A method and related apparatus for detecting the fullness of passion fruit juice based on an AI workstation

Abstract

This application provides a method and related apparatus for detecting the fullness of passion fruit pulp based on an AI workstation. The AI ​​workstation acquires multiple near-infrared diffuse reflectance spectra of a single passion fruit; based on these spectra, it obtains water speciation characteristic parameters and internal structure scattering parameters, and constructs a multi-dimensional feature parameter set by combining fruit density characteristic parameters and peel attenuation interference correction parameters; the multi-dimensional feature parameters are fused and calculated to obtain a pulp fullness index, which characterizes the relative abundance and distribution of the flowable pulp inside the passion fruit; and grading is performed based on the pulp fullness index. This application achieves simultaneous and non-destructive characterization of multiple indicators inside the passion fruit through near-infrared diffuse reflectance spectroscopy combined with multi-parameter fusion, reducing interference, improving the accuracy and stability of the judgment, adapting to the real-time processing capabilities of the AI ​​workstation, and meeting the detection needs of rapid online grading of passion fruit in industrial sorting scenarios.

Description

Technical Field

[0001] This application relates to the field of industrial automation testing technology, and in particular to a method and related apparatus for detecting the fullness of passion fruit juice based on an AI workstation. Background Technology

[0002] In the industrialized fruit and vegetable sorting sector, server-based sorting line control systems typically include a real-time quality inspection subsystem built upon front-end testing equipment. This real-time quality inspection subsystem is usually customized according to the type of fruit and vegetable being sorted. Taking passion fruit sorting lines as an example, the fullness of the internal juice of passion fruit is an important indicator for evaluating the commercial quality and edible value of the fruit. It has key testing significance in fruit grading, storage, and processing. The real-time quality inspection subsystem needs to achieve accurate and real-time detection of the aforementioned indicators.

[0003] Existing technologies employ X-ray micro-computed tomography (CT) to detect the internal quality of passion fruit. A real-time quality inspection subsystem's computing equipment (such as a server) processes the perceived data, labels the results, and controls the sorting line's action units. The computing equipment acquires morphological and structural parameters such as fruit volume, peel thickness, pulp volume, and cavity size and distribution, enabling visualization of the internal three-dimensional structure and quantification of tissue segmentation. However, the detection results primarily reflect density differences and spatial morphology, only indicating internal fullness and cavity conditions. They cannot accurately characterize the relative abundance and distribution of flowable juice, making it difficult to accurately determine juice fullness. Furthermore, the computing equipment faces computational bottlenecks when the sorting line's conveyor belt is fully loaded and passion fruit is transported at high speed. Summary of the Invention

[0004] This application provides a method and related device for detecting the fullness of passion fruit pulp based on an AI workstation. By processing the near-infrared diffuse reflectance spectrum through the AI ​​workstation and combining it with multi-dimensional feature parameter fusion, it can simultaneously and non-destructively characterize the internal moisture state, pulp scattering, fruit density, and peel attenuation of passion fruit. This accurately reflects the relative abundance and distribution uniformity of the pulp, reduces the interference of the peel and internal cavity on the detection signal, and improves the stability and accuracy of pulp fullness determination, meeting the physical detection needs of rapid online grading. The detection process is adapted to the real-time data processing capabilities of the AI ​​workstation, without the need for additional computing power, and can quickly respond to the real-time detection needs of industrial sorting.

[0005] In a first aspect, embodiments of this application provide a method for detecting the fullness of passion fruit juice based on an AI workstation, applied to the AI ​​workstation. The method includes: acquiring multiple near-infrared diffuse reflectance spectra of a single passion fruit; acquiring water morphology characteristic parameters and internal structure scattering parameters of the passion fruit based on the near-infrared diffuse reflectance spectra, and constructing a multi-dimensional feature parameter set by combining the fruit density characteristic parameters and peel attenuation interference correction parameters of the passion fruit; performing fusion calculation on the multi-dimensional feature parameters to obtain a juice fullness index characterizing the relative abundance and distribution state of the flowable juice inside the passion fruit; and completing the grading determination of the fullness of the passion fruit juice based on the juice fullness index.

[0006] Secondly, this application provides a non-destructive testing device for passion fruit pulp fullness based on an AI workstation. The device includes: an acquisition unit for acquiring multiple near-infrared diffuse reflectance spectra of a single passion fruit; a processing unit for acquiring water morphology characteristic parameters and internal structure scattering parameters of the passion fruit based on the near-infrared diffuse reflectance spectra, and constructing a multi-dimensional feature parameter set by combining the fruit density characteristic parameters and peel attenuation interference correction parameters of the passion fruit; performing fusion calculation on the multi-dimensional feature parameters to obtain a pulp fullness index characterizing the relative abundance and distribution state of the flowable pulp inside the passion fruit; and completing the grading determination of the passion fruit pulp fullness based on the pulp fullness index.

[0007] Thirdly, embodiments of this application provide an AI workstation, including a processor, a memory, a communication interface, and a computer program, the computer program being stored in the memory and configured to be executed by the processor, the program including instructions for performing steps in the method as described in any one of the first aspects.

[0008] As can be seen, in this embodiment, the AI ​​workstation acquires near-infrared spectral data of passion fruit, extracts multi-dimensional feature parameters such as water form and internal scattering, constructs a multi-dimensional feature parameter set, and combines fruit density feature parameters and peel attenuation interference correction parameters to construct the multi-dimensional feature parameter set; the multi-dimensional feature parameters are fused and calculated to obtain the juice fullness index, which characterizes the relative abundance and distribution of the flowable juice inside the passion fruit; the grading is completed based on the juice fullness index. It is evident that this application, through the processing of near-infrared diffuse reflectance spectra by the AI ​​workstation and the fusion of multi-dimensional feature parameters, can achieve simultaneous and non-destructive characterization of the internal water state, pulp scattering, fruit density, and peel attenuation of passion fruit, accurately reflecting the relative abundance and distribution uniformity of the juice, reducing the interference of the peel and internal cavity on the detection signal, improving the stability and accuracy of juice fullness determination, and meeting the physical detection requirements for rapid online grading; the detection process is adapted to the real-time data processing capabilities of the AI ​​workstation, without requiring additional computing power, and can quickly respond to the real-time detection needs of industrial sorting. This solves the problem that traditional testing only focuses on structural state and cannot accurately reflect the condition of the pulp. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a schematic diagram of an existing X-ray micro-CT passion fruit detection device. Figure 2 Three-dimensional reconstruction and tomographic slices of passion fruit using X-ray micro-CT. Figure 3 This is a schematic diagram of the tower AI workstation in the embodiments of this application; Figure 4 A schematic diagram of an intelligent automated sorting line for passion fruit; Figure 5 This is a flowchart illustrating a method for detecting the fullness of passion fruit pulp based on an AI workstation. Figure 6 This is a schematic diagram of a near-infrared spectroscopy detection device; Figure 7 This is a functional unit structure diagram of a passion fruit juice fullness detection device based on an AI workstation. Figure 8 This is a schematic diagram of the server structure provided in an embodiment of this application; Figure 9 This is a schematic diagram of the near-infrared diffuse reflectance inside the near-infrared spectroscopy detection device. Detailed Implementation

[0011] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0012] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but in some embodiments includes steps or units not listed, or in some embodiments includes other steps or units inherent to these processes, methods, products, or apparatuses.

[0013] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0014] In the embodiments of this application, "and / or" describes the relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent the following three situations: A exists alone; A and B exist simultaneously; B exists alone. Among them, A and B can be singular or plural.

[0015] In this embodiment, the symbol " / " can indicate that the preceding and following objects are in an "or" relationship. Alternatively, the symbol " / " can also represent a division sign, i.e., performing a division operation. For example, A / B can mean A divided by B.

[0016] In the embodiments of this application, "at least one item" or its similar expression refers to any combination of these items, including any combination of a single item or a plurality of items. "One or more" means one or more, while "multiple" means two or more. For example, "at least one item" of a, b, or c can represent the following seven cases: a, b, c; a and b; a and c; b and c; a, b, and c. Each of a, b, and c can be an element or a set containing one or more elements.

