response and the damping growth value. The region-by-region integration algorithm is used to calculate the soft-toughness integral value within the region, and then all the regional integral values ​​are constructed into a three-dimensional spatial structure characteristic map, and the pulp soft-toughness evaluation data is output.

[0127] Step S33 includes the following steps:

[0128] Step S331: performing pulp-related absorption attenuation measurement mapping based on the attenuation structure localization difference data to obtain pulp-related absorption attenuation measurement data;

[0129] Step S332: Deducing the regional pulp density difference interval based on the pulp-related absorption attenuation measurement data to obtain the regional pulp density difference interval;

[0130] Step S333: performing pulp shear viscosity increment fitting based on the pulp-associated absorption attenuation measurement data and the regional pulp density difference interval to obtain pulp shear viscosity increment fitting data;

[0131] Step S334: performing a pulp volume viscosity ratio analysis based on the pulp-associated absorption attenuation measurement data, the pulp shear viscosity increment fitting data, and the pulp density difference interval to obtain the pulp volume viscosity ratio;

[0132] Step S335: performing a simulation evaluation of the pulp viscosity increment based on the pulp shear viscosity increment fitting data and the pulp volume viscosity ratio to obtain pulp viscosity increment data.

[0133] As an example of the present invention, refer to Figure 3 As shown, in this example, step S33 includes:

[0134] Step S331: performing pulp-related absorption attenuation measurement mapping based on the attenuation structure localization difference data to obtain pulp-related absorption attenuation measurement data;

[0135] In an embodiment of the present invention, based on the localized difference data of the attenuation structure obtained in step S24, the energy attenuation rate in each scanning area is integrated and extracted frame by frame in the frequency band of 2.5 MHz to 3.0 MHz to construct a regional frequency domain attenuation mapping diagram, and the regional main frequency energy transfer density distribution function is used to calculate the acoustic energy dissipation value per unit volume. The energy distribution balance of different areas is adjusted and corrected in combination with the scattering enhancement gradient distribution curve. After the non-pulp tissue reflection interference component is eliminated by the Brownian distribution noise elimination method, the correlation relationship between energy attenuation and local tissue absorption capacity is established using the matrix linear mapping method, and finally a quantization matrix of absorption attenuation value corresponding to each spatial coordinate is formed as the pulp-related absorption attenuation measurement data.

[0136] Step S332: Deducing the regional pulp density difference interval based on the pulp-related absorption attenuation measurement data to obtain the regional pulp density difference interval;

[0137] In an embodiment of the present invention, based on the pulp-associated absorption attenuation measurement data obtained in step S331, the two-dimensional regional attenuation matrix is ​​projected into the density change interval mapping space. By fitting the nonlinear relationship curve between the energy dissipation rate and the tissue physical density, a pear pulp sample with a reference density of 1.02 g / cm³ is set as a benchmark, and a Gaussian fitting is performed with its attenuation coefficient of 0.015 dB / mm as the center point. By analyzing the correspondence between each attenuation segment and the known density standard, a density distribution interval mapping table is constructed. Then, according to the mapping table, the density distribution of all regions is estimated, and the maximum and minimum ranges of the density values ​​of each region are extracted to construct a density difference matrix to obtain the regional pulp density difference interval.

[0138] Step S333: performing pulp shear viscosity increment fitting based on the pulp-associated absorption attenuation measurement data and the regional pulp density difference interval to obtain pulp shear viscosity increment fitting data;

[0139] In an embodiment of the present invention, based on the pulp-associated absorption attenuation measurement data obtained in step S331 and the regional pulp density difference interval obtained in step S332, the density difference interval of the center point of each region is selected as the fitting parameter boundary, and a multinomial fitting function is used to perform bivariate fitting of the absorption value and the density interval. The minimum value of the second-order derivative is used as the stable point of the shear viscosity increment change, and the least squares method is used to estimate the relationship between the acoustic wave viscosity dissipation and the tissue stress difference per unit area at this point. Combined with the pulp shear deformation response curve obtained in the experiment, a linearized incremental model is established for each interval to form pulp shear viscosity incremental fitting data for subsequent volume scale transformation.