[0017] In the embodiments of this application, "equal to" can be used with "greater than" and is applicable to technical solutions used when "greater than" is used; it can also be used with "less than" and is applicable to technical solutions used when "less than" is used. When "equal to" is used with "greater than", it is not used with "less than"; when "equal to" is used with "less than", it is not used with "greater than".

[0018] In existing technologies, X-ray microcomputed tomography (XCT) is used to detect the internal quality of passion fruit. During the detection process, if... Figure 1 As shown, a whole passion fruit is placed in the rotating stage detection area of ​​an X-ray micro-computed tomography (XCT) scanner. The scanner is equipped with a motion platform that can move along the X and Y axes, enabling omnidirectional, multi-angle positioning of the passion fruit in conjunction with the rotating stage. A conical X-ray beam is emitted from an X-ray source to perform omnidirectional, multi-angle tomographic scanning of the passion fruit. A flat panel detector collects the transmission signals, obtaining two-dimensional tomographic projection data of the passion fruit's interior. Subsequently, the scanned image data is transmitted to the computing device (such as a server / AI workstation) of the real-time quality inspection subsystem. This computing device performs three-dimensional reconstruction, image segmentation, and feature extraction on the image data, ultimately obtaining morphological and structural parameters of the passion fruit, specifically including fruit volume, peel thickness, pulp volume, cavity size, and cavity distribution. Based on these parameters, the computing device can visualize the three-dimensional structure of the passion fruit's interior, as well as segment and quantify the peel, pulp, and cavities, completing the detection and data output of the passion fruit's internal structure.

[0019] like Figure 2 As shown, Figure 2This image showcases the visualization results of internal examination of passion fruit using X-ray micro-CT technology. The upper left area of ​​the image is a 3D reconstructed model of the passion fruit, visually presenting the overall outline of the fruit through pseudo-color rendering. The other three view areas are orthogonal tomographic slices (cross-sectional / coronal views), clearly showing the internal tissue structure of the passion fruit based on density differences. The slice images clearly identify the outer peel (high-brightness ring area), the middle pulp (medium-density area), and the central internal cavity (large area of ​​low brightness or black voids). By combining 3D reconstruction and tomographic slicing, existing technology can quantify morphological and structural parameters such as fruit volume, peel thickness, cavity size, and spatial distribution. However, due to the limitations of the imaging principle in determining density differences, the boundaries between liquid juice and solid pulp / air cavities in the image are blurred, making it impossible to effectively distinguish them using grayscale or density values. This makes it difficult to directly characterize the relative abundance and uniformity of flowable juice, resulting in difficulties in accurately determining the juice fullness. Meanwhile, this technology relies on the three-dimensional reconstruction and complex calculation of a large amount of tomographic image data. When the sorting line is running at high speed and full load, it is easy to cause a bottleneck in the computing power of the computing equipment. It cannot meet the needs of accurate and real-time detection of the core quality indicator (juice fullness) of passion fruit in the industrial sorting of fruits and vegetables. It has obvious technical defects and urgently needs a better detection method to solve the problem.

[0020] To address the shortcomings of existing technologies, this application proposes a targeted improvement scheme. This application abandons the traditional X-ray micro-CT technique that relies solely on density difference imaging, and instead employs a detection scheme combining near-infrared diffuse reflectance spectroscopy with multi-dimensional feature parameter fusion. It is adapted to the real-time data processing capabilities of AI workstations, fundamentally solving the pain points of existing technologies, such as their inability to accurately characterize flowable pulp and computational bottlenecks. This application directly obtains characteristic information related to the pulp within passion fruit through near-infrared spectroscopy, constructs a multi-dimensional feature set, and performs fusion calculations via an AI workstation to generate a pulp fullness index that directly characterizes the abundance and distribution of flowable pulp. This fundamentally solves the problem of X-ray technology's inability to distinguish between liquid pulp and solid fruit flesh. Simultaneously, it improves detection accuracy through interference correction parameters and eliminates the need for massive image reconstruction calculations, significantly reducing computational load and adapting to the high-speed online detection requirements of sorting lines, providing reliable technical support for the industrial-scale quality grading of passion fruit.

[0021] The present application will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0022] The passion fruit pulp fullness detection method of this application, through methods such as... Figure 3The AI ​​workstation shown serves as the core computing platform, and its overall architecture includes at least a core computing module, a graphics processing expansion module, a storage module, a communication interface module, and a power supply module. This hardware structure represents a general configuration for AI workstations. Based on this, and specifically addressing the core requirements of this application—near-infrared spectral processing, multidimensional dataset construction, and rapid determination of sap fullness—some hardware structures have been adaptively adjusted to ensure the efficient, stable, and real-time execution of the detection method.

[0023] The core computing module, as the hardware core, is used to complete the analysis, preprocessing (noise reduction, normalization), multi-dimensional feature extraction, and feature fusion calculation of the raw near-infrared spectral signal. It can perform dimensional decomposition of the near-infrared spectral signal to construct a multi-dimensional dataset containing water morphology features, internal structure scattering features, and peel attenuation correction parameters, providing basic computing power support for the calculation of the juiciness index. The graphics processing extension module is used to run AI inference models, performing parallel inference and fast computation on the multi-dimensional spectral dataset, improving the mapping speed from spectral data to the juiciness index, breaking through the computing power bottleneck of existing technologies, and meeting the needs of industrial applications. The system meets the real-time requirements of line detection; the storage module temporarily stores the raw spectral data, intermediate processing results, and constructed multidimensional datasets uploaded in real time by the near-infrared spectral acquisition device, while also storing detection model parameters and historical detection records to ensure that the data read / write rate matches the spectral acquisition rate and avoid data loss or processing delays; the communication interface module establishes stable real-time communication with the near-infrared spectral acquisition device, enabling continuous reception of spectral data and feedback of detection results, and achieving coordinated linkage with the sorting line control system; the power supply module adopts an industrial-grade stable power supply design to ensure the reliability of the AI ​​workstation under long-term continuous operation in the sorting workshop.

[0024] In this embodiment, as an example, the AI ​​workstation adopts a tower structure, and its specific hardware configuration is as follows: The core computing module is configured with a dual-socket high-performance multi-core processor, supporting up to 64 cores and 128 threads for parallel analysis and feature extraction of multidimensional spectral data; the graphics processing expansion module is equipped with 5×PCIe 4.0×16 graphics card slots, supporting multi-GPU parallel computing, with maximum computing power sufficient to meet the efficient inference needs of AI models, and is compatible with domestically developed GPUs, adapting to domestic deployment in different scenarios. Furthermore, for the high-dimensional matrix operation requirements of near-infrared spectroscopy, GPU operation parameters are optimized to improve the processing efficiency of multidimensional datasets; the storage module includes 8×DDR4 ECC memory (expandable up to 2TB) and M.2... The system features NVMe high-speed solid-state drives and SATA hard disk drives. High-speed memory is used for real-time caching of raw spectral data and intermediate processing results, while the high-speed solid-state drives and hard disk drives are used for long-term storage of multidimensional datasets and model parameters. The communication interface module is equipped with dual 10 Gigabit adaptive Ethernet ports and multiple I / O interfaces such as USB 3.2 and Type-C. The network communication protocol is optimized for near-infrared spectral data transmission requirements, ensuring the stability and real-time performance of spectral data transmission. It can be directly connected to near-infrared spectral acquisition equipment in sorting lines. The power supply module uses an ATX 2000W high-power supply with an operating temperature range of 0-40℃, adapting to the environment of industrial sorting workshops and ensuring the stability of the hardware under full load.