[0140] Step S334: performing a pulp volume viscosity ratio analysis based on the pulp-associated absorption attenuation measurement data, the pulp shear viscosity increment fitting data, and the pulp density difference interval to obtain the pulp volume viscosity ratio;

[0141] In an embodiment of the present invention, the pulp-associated absorption attenuation measurement data obtained in step S331, the pulp shear viscosity increment fitting data obtained in step S333, and the pulp density difference interval obtained in step S332 are input into a three-dimensional tensor fitting function. The energy absorption capacity change field of each region at the volume scale is constructed based on the tensor kernel expansion algorithm. The unit analysis volume is set to 1 mm³, and the partial derivative of the product of the energy absorption change and the density disturbance and the viscosity increment in the volume is calculated. The composite triangular grid integration method is used to obtain the shear viscosity ratio per unit volume of each region, and finally the pulp volume viscosity ratio data is formed.

[0142] Step S335: performing a simulation evaluation of the pulp viscosity increment based on the pulp shear viscosity increment fitting data and the pulp volume viscosity ratio to obtain pulp viscosity increment data.

[0143] In an embodiment of the present invention, based on the pulp shear viscosity increment fitting data obtained in step S333 and the pulp volume viscosity ratio data obtained in step S334, the local shear response rate and energy dissipation efficiency of each region in the 3 MHz frequency band are extracted respectively, and the two are normalized and constructed into a two-dimensional function curve map. The viscosity change surface is smoothly fitted using a Bessel third-order curve, and the impedance change in the energy propagation path at high frequency is piecewise integrated. The viscosity change increment per unit thickness of each region is extracted, and weighted integration is performed in combination with the distribution density of the difference structure between regions. Finally, the pulp viscosity increment data is output and marked in the spatial area distribution map for subsequent comprehensive evaluation of softness and toughness.

[0144] The present invention also provides an agricultural product maturity detection system for executing the agricultural product maturity detection method described above, the agricultural product maturity detection system comprising:

[0145] The acoustic wave bandwidth variation analysis module is used to scan the pears at multiple angles using an ultrasonic detector, and then perform frequency domain analysis on the regional division to generate regional frequency domain data of the reflected acoustic wave; the acoustic wave bandwidth variation analysis is performed on the regional frequency domain data of the reflected acoustic wave to obtain regional acoustic wave bandwidth variation data;

[0146] The high-frequency energy attenuation structure analysis module is used to integrate the high-to-low-frequency energy ratio between different regions based on the regional acoustic wave bandwidth change data to obtain regional high-to-low-frequency energy ratio integral data; based on the regional high-to-low-frequency energy ratio integral data, the high-frequency energy attenuation structure analysis between different regions is performed to obtain localized difference data of the attenuation structure;

[0147] The pulp softness and toughness evaluation module is used to simulate and evaluate the pulp viscosity increment based on the attenuation structure localization difference data to obtain the pulp viscosity increment data; and to evaluate the pulp softness and toughness based on the pulp viscosity increment data to obtain the pulp softness and toughness evaluation data;

[0148] The maturity evaluation module is used to evaluate the maturity of pears based on the flesh softness and toughness evaluation data, thereby obtaining the pear maturity evaluation data.