[0025] It should be noted that the AI ​​workstation hardware configuration, tower structure, specific processor, interface specifications, etc., described in this embodiment are merely illustrative examples to adapt to the near-infrared spectral processing and multidimensional dataset construction requirements of this application, and should not constitute a limitation on the hardware structure, model, or configuration of the AI ​​workstation of this application. Any computing device that possesses the core computing capabilities, graphics processing capabilities, high-speed storage capabilities, and stable communication capabilities described in this application, and is capable of adapting to near-infrared spectral processing and multidimensional feature fusion calculation, falls within the scope of protection of this application.

[0026] Based on the above description of the AI ​​workstation hardware structure, the following section provides a detailed explanation of the complete application process of the passion fruit juice fullness detection method described in this application, taking into account the actual application scenario of automated fruit and vegetable sorting.

[0027] Please see Figure 4The detection method described in this application is applied to an intelligent automated passion fruit sorting line, which is the core equipment for post-harvest commercial processing. It is mainly used for integrated non-destructive testing and automated grading of passion fruit's appearance and internal quality. Its overall structure includes a feeding and conveying device, a single-fruit sorting device, an appearance visual inspection device, a near-infrared spectroscopy detection device, an AI workstation, and a grading and discharging device. The AI ​​workstation is communicatively connected to the appearance visual inspection device and the near-infrared spectroscopy detection device. These devices are sequentially connected and coordinated to achieve fully automated operation of the passion fruit process, from feeding, conveying, detection, analysis to grading and discharging.

[0028] The complete sorting process is as follows: After being sorted into individual fruits, the passion fruits are fed into fruit cups one by one. The fruit cups are mounted on a conveyor belt and are transported forward synchronously with the belt. The fruit cups can be continuously rotated during the transport process, causing the passion fruits to rotate synchronously and present different orientations during the transport process, providing a basis for subsequent comprehensive testing.

[0029] Subsequently, the passion fruit cups are sequentially placed into the appearance visual inspection device. The appearance visual inspection device uses a multi-view imaging unit to collect images of the passion fruit in its flipped state from all angles, obtaining appearance image information of the passion fruit without any blind spots. Based on the collected appearance image information, the AI ​​workstation comprehensively identifies and accurately judges the appearance quality indicators of the passion fruit, such as fruit diameter, peel color, surface defects and blemishes.

[0030] After the appearance inspection is completed, the cup containing the passion fruit is transported to the near-infrared spectroscopy detection device. The cup is continuously rotated, causing the passion fruit to rotate synchronously. This allows the near-infrared spectroscopy detection device to collect non-destructive spectral signals from all angles without blind spots, obtaining complete raw near-infrared spectral data of the passion fruit. This raw near-infrared spectral data is then transmitted to the AI ​​workstation in real time. The AI ​​workstation preprocesses the raw near-infrared spectral data, extracts multi-dimensional features, and performs feature fusion calculations to construct a multi-dimensional dataset related to the internal quality of the passion fruit. Based on this multi-dimensional dataset, the juice fullness detection result of the passion fruit is obtained, enabling accurate determination of the core internal quality indicators of the passion fruit.

[0031] Furthermore, the AI ​​workstation also uses near-infrared spectroscopy to obtain conventional internal quality parameters such as sweetness and acidity, and combines them with the juice fullness to jointly determine the internal quality of passion fruit. This ensures that the test results comprehensively cover the internal quality dimensions of passion fruit, providing a more comprehensive and accurate basis for subsequent grading, avoiding grading deviations caused by relying on only a single indicator. It also meets the actual needs of passion fruit quality evaluation, making the test results more valuable and practical.

[0032] After completing the appearance and internal quality inspections, the AI ​​workstation drives the corresponding actuators to sort the passion fruit from the fruit cups to the corresponding graded discharge devices according to the preset grading standards, thus completing the fully automated inspection and grading of the passion fruit.

[0033] Among them, the near-infrared spectroscopy detection device, the AI ​​workstation, and the process of collecting near-infrared spectra of passion fruit, constructing multidimensional datasets, and determining the fullness of juice are the key technical aspects that this application focuses on improving and protecting. These are used to solve the technical defects of existing X-ray detection technology, such as the inability to accurately identify liquid juice, low detection efficiency, and high computing power consumption. The continuous flipping structure of passion fruit achieved by the fruit cup and belt is a key supporting structure for the sorting line of this application to achieve all-round detection without blind spots, ensuring the comprehensiveness and accuracy of appearance and internal quality detection.

[0034] In this embodiment, the intelligent automated sorting line for passion fruit can refer to the structural layout of existing intelligent sorting equipment for fruits and vegetables. The difference is that this application replaces the internal quality detection object with the sugar content detection to the passion fruit juice fullness detection, and specifically configures a near-infrared spectroscopy detection device and a suitable AI workstation. At the same time, it relies on the fruit cup flipping structure to achieve all-round spectral acquisition, so as to achieve accurate and efficient detection of the unique internal quality indicators of passion fruit.

[0035] The above describes the application scenarios and overall implementation path of the passion fruit pulp fullness detection method of this application, based on the overall structure and workflow of the intelligent automated passion fruit sorting line. To enable those skilled in the art to more clearly and completely understand the technical solution of this application, the specific implementation steps, data processing flow, and parameter setting methods of the passion fruit pulp fullness detection method described in this application are described in detail below with reference to specific embodiments. This embodiment is only used as an illustrative example of the technical solution of this application and does not constitute a limitation on the scope of protection of this application.

[0036] Please see Figure 5 , Figure 5 This is a flowchart illustrating a method for detecting the fullness of passion fruit juice based on an AI workstation. This method is applied to applications such as... Figure 3 The AI ​​workstation shown in the figure includes the following steps S501-S504: Step S501: Obtain multiple near-infrared diffuse reflectance spectra of a single passion fruit.

[0037] This step is performed in the internal quality inspection stage of the passion fruit intelligent automated sorting line. In the corresponding sorting process, after the passion fruit is sorted into individual fruits, transported in fruit cups, and undergoes visual inspection, it enters the stage of the near-infrared spectroscopy detection device.

[0038] Specifically, after being sorted into individual passion fruits, they are sequentially fed into fruit cups. The fruit cups are mounted on a belt and move forward synchronously with the conveyor line. During the conveying process, the fruit cups continuously rotate, causing the passion fruits to rotate synchronously and present different orientations during the conveying process. When the fruit cups containing passion fruits enter the detection area of ​​the near-infrared spectroscopy detection device, the near-infrared spectroscopy detection device collects near-infrared diffuse reflectance spectra of the passion fruits in different orientations from multiple angles, thereby obtaining multiple near-infrared diffuse reflectance spectra of a single passion fruit. This provides a complete and comprehensive raw data foundation for subsequent multi-dimensional feature extraction and juice fullness calculation.

[0039] Please refer to Figure 6 , Figure 9 , Figure 6 This is a schematic diagram of a near-infrared spectroscopy detection device. Figure 9 This is a schematic diagram of the near-infrared diffuse reflectance inside a near-infrared spectroscopy detection device. This device is a dedicated instrument for acquiring near-infrared diffuse reflectance spectra. It adopts a modular design and internally includes at least: a near-infrared emission module, a spectral acquisition module, a signal processing module, and a data transmission module. These modules work together to achieve stable acquisition, processing, and transmission of near-infrared spectra, as detailed below: Near-infrared emission module: As the core of the light source for spectral acquisition, it has a built-in near-infrared light source and optical path guiding components. It is used to emit light signals with stable intensity and wavelength covering the range required for near-infrared detection, accurately adapting to the spectral response requirements of the relevant characteristics of the pulp inside the passion fruit. The light signal is then directionally transmitted to the surface of the passion fruit through the optical path components.

[0040] Spectral acquisition module: The core is a spectral detector, equipped with an optical fiber assembly, which is used to receive near-infrared light signals after diffuse reflection from the internal tissue of passion fruit. The optical fiber assembly stably transmits the light signals, and the spectral detector converts the light signals into processable electrical signals, accurately capturing the differences in spectral response at different wavelengths, providing the original foundation for subsequent signal processing.

[0041] Signal processing module: Integrates core functions such as signal amplification and analog-to-digital conversion. It is used to initially amplify the raw electrical signal output by the spectrum acquisition module to compensate for the weak strength of the raw signal. Then, it converts the amplified analog electrical signal into standardized digital spectral data to ensure that the data can be recognized and processed by subsequent equipment.

[0042] Data transmission module: It has a built-in communication interface for transmitting near-infrared spectral data processed by the signal processing module to the AI ​​workstation in real time. It is precisely adapted to the communication interface module of the AI ​​workstation to ensure the real-time performance and stability of data transmission and realize the synchronous linkage between spectral data and AI workstation.

[0043] By combining the modular structure with the fruit cup flipping structure, the near-infrared spectroscopy detection device can collect multiple near-infrared diffuse reflectance spectra from different parts and orientations of a single passion fruit. This avoids the defect that a single-view spectrum cannot fully characterize the distribution of juice inside the passion fruit, ensuring the accuracy and reliability of subsequent juice fullness detection from the data source.

[0044] Step S502: Obtain the water morphology characteristic parameters and internal structure scattering parameters of passion fruit based on the near-infrared diffuse reflectance spectrum, and construct a multi-dimensional characteristic parameter set by combining the fruit density characteristic parameters and peel attenuation interference correction parameters of passion fruit.

[0045] In this step, the AI ​​workstation extracts the water speciation characteristic parameters and internal structure scattering parameters of passion fruit based on near-infrared diffuse reflectance spectroscopy. These parameters are then combined with fruit density characteristic parameters and peel attenuation interference correction parameters to construct a multi-dimensional feature parameter set. Among these: The water form characteristic parameter is used to characterize the relative ratio of free water and cell-bound water inside passion fruit, reflecting the form and content level of the juice. The larger the parameter value, the more fluid the juice, and the smaller the value, the less juice and the drier the fruit.

[0046] The peel attenuation correction parameter is used to counteract the absorption and attenuation interference of the peel on near-infrared light and eliminate the detection deviation caused by the difference in peel thickness. The thicker the peel, the greater the correction range of this parameter, making the detection results more accurate.

[0047] Internal structure scattering parameters are used to quantify the uniformity of passion fruit pulp, the degree of cavity and shrinkage disturbance, and reflect the influence of pulp structure on juice distribution; the higher the parameter value, the more uneven the internal structure and the more obvious the cavity or shrinkage.

[0048] Fruit density is a characteristic parameter used to characterize the overall fullness and plumpness of passion fruit, and to help improve the stability of internal quality assessment; normal density indicates that the fruit is plump, while low density usually indicates that the fruit is hollow or has loose flesh.

[0049] The above parameters together constitute a multi-dimensional feature parameter set, providing a comprehensive and stable data foundation for subsequent feature fusion calculations and slurry fullness determination.

[0050] In some embodiments, the water speciation characteristic parameters are obtained through the following steps: averaging the multiple near-infrared diffuse reflectance spectra to obtain an average spectrum; locating the characteristic absorption peaks corresponding to free water and cell-bound water based on the average spectrum, and fitting the characteristic absorption peaks to obtain the actual absorbance of the two characteristic absorption peaks; correcting the actual absorbance based on the peak shape correction factor of the characteristic absorption peaks to eliminate errors caused by peak shape distortion; the peak shape correction factor is obtained by calibration using standard passion fruit samples and determined through iterative optimization of fitting residuals; and calculating the water speciation characteristic parameters based on the corrected actual absorbance, wherein the water speciation characteristic parameters are used to characterize the relative proportion of free water and cell-bound water.

[0051] The water speciation characteristic parameters were obtained through systematic analysis and data processing of near-infrared diffuse reflectance spectra. The core principle is to utilize the characteristic interaction between near-infrared light and the internal water (free water and cell-bound water) of passion fruit. Through the extraction, fitting, correction, and calculation of spectral signals, characteristic parameters that can accurately characterize the relative proportion of free water and cell-bound water are finally obtained, providing core data support for the subsequent calculation of the juice fullness index. The specific implementation method is as follows: First, the average spectrum of the near-infrared diffuse reflectance spectra of a single passion fruit collected in step S501 is obtained by averaging multiple near-infrared diffuse reflectance spectra. These multiple near-infrared diffuse reflectance spectra are spectral data collected from different orientations and parts of the passion fruit during the fruit cup rotation process. Arithmetic averaging effectively reduces random noise, local texture differences on the fruit surface, minor fluctuations in the optical path, and spectral deviations caused by fruit cup rotation during the detection process, thus improving the stability and accuracy of the spectral signal. The average spectrum is calculated using the following formula: ; in, Let λ be the average spectral intensity at wavelength λ, and n be the total number of near-infrared diffuse reflectance spectra collected from a single fruit. Let be the spectral intensity of the i-th spectrum at wavelength λ, where λ is any wavelength point within the near-infrared detection band.

[0052] Subsequently, the characteristic absorption peaks corresponding to free water and cell-bound water were located based on the obtained average spectrum. In the near-infrared band, the overtones and combination frequencies of the O-H stretching vibrations of water have clear and industry-recognized characteristic absorption positions, which are the core basis for distinguishing between free water and cell-bound water. The characteristic absorption wavelength corresponding to free water is 1450 nm, and the characteristic absorption wavelength corresponding to cell-bound water is 1940 nm. Based on the average spectrum, the positions of the two characteristic absorption peaks were located through baseline correction and peak identification algorithms, and the spectral intensity range of the two absorption peaks was determined.

[0053] Next, the two located characteristic absorption peaks are fitted to obtain their actual absorbance. Since the average spectrum contains background baseline interference, superposition interference from adjacent bands, and noise interference, the true absorbance of the characteristic absorption peaks cannot be directly read. Therefore, a Gaussian fitting function is used to fit the peak shapes of the two characteristic absorption peaks separately, subtracting the background baseline and interference signals. The fitting formula is as follows: ; in, The value is the fitted spectral intensity, and 'a' is the peak height of the characteristic absorption peak. σ represents the center wavelength of the characteristic absorption peaks (1450 nm for free water and 1940 nm for cell-bound water), b represents the peak width coefficient, and b represents the background baseline intensity. The actual absorbance of the characteristic absorption peaks of free water was obtained through fitting calculations. The actual absorbance of the characteristic absorption peak of cell-bound water .

[0054] Subsequently, the actual absorbance is corrected for deviation based on the peak shape correction factor of the characteristic absorption peak to eliminate errors caused by peak shape distortion. In actual testing, factors such as passion fruit peel scattering, light path offset, and slight changes in ambient temperature can cause slight distortion of the characteristic absorption peak, thus affecting the accuracy of absorbance. Therefore, deviation correction using a peak shape correction factor is necessary. The correction formula is as follows: ; In the above formula, , These are the corrected absorbance values ​​for free water and bound water, respectively. , These are peak shape correction factors for the characteristic absorption peaks of free water and cell-bound water, respectively. The peak shape correction factors were obtained by calibration of standard passion fruit samples and determined by iterative optimization through fitting residuals to ensure that the corrected absorbance can truly reflect the internal water state of the fruit.

[0055] Finally, the water speciation characteristic parameters are calculated based on the corrected actual absorbance. These parameters characterize the relative ratio of free water to cell-bound water, and the calculation uses a ratio formula, as follows: ; Here, M represents the water form characteristic parameter; a larger M value indicates a higher proportion of free water inside the passion fruit and more abundant flowable juice; a smaller M value indicates a lower proportion of free water and less or drier juice. This parameter directly reflects the form and content level of passion fruit juice and is the core input parameter for subsequent juice fullness index calculation.

[0056] In some embodiments, the peel attenuation correction parameter is obtained through the following steps: extracting the baseline slope of the characteristic response band of the peel in the average spectrum, wherein the baseline slope is positively correlated with the peel thickness; calculating the peel thickness based on the baseline slope and a pre-established correlation between the peel thickness and the baseline slope, wherein the correlation is calibrated by fitting standard samples of different peel thicknesses; and calculating the peel attenuation correction parameter based on the peel thickness using an adaptive compensation mechanism.