[0149] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A method for detecting the maturity of agricultural products, characterized in that: The following steps are involved: Step S1: Scanning the pear at multiple angles using an ultrasonic detector, then performing frequency domain analysis on the regional division to generate regional frequency domain data of reflected sound waves; performing sound wave bandwidth variation analysis on the regional frequency domain data of reflected sound waves to obtain regional sound wave bandwidth variation data; Step S2: performing high-low frequency energy ratio integration processing on the regional acoustic wave bandwidth change data between different regions to obtain regional high-low frequency energy ratio integration data; Based on the regional high-low frequency energy ratio integral data, high-frequency energy attenuation structure analysis is performed between different regions to obtain localized difference data of attenuation structure; wherein step S2 includes: Step S21: calculating the sound wave frequency change rate ratios between different regions based on the regional sound wave bandwidth change data to obtain the regional sound wave frequency change rate ratios; Step S22: performing high-low frequency energy ratio integration processing on the regional sound wave bandwidth change data between different regions according to the regional sound wave frequency change rate ratio to obtain regional high-low frequency energy ratio integration data; Step S23: performing high-frequency energy attenuation structure analysis between different regions based on the regional high-low frequency energy ratio integral data to obtain regional high-frequency energy attenuation structure data; Step S24: performing multi-scale localized difference analysis between different regions on the regional high-frequency energy attenuation structure data to obtain attenuation structure localized difference data; Step S3: performing a simulation evaluation of the pulp viscosity increment based on the attenuation structure localization difference data to obtain pulp viscosity increment data; performing a pulp softness and toughness evaluation based on the pulp viscosity increment data to obtain pulp softness and toughness evaluation data; Step S4: evaluating the maturity of the pears according to the pulp softness and toughness evaluation data, thereby obtaining the pear maturity evaluation data.

2. The method for detecting the maturity of agricultural products according to claim 1, wherein: Step S1 includes the following steps: Step S11: Scan the pear at multiple angles using an ultrasonic detector to obtain multi-angle reflected sound waves from the pear; Step S12: performing noise filtering on the pear multi-angle reflected sound waves to obtain the pear multi-angle reflected filtered sound waves; Step S13: performing frequency domain analysis on the pear multi-angle reflected filtered sound waves to generate frequency domain data of the reflected sound waves; Step S14: performing acoustic wave bandwidth change analysis on the regional frequency domain data of the reflected acoustic wave to obtain regional acoustic wave bandwidth change data.

3. The method for detecting the maturity of agricultural products according to claim 1, wherein: Step S22 includes the following steps: Step S221: performing high-low frequency change morphology identification of sound waves between different regions based on the regional sound wave frequency change rate ratio and the regional sound wave bandwidth change data to obtain high-low frequency change morphology data of sound waves between different regions; Step S222: performing high-low frequency discrete entropy regression analysis on the high-low frequency variation morphological data of the sound waves between different regions to obtain high-low frequency discrete entropy regression data; Step S223: performing nonlinear fluctuation polynomial interpolation on the high-low frequency variation morphological data of the sound waves between different regions based on the high-low frequency discrete entropy value regression data to obtain high-low frequency nonlinear fluctuation interpolation data; Step S224: calculating the variance of the segmented fluctuation slope growth rate on the high-low frequency nonlinear fluctuation interpolation data to obtain the variance of the high-low frequency segmented fluctuation slope growth rate; Step S225: performing high-low frequency energy ratio integration processing on the variance of the high-low frequency segmented fluctuation slope growth rate between different regions according to the Romberg integration method to obtain regional high-low frequency energy ratio integration data.

4. The method for detecting the maturity of agricultural products according to claim 1, wherein: Step S23 includes the following steps: Step S231: identifying high-frequency energy attenuation fluctuation frequency bands between different regions based on regional high-low frequency energy ratio integral data, and obtaining high-frequency energy attenuation fluctuation frequency bands between different regions; Step S232: performing attenuation concentration density analysis on the high-frequency energy attenuation fluctuations between different regions to obtain the high-frequency energy attenuation concentration density between different regions; Step S233: performing non-uniform complex wavenumber gradient differential processing on the high-frequency energy attenuation fluctuation frequency band based on the high-frequency energy attenuation concentration density to obtain high-frequency attenuation complex wavenumber gradient differential data; Step S234: calculating the attenuation acceleration variance between different regions of the high-frequency energy attenuation fluctuation frequency band according to the high-frequency attenuation complex wave number gradient differential data to obtain the attenuation acceleration variance; Step S235: performing high-frequency energy attenuation structure analysis between different regions based on the high-frequency attenuation complex wavenumber gradient differential data and the attenuation acceleration variance to obtain regional high-frequency energy attenuation structure data.