[0057] The peel attenuation correction parameter is obtained through baseline feature analysis and adaptive compensation calculation of the average spectrum. Its core principle is to establish the correlation between peel thickness and the spectral baseline by utilizing the spectral response law of the peel in the near-infrared band. The peel thickness is then inverted through the baseline slope, and an adaptive attenuation correction amount is generated accordingly to offset the absorption, scattering, and attenuation interference of peels of different thicknesses on near-infrared light, ensuring the accuracy of subsequent detection of water morphology parameters and juice fullness. The specific implementation method is as follows: First, the baseline slope of the characteristic response band of the peel in the average spectrum obtained in the preceding steps is extracted. The peel exhibits stable spectral absorption and scattering characteristics within the near-infrared detection range, corresponding to a clearly defined characteristic response band. The trend of the spectral baseline within this band directly reflects the thickness and density of the peel, and the baseline slope is positively correlated with peel thickness; that is, the thicker the peel, the greater the baseline slope; the thinner the peel, the smaller the baseline slope. By performing baseline fitting and differentiation on the spectrum in this band, the baseline slope value of the corresponding average spectrum is obtained.

[0058] Subsequently, based on the baseline slope, the peel thickness is calculated using a pre-established correlation between peel thickness and the baseline slope. This correlation was obtained by collecting a large number of standard passion fruit samples of different varieties, maturity levels, and peel thicknesses, measuring their actual peel thickness, and collecting corresponding near-infrared spectra. After fitting and calibration, it possesses uniqueness and stability. By substituting the calculated baseline slope into this correlation, the peel thickness of the current passion fruit to be tested can be quickly retrieved.

[0059] Finally, based on the peel thickness, an adaptive compensation mechanism is used to calculate the peel attenuation correction parameters. The adaptive compensation mechanism automatically adjusts the correction intensity according to the peel thickness; the thicker the peel, the stronger the attenuation effect on near-infrared light, and the larger the corresponding correction parameter compensation amount; the thinner the peel, the smaller the correction amount.

[0060] By using the peel attenuation correction parameter, the water morphology characteristic parameters can be corrected in real time during subsequent calculations, effectively eliminating the detection bias caused by the difference in peel thickness, and making the final juice fullness result more realistic and stable.

[0061] In some embodiments, the internal structure scattering parameters are obtained through the following steps: calculating the absorbance difference at each wavelength point across the entire near-infrared diffuse reflectance spectrum of multiple near-infrared spectra, wherein the absorbance difference is used to characterize the degree of local scattering fluctuation in the single fruit spectrum; integrating the absorbance difference across the entire spectrum to calculate the internal structure scattering parameters, wherein the internal structure scattering parameters are used to quantify the uniformity of the passion fruit pulp, the degree of cavity and shrinkage disturbance.

[0062] The internal structure scattering parameters are obtained through fluctuation difference analysis and integral calculation of near-infrared diffuse reflectance spectra from multiple angles. Addressing the characteristic that structural differences in passion fruit pulp density, cavities, and shrinkage cause fluctuations in near-infrared light scattering, this application characterizes local scattering perturbations through absorbance differences between multiple angle spectra and quantifies the overall structural uniformity through full-band integration. This accurately reflects the impact of internal cavities, shrinkage, and looseness on the juice fullness. The specific implementation method is as follows: First, the absorbance difference of multiple near-infrared diffuse reflectance spectra at each wavelength point across the entire wavelength range was calculated. Multiple diffuse reflectance spectra were collected from the same passion fruit during the turning process. The differences in spectra at different angles were mainly caused by uneven internal pulp structure, cavities, and localized drying. For each identical wavelength point across the entire wavelength range, the absorbance difference between the multiple spectra was calculated. : ; in, , These represent the absorbance of two diffuse reflectance spectra at different angles for the same wavelength. This is used to characterize the degree of local scattering fluctuation of a single fruit at a given wavelength. The more uneven the flesh, the more pronounced the cavities or shrunkenness, and the greater the spectral differences at various angles. The higher the overall value.

[0063] Subsequently, the absorbance difference across the entire wavelength range is integrated to obtain the internal structure scattering parameters. The absorbance differences at all wavelengths across the entire wavelength range are then integrated and summed to obtain the characteristic parameter D, which can comprehensively characterize the uniformity of the internal structure. ; in, , These are the start and end wavelengths of the near-infrared detection band, respectively, and D is the internal structure scattering parameter. The internal structure scattering parameter is used to quantify the uniformity of the passion fruit pulp, the degree of cavity and shrinkage disturbance; the larger the D value, the more uneven the pulp structure, and the more significant the cavity, shrinkage or local looseness phenomena; the smaller the D value, the denser and more uniform the pulp structure, and the more stable the juice distribution.

[0064] In some embodiments, the fruit density characteristic parameters are obtained through the following steps: acquiring a three-dimensional contour image of passion fruit, extracting the external size parameters of passion fruit and calculating the actual volume of passion fruit; obtaining the weight of a single passion fruit, and calculating the fruit density characteristic parameters in combination with the actual volume.

[0065] Among them, the fruit density characteristic parameters are obtained by combining three-dimensional contour detection and weight detection.

[0066] In some embodiments, the fruit cup is also equipped with a weight detection device for real-time acquisition of the weight of a single passion fruit. As the passion fruit moves with the conveyor mechanism, the weight detection device can simultaneously acquire the weight of a single fruit, working in conjunction with 3D contour image acquisition and near-infrared spectroscopy detection to achieve simultaneous acquisition of volume, weight, and spectral information. This application uses fruit density as an external fullness evaluation index, which is corroborated with near-infrared spectral characteristics, thereby improving the reliability of the juice fullness determination. The specific implementation method is as follows: First, a 3D contour image of the passion fruit is acquired. Contour analysis is then used to extract the fruit's external dimensions, allowing for the calculation of its actual volume. Using a 3D vision approach can accommodate the irregular shape of the passion fruit, avoiding volume calculation errors caused by 2D images and accurately reflecting the overall size of the fruit.

[0067] Subsequently, the weight of a single passion fruit was obtained, and combined with the obtained actual volume, the fruit density characteristic parameter was calculated. Fruit density directly reflects the fullness of the pulp and is an important external basis for judging whether the fruit is hollow, shriveled, or has loose pulp.

[0068] Fruit density characteristic parameters are used to characterize the overall fullness and plumpness of passion fruit, and help improve the stability of internal quality judgment. Normal density indicates that the fruit is full and plump, while low density indicates that there may be hollow, loose or insufficient juice inside, which can provide a reliable basis for subsequent correction of the juice fullness index.

[0069] Step S503: The multi-dimensional feature parameters are fused and calculated to obtain the juice fullness index, which characterizes the relative abundance and distribution of the flowable juice inside the passion fruit.

[0070] This step organically integrates the water speciation parameters, peel attenuation correction parameters, internal structure scattering parameters, and fruit density parameters obtained in the preceding steps. By comprehensively considering the internal information of the fruit reflected by these various dimensions, the limitations of a single parameter are eliminated, ultimately resulting in a juice fullness index that can comprehensively and stably characterize the relative abundance and uniform distribution of the flowable juice inside the passion fruit. This index comprehensively considers the combined effects of juice content, pulp structure, peel interference, and fruit fullness, objectively reflecting the true juice fullness of the passion fruit and providing a unified and intuitive evaluation basis for subsequent grading.

[0071] In specific implementations, in some embodiments, the process of fusing and calculating the multi-dimensional feature parameters to obtain a juice fullness index characterizing the relative abundance and distribution of the fluid juice inside the passion fruit includes: correcting the water form feature parameters according to the peel attenuation correction parameters to eliminate the interference of peel thickness on the water form feature parameters; fusing the corrected water form feature parameters with the fruit density feature parameters to obtain a basic fusion value, which is used to characterize the basic reserve of obtainable juice and the fullness of the pulp inside the passion fruit; correcting the basic fusion value according to the internal structure scattering parameters to weaken the interference of internal cavities, shrinkage, and uneven juice distribution on the determination of juice fullness; and optimizing the corrected basic fusion value overall by combining pre-trained adaptive weight coefficients to obtain the juice fullness index.

[0072] The process of fusing and calculating multi-dimensional feature parameters to obtain the juice fullness index is constructed as an optimization problem aimed at accurately characterizing the juice fullness inside passion fruit. This process achieves iterative optimization by combining a loss function with the dual objectives of maximizing the contribution of effective juice information and minimizing structural anomalies and external disturbances, resulting in a more stable, accurate, and robust juice fullness index.

[0073] This embodiment transforms the solution of the slurry filling index into a constrained optimization problem. The overall optimization objective is to maximize the effective feature contribution and minimize structural disturbance and detection bias. The objective function is expressed as: ; in, This indicates that the sap filling index JFI is used as the optimization variable, and the objective function is... The solution is maximized; ω is the global adaptive weight coefficient. , λ represents the specific contribution weights of moisture and density characteristics, respectively; λ is the structural anomaly penalty coefficient; and μ is the regularization coefficient. The effective characteristic of water content after peel correction is as follows: Contribution to fruit density This is a penalty term for internal structural disturbances. This is the overall loss function.

[0074] To achieve iterative calibration and error control of the model, the loss function is constructed as follows: ; in, The slurry filling index is the output of the current model. Here, D represents the true index of the standard sample label, and D represents the internal structure scattering parameter. Let be the mean scattering parameters corresponding to a normal fruit structure, and α be the structure regularization coefficient. This loss function can continuously reduce the deviation between predicted and true values ​​during the iteration process, while suppressing interference from structural anomalies such as internal cavities and shrinkage.

[0075] In the optimization process, the water speciation characteristic parameters are first corrected according to the peel attenuation correction parameter to eliminate the interference of peel thickness on the spectral signal and obtain the water characteristics that truly reflect the juice content: ; Subsequently, the corrected moisture characteristics and fruit density characteristics were treated as independent effective contributors, with the fruit density characteristics being directly used as the density contributor to characterize the overall fullness and plumpness of the fruit. ; Based on this, a perturbation penalty term is constructed according to the internal structure scattering parameters to mitigate the impact of abnormalities such as cavities, shrinkage, and uneven pulp distribution. ; After determining each component, the system will perform iterative calculations based on the aforementioned optimization objective and loss function. During the iteration process, the system will simultaneously verify each constraint to ensure that the optimization results conform to the physical meaning and the actual detection scenario. Specific constraints include: Parameter value constraints: Set reasonable value ranges for each input parameter to avoid interference from outliers, i.e., 0. <M≤M max 0 <K≤K max 0≤D≤D max ρ min <ρ≤ρ max The thresholds were obtained by statistical calibration of a large number of standard samples of passion fruit of different varieties and maturity levels. Physical constraints: ensuring that valid feature terms are positive and interference penalty terms are non-negative, i.e. >0、 >0、 ≥0 ensures that the optimization results conform to the actual logic that "sap content and fullness contribute positively, while structural anomalies are penalized negatively"; Output range constraint of the index: Set a reasonable output range for the slurry filling index, i.e., JFI∈[J min J max This range is calibrated by the actual pulp fullness of the standard sample to ensure the consistency and comparability of the test results; Iterative convergence constraint: Set a convergence threshold for the loss function. When the difference between the loss functions of two consecutive iterations is less than a preset threshold (e.g., 10), the convergence is achieved. -3 ), and the loss function value When the value is ≤0.05, the iteration is considered to have converged, and the calculation is stopped.

[0076] When the iteration converges and all constraints are satisfied, the final slurry filling index is output: ; The juice fullness index obtained through the above optimization method can comprehensively reflect the relative abundance, distribution uniformity, and pulp fullness of the fluid juice inside the passion fruit. Compared with the conventional simple fusion method, it has higher accuracy and stability and can provide a reliable evaluation basis for subsequent grading.

[0077] Step S504: The passion fruit juice fullness is graded and determined according to the juice fullness index.

[0078] This step, based on the juice fullness index (JFI) obtained in step S503, and combined with grading threshold ranges pre-calibrated using a large number of standard samples, classifies passion fruit into different juice fullness grades, achieving automated and standardized grading of fruit quality. During the grading process, the system compares the real-time calculated JFI with the preset thresholds, matches the corresponding grades, and outputs the grading results.

[0079] Furthermore, the AI ​​workstation can combine the appearance inspection results (such as fruit shape, color, blemishes, and damage) detected by conventional technical means with other internal quality results (such as sweetness, acidity, and soluble solids content) and juice fullness index to conduct multi-dimensional fusion evaluation and jointly determine the overall quality of a single passion fruit.

[0080] In some embodiments, the step of determining the grading of passion fruit juice fullness based on the juice fullness index includes: determining the grading threshold of the juice fullness index and the structural anomaly threshold of the internal structure scattering parameter based on sample training; and classifying the passion fruit juice fullness into four levels: full juice, normal juice, insufficient juice / dry, and hollow / empty shell / severely shriveled, based on the threshold range of the juice fullness index and whether the internal structure scattering parameter exceeds the structural anomaly threshold.

[0081] It should be noted that, in this embodiment, the core function of the juice fullness index is to quantitatively characterize the concentration of the flowable juice in the passion fruit pulp itself. The internal structure scattering parameter is introduced in its calculation process to correct the interference of structural factors such as loose pulp and local cavities on the juice content detection signal, eliminate the detection deviation caused by structural interference, and ensure that the juice fullness index can truly reflect the degree of juice enrichment in the pulp itself, rather than a false value diluted by the sparse structure and local voids.

[0082] The internal structure scattering parameter is introduced again in the grading stage. Its function is completely different from that in the calculation stage. Its core purpose is to determine whether there are structural defects such as large areas of hollowness, empty shells, or severe shriveling in the overall structure of the fruit. The juice filling index can only reflect the juice concentration of the pulp itself and cannot distinguish between two types of fruits: "normal juice concentration but hollow overall" and "normal juice concentration and dense structure". Therefore, it is necessary to supplement the judgment by internal structure scattering parameter to clarify the integrity of the overall structure of the fruit.

[0083] This can be further illustrated with specific examples of detection scenarios: Fruit A: The juiciness index is within the normal range, and the internal structure scattering parameters do not exceed the abnormal threshold, indicating that the pulp juiciness concentration is normal and the structure is dense, and it is judged to have normal juiciness. Fruit B: The juiciness index is also within the normal range, but the internal structure scattering parameter is significantly out of limit, indicating that the juiciness concentration of the pulp itself is normal, but there is a large area of ​​hollowness. It is judged as hollow / empty shell / severely shriveled to avoid misjudgment. Fruit C: The juiciness index is low, and the internal structure scattering parameters are normal, indicating that the pulp structure is dense but the juiciness concentration is insufficient. It is judged as insufficient juiciness / dryness. Fruit D: The juiciness index is high and the internal structure scattering parameters are normal, indicating that the juiciness concentration is high and the structure is dense, thus it is judged to be full of juiciness.

[0084] In summary, the internal structure scattering parameter is introduced in the calculation stage to "correct the signal and accurately calculate the pulp and juice concentration," while the parameter is introduced in the grading stage to "identify overall structural defects and distinguish fruits of different quality types." The two have completely different uses, stages, and purposes and are not used repeatedly, but rather complement each other to achieve accurate grading of passion fruit pulp fullness, ensuring that the grading results reflect both the pulp content and the actual structural state of the fruit.

[0085] In some embodiments, before classifying passion fruit sap fullness into four levels—full sap, normal sap, insufficient sap / dry, and hollow / empty shell / severely shriveled—based on the threshold range of the sap fullness index and whether the internal structure scattering parameter exceeds the structural abnormality threshold, the method further includes: when the fruit density characteristic parameter is lower than a set threshold, adjusting the sap fullness index downwards, wherein the set threshold is determined by calibration based on the normal density range of the corresponding passion fruit variety.