5. The method for detecting the maturity of agricultural products according to claim 4, wherein: Step S233 includes the following steps: Based on the high-frequency energy attenuation concentration density, the frequency decay node non-uniformity analysis of the concentrated center waveform of the high-frequency energy attenuation fluctuation frequency band is performed to obtain the frequency decay node non-uniformity data; According to the non-uniform data of the frequency decay node, the high-frequency energy attenuation fluctuation frequency band is processed by complex spectrum construction to obtain a non-uniformly distributed complex spectrum matrix; Perform first-order complex wave number differentiation on the non-uniformly distributed complex spectrum matrix to obtain complex wave number gradient change data; The complex domain phase singular spectrum entropy is calculated for the complex wave number gradient change data to obtain the complex domain phase singular spectrum entropy; The high-frequency attenuated complex wave number gradient differential data are obtained by performing non-uniform complex wave gradient differential processing according to the complex domain phase singular spectrum entropy.

6. The method for detecting the maturity of agricultural products according to claim 1, wherein: Step S3 includes the following steps: Step S31: identifying the acoustic wave scattering enhancement gradients between different regions based on the attenuation structure localization difference data to obtain the acoustic wave scattering enhancement gradients between different regions; Step S32: performing a simulation evaluation of the pear intercellular expansion based on the acoustic wave scattering enhancement gradients between different regions to obtain intercellular expansion data; Step S33: performing a simulation evaluation of the pulp viscosity increment based on the attenuation structure localization difference data to obtain pulp viscosity increment data; Step S34: Evaluate the softness and toughness of the pulp based on the intercellular space expansion data and the pulp viscosity increment data to obtain pulp softness and toughness evaluation data.

7. The method for detecting the maturity of agricultural products according to claim 6, wherein: Step S33 includes the following steps: Step S331: performing pulp-related absorption attenuation measurement mapping based on the attenuation structure localization difference data to obtain pulp-related absorption attenuation measurement data; Step S332: Deducing the regional pulp density difference interval based on the pulp-related absorption attenuation measurement data to obtain the regional pulp density difference interval; Step S333: performing pulp shear viscosity increment fitting based on the pulp-associated absorption attenuation measurement data and the regional pulp density difference interval to obtain pulp shear viscosity increment fitting data; Step S334: performing a pulp volume viscosity ratio analysis based on the pulp-associated absorption attenuation measurement data, the pulp shear viscosity increment fitting data, and the pulp density difference interval to obtain the pulp volume viscosity ratio; Step S335: performing a simulation evaluation of the pulp viscosity increment based on the pulp shear viscosity increment fitting data and the pulp volume viscosity ratio to obtain pulp viscosity increment data.

8. A system for detecting the maturity of agricultural products, characterized in that: For executing the agricultural product maturity detection method according to claim 1, the agricultural product maturity detection system comprises: The acoustic wave bandwidth variation analysis module is used to scan the pears at multiple angles using an ultrasonic detector, and then perform frequency domain analysis on the regional division to generate regional frequency domain data of the reflected acoustic wave; the acoustic wave bandwidth variation analysis is performed on the regional frequency domain data of the reflected acoustic wave to obtain regional acoustic wave bandwidth variation data; The high-frequency energy attenuation structure analysis module is used to integrate the high-to-low-frequency energy ratio between different regions based on the regional acoustic wave bandwidth change data to obtain regional high-to-low-frequency energy ratio integral data; based on the regional high-to-low-frequency energy ratio integral data, the high-frequency energy attenuation structure analysis between different regions is performed to obtain localized difference data of the attenuation structure; The pulp softness and toughness evaluation module is used to simulate and evaluate the pulp viscosity increment based on the attenuation structure localization difference data to obtain the pulp viscosity increment data; and to evaluate the pulp softness and toughness based on the pulp viscosity increment data to obtain the pulp softness and toughness evaluation data; The maturity evaluation module is used to evaluate the maturity of pears based on the flesh softness and toughness evaluation data, thereby obtaining the pear maturity evaluation data.

Images