[0086] In actual testing, fruit density parameters directly reflect the firmness, pulp fullness, and water loss of passion fruit. For passion fruit of the same variety and at the same maturity, lower fruit density generally indicates looser pulp and poorer overall fullness. Even if the calculated juice fullness index is within the normal range, the actual edible juice volume and pulp quality are still lower than normal fruit. Therefore, adding a pre-correction step based on fruit density parameters before final grading can further improve the consistency between the juice fullness index and actual quality.

[0087] Specifically, the AI ​​workstation pre-calculates and determines a reasonable range of fruit density based on a large number of normal samples of the corresponding passion fruit variety, and sets a density judgment threshold accordingly. After calculating the juiciness index and before formally proceeding to grading, the density characteristic parameters of the fruit to be tested are compared with this set threshold. If the density characteristic parameters are not lower than the threshold, it indicates that the overall density of the fruit is normal, and the original juiciness index can be directly used for subsequent grading. If the density characteristic parameters are lower than the threshold, it indicates that the overall fullness of the fruit is insufficient, and there are issues such as being too loose, too light, or premature water loss. In this case, the original juiciness index is adjusted downward to make the corrected index more closely reflect the actual juiciness level of the fruit.

[0088] By using the above density pre-correction steps, misjudgments caused by a low overall fruit density leading to a high juice filling index can be effectively avoided. This allows the index to simultaneously take into account both the juice concentration inside the pulp and the overall fullness of the fruit, further improving the accuracy and reliability of subsequent grading.

[0089] As can be seen, in this embodiment, the AI ​​workstation acquires near-infrared spectral data of passion fruit, extracts multi-dimensional feature parameters such as water form and internal scattering, constructs a multi-dimensional feature parameter set, and combines fruit density feature parameters and peel attenuation interference correction parameters to construct the multi-dimensional feature parameter set; the multi-dimensional feature parameters are fused and calculated to obtain the juice fullness index, which characterizes the relative abundance and distribution of the flowable juice inside the passion fruit; the grading is completed based on the juice fullness index. It is evident that this application, through the processing of near-infrared diffuse reflectance spectra by the AI ​​workstation and the fusion of multi-dimensional feature parameters, can achieve simultaneous and non-destructive characterization of the internal water state, pulp scattering, fruit density, and peel attenuation of passion fruit, accurately reflecting the relative abundance and distribution uniformity of the juice, reducing the interference of the peel and internal cavity on the detection signal, improving the stability and accuracy of juice fullness determination, and meeting the physical detection requirements for rapid online grading; the detection process is adapted to the real-time data processing capabilities of the AI ​​workstation, without requiring additional computing power, and can quickly respond to the real-time detection needs of industrial sorting. This solves the problem that traditional testing only focuses on structural state and cannot accurately reflect the condition of the pulp.

[0090] The above primarily describes the solutions of the embodiments of this application from the perspective of the method execution process. It is understood that, in order to achieve the above functions, the server includes the corresponding hardware structure and / or software modules for executing each function. Those skilled in the art should readily recognize that, based on the units and algorithm steps of the examples described in the embodiments provided herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0091] This application embodiment can divide the server into functional units according to the above method example. For example, each function can be divided into different functional units, or two or more functions can be integrated into one processing module. The integrated unit can be implemented in hardware or as a software program module. It should be noted that the unit division in this application embodiment is illustrative and only represents a logical functional division, while other division methods may be used in actual implementation.

[0092] In the case of using integrated units, please refer to Figure 7 , Figure 7 This is a functional unit structure diagram of a passion fruit juice fullness detection device based on an AI workstation. The passion fruit juice fullness detection device 7 based on the AI ​​workstation includes: Acquisition unit 701 is used to acquire multiple near-infrared diffuse reflectance spectra of a single passion fruit; Processing unit 702 is used to obtain the water morphology characteristic parameters and internal structure scattering parameters of passion fruit based on the near-infrared diffuse reflectance spectrum, and to construct a multi-dimensional characteristic parameter set by combining the fruit density characteristic parameters and peel attenuation interference correction parameters of passion fruit; to perform fusion calculation on the multi-dimensional characteristic parameters to obtain the juice fullness index, which characterizes the relative abundance and distribution of the flowable juice inside the passion fruit; and to complete the grading and determination of the juice fullness of passion fruit based on the juice fullness index.

[0093] As can be seen, in this embodiment, the AI ​​workstation acquires near-infrared spectral data of passion fruit, extracts multi-dimensional feature parameters such as water form and internal scattering, constructs a multi-dimensional feature parameter set, and combines fruit density feature parameters and peel attenuation interference correction parameters to construct the multi-dimensional feature parameter set; the multi-dimensional feature parameters are fused and calculated to obtain the juice fullness index, which characterizes the relative abundance and distribution of the flowable juice inside the passion fruit; the grading is completed based on the juice fullness index. It is evident that this application, through the processing of near-infrared diffuse reflectance spectra by the AI ​​workstation and the fusion of multi-dimensional feature parameters, can achieve simultaneous and non-destructive characterization of the internal water state, pulp scattering, fruit density, and peel attenuation of passion fruit, accurately reflecting the relative abundance and distribution uniformity of the juice, reducing the interference of the peel and internal cavity on the detection signal, improving the stability and accuracy of juice fullness determination, and meeting the physical detection requirements for rapid online grading; the detection process is adapted to the real-time data processing capabilities of the AI ​​workstation, without requiring additional computing power, and can quickly respond to the real-time detection needs of industrial sorting. This solves the problem that traditional testing only focuses on structural state and cannot accurately reflect the condition of the pulp.

[0094] In some embodiments, the water speciation characteristic parameters are obtained through the following steps: averaging the multiple near-infrared diffuse reflectance spectra to obtain an average spectrum; locating the characteristic absorption peaks corresponding to free water and cell-bound water based on the average spectrum, and fitting the characteristic absorption peaks to obtain the actual absorbance of the two characteristic absorption peaks; correcting the actual absorbance based on the peak shape correction factor of the characteristic absorption peaks to eliminate errors caused by peak shape distortion; the peak shape correction factor is obtained by calibration using standard passion fruit samples and determined through iterative optimization of fitting residuals; and calculating the water speciation characteristic parameters based on the corrected actual absorbance, wherein the water speciation characteristic parameters are used to characterize the relative proportion of free water and cell-bound water.

[0095] In some embodiments, the peel attenuation correction parameter is obtained through the following steps: extracting the baseline slope of the characteristic response band of the peel in the average spectrum, wherein the baseline slope is positively correlated with the peel thickness; calculating the peel thickness based on the baseline slope and a pre-established correlation between the peel thickness and the baseline slope, wherein the correlation is calibrated by fitting standard samples of different peel thicknesses; and calculating the peel attenuation correction parameter based on the peel thickness using an adaptive compensation mechanism.

[0096] In some embodiments, the internal structure scattering parameters are obtained through the following steps: calculating the absorbance difference at each wavelength point across the entire near-infrared diffuse reflectance spectrum of multiple near-infrared spectra, wherein the absorbance difference is used to characterize the degree of local scattering fluctuation in the single fruit spectrum; integrating the absorbance difference across the entire spectrum to calculate the internal structure scattering parameters, wherein the internal structure scattering parameters are used to quantify the uniformity of the passion fruit pulp, the degree of cavity and shrinkage disturbance.

[0097] In some embodiments, the fruit density characteristic parameters are obtained through the following steps: acquiring a three-dimensional contour image of passion fruit, extracting the external size parameters of passion fruit and calculating the actual volume of passion fruit; obtaining the weight of a single passion fruit, and calculating the fruit density characteristic parameters in combination with the actual volume.

[0098] In some embodiments, in the process of fusing and calculating the multi-dimensional feature parameters to obtain the juice fullness index, which characterizes the relative abundance and distribution of the fluid juice inside the passion fruit, the processing unit 702 is configured to: correct the water form feature parameters according to the peel attenuation correction parameters to eliminate the interference of peel thickness on the water form feature parameters; fuse the corrected water form feature parameters with the fruit density feature parameters to obtain a basic fusion value, which is used to characterize the basic reserve of obtainable juice and the fullness of the pulp inside the passion fruit; correct the basic fusion value according to the internal structure scattering parameters to weaken the interference of internal cavities, shrinkage, and uneven juice distribution on the determination of juice fullness; and optimize the corrected basic fusion value as a whole by combining pre-trained adaptive weight coefficients to obtain the juice fullness index.

[0099] In some embodiments, in determining the grading of passion fruit juice fullness based on the juice fullness index, the processing unit 702 is configured to: determine the grading threshold of the juice fullness index and the structural anomaly threshold of the internal structure scattering parameter based on sample training; and classify the passion fruit juice fullness into four levels: full juice, normal juice, insufficient juice / dry, and hollow / empty shell / severely shriveled, based on the threshold range of the juice fullness index and whether the internal structure scattering parameter exceeds the structural anomaly threshold.

[0100] In some embodiments, before classifying passion fruit sap fullness into four levels—full sap, normal sap, insufficient sap / dry, and hollow / empty shell / severely shriveled—based on the threshold range of the sap fullness index and whether the internal structure scattering parameter exceeds the structural abnormality threshold, the processing unit 702 is used to: adjust the sap fullness index downward when the fruit density characteristic parameter is lower than a set threshold, wherein the set threshold is determined by calibration based on the normal density range of the corresponding passion fruit variety.

[0101] Please see Figure 8 , Figure 8 This is a schematic diagram of the server structure provided in the embodiments of this application, such as... Figure 8 As shown, the server 8 includes a processor 801, a memory 803, a communication interface 802, and a computer program 8031. The computer program 8031 ​​is stored in the memory 803 and configured to be executed by the processor 801. The program includes a method for performing a passion fruit juice fullness detection method based on an AI workstation as described in the above embodiments.

[0102] This application provides a computer-readable storage medium storing a computer program / instructions thereon, which, when executed by a processor, implement the steps of any possible embodiment of the method.

[0103] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0104] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0105] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical or other forms.

[0106] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0107] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0108] If the aforementioned integrated units are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0109] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage device, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0110] The embodiments of this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for detecting the fullness of passion fruit pulp based on an AI workstation, characterized in that, Applied to the AI ​​workstation, the method includes: Obtain multiple near-infrared diffuse reflectance spectra of a single passion fruit; Based on the near-infrared diffuse reflectance spectrum, the water morphology characteristic parameters and internal structure scattering parameters of passion fruit were obtained. Combined with the fruit density characteristic parameters and peel attenuation interference correction parameters of passion fruit, a multi-dimensional characteristic parameter set was constructed. The multi-dimensional feature parameters are fused and calculated to obtain the juice fullness index, which characterizes the relative abundance and distribution of the flowable juice inside the passion fruit. The grading of passion fruit juice fullness is determined based on the juice fullness index.

2. The method according to claim 1, characterized in that, The water speciation characteristic parameters are obtained through the following steps: The average spectrum is obtained by averaging the multiple near-infrared diffuse reflectance spectra. Based on the average spectrum, the characteristic absorption peaks corresponding to free water and cell-bound water are located, and the actual absorbance of the two characteristic absorption peaks is obtained by fitting the characteristic absorption peaks. The actual absorbance is corrected for deviation based on the peak shape correction factor of the characteristic absorption peak to eliminate the error caused by peak shape distortion; the peak shape correction factor is obtained by calibration of standard passion fruit samples and determined by iterative optimization through fitting residuals; The water speciation characteristic parameters are calculated based on the corrected actual absorbance. These parameters are used to characterize the relative proportion of free water to cell-bound water.

3. The method according to claim 2, characterized in that, The peel attenuation correction parameters are obtained through the following steps: The baseline slope of the characteristic response band of the pericarp in the average spectrum is extracted, and the baseline slope is positively correlated with the pericarp thickness; The peel thickness is calculated based on the baseline slope and a pre-established correlation between the peel thickness and the baseline slope. The correlation is calibrated by fitting standard samples with different peel thicknesses. Based on the peel thickness, an adaptive compensation mechanism is used to calculate the peel attenuation correction parameters.

4. The method according to claim 1, characterized in that, The internal structure scattering parameters are obtained through the following steps: Calculate the absorbance difference at each wavelength point across the entire near-infrared diffuse reflectance spectrum of multiple near-infrared spectra, and use the absorbance difference to characterize the degree of local scattering fluctuation in the single fruit spectrum; The absorbance difference across the entire wavelength range is integrated to calculate the internal structure scattering parameters, which are used to quantify the uniformity of the passion fruit pulp, the degree of cavity and shrinkage disturbance.

5. The method according to claim 1, characterized in that, The fruit density characteristic parameters are obtained through the following steps: Collect a three-dimensional outline image of passion fruit, extract the external size parameters of passion fruit, and calculate the actual volume of passion fruit; The weight of a single passion fruit is obtained, and the fruit density characteristic parameters are calculated based on the actual volume.

6. The method according to claim 1, characterized in that, The process of fusing and calculating the multi-dimensional feature parameters yields the sap fullness index, which characterizes the relative abundance and distribution of the flowable sap within the passion fruit, including: The water morphology characteristic parameters are corrected according to the peel attenuation correction parameters to eliminate the interference of peel thickness on the water morphology characteristic parameters. The corrected water morphology characteristic parameters are fused with the fruit density characteristic parameters to obtain a basic fusion value, which is used to characterize the basic reserve of available juice and the fullness of the pulp inside the passion fruit. The basic fusion value is corrected based on the internal structure scattering parameters to weaken the interference of passion fruit internal cavities, shrinkage, and uneven juice distribution on the juice fullness determination. The modified base fusion value is optimized by combining the pre-trained adaptive weight coefficients to obtain the juice fullness index.

7. The method according to any one of claims 1-6, characterized in that, The step of grading the passion fruit pulp fullness based on the pulp fullness index includes: Based on sample training, the grading threshold of the sap filling index and the structural anomaly threshold of the internal structure scattering parameter are determined. Based on the threshold range of the juice fullness index and whether the internal structure scattering parameters exceed the structural abnormality threshold, the passion fruit juice fullness is divided into four levels: full juice, normal juice, insufficient juice / dry water, hollow / empty shell / severely shriveled.

8. The method according to claim 7, characterized in that, Before classifying passion fruit pulp fullness into four levels—full pulp, normal pulp, insufficient pulp / dry, and hollow / empty shell / severely shriveled—based on the threshold range of the pulp fullness index and whether the internal structure scattering parameters exceed the structural abnormality threshold, the method further includes: When the fruit density characteristic parameter is lower than the set threshold, the juice filling index is adjusted downward. The set threshold is determined by calibration based on the normal density range of the corresponding passion fruit variety.

9. A non-destructive testing device for passion fruit pulp fullness based on an AI workstation, characterized in that, The device includes: The acquisition unit is used to acquire multiple near-infrared diffuse reflectance spectra of a single passion fruit; The processing unit is used to obtain the water morphology characteristic parameters and internal structure scattering parameters of passion fruit based on the near-infrared diffuse reflectance spectrum, and to construct a multi-dimensional characteristic parameter set by combining the fruit density characteristic parameters and peel attenuation interference correction parameters of passion fruit; to perform fusion calculation on the multi-dimensional characteristic parameters to obtain the juice fullness index, which characterizes the relative abundance and distribution of the flowable juice inside the passion fruit; and to complete the grading and determination of the juice fullness of passion fruit based on the juice fullness index.

10. An AI workstation, characterized in that, It includes a processor, a memory, a communication interface, and a computer program, the computer program being stored in the memory and configured to be executed by the processor, the program including instructions for performing the steps of the method as described in any one of claims 1-8.

Images