A method and system for predicting water loss and peel puffiness in tangerines based on fruit quality modeling
By acquiring hyperspectral reflectance data of mandarin orange peel and impulse response data of fruit texture, and combining them with a hidden Markov model, a fruit quality model was constructed, which solved the problem of early and accurate early warning of mandarin orange peel floating and water loss, and achieved high-precision prediction and management.
Patent Information
- Application Number
- CN202611185128.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-06
- Publication Date
- 2026-11-10
AI Technical Summary
Existing technologies are insufficient for early and accurate warnings of waterlogging in tangerines, and existing prediction methods rely on manual observation or single physicochemical indicators, which cannot achieve early and accurate warnings and large-scale coverage of waterlogging.
By acquiring hyperspectral reflectance data of mandarin orange peel and impulse response data of fruit texture, the intensity of characteristic peaks of cellulose degradation and elastic modulus of peel-flesh interface are extracted. Combined with a hidden Markov model, fruit quality modeling is constructed, the critical value of fruit quality is determined, and the potential evolution probability of floating peel and water loss is predicted, thereby generating targeted treatment strategies.
This study enabled quantitative characterization and early warning of water loss in mandarin orange peels, improving prediction accuracy, reducing economic losses, and providing a scientific basis for post-harvest management.
Smart Images

Figure CN123070548A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of smart agriculture monitoring technology, and in particular to a method and system for predicting water loss and peel dehydration in tangerines based on fruit quality modeling. Background Technology
[0002] Mandarin oranges are known for their excellent flavor and high nutritional value, but they are highly susceptible to physiological disorders such as peel dehydration during and after harvest. Peel dehydration manifests as separation of the peel from the pulp, and shrinkage and disintegration of juice vesicles, leading to reduced fruit weight and deteriorated taste, and is one of the main causes of post-harvest economic losses. Existing prediction technologies mainly rely on manual visual observation or detection of single physicochemical indicators, which have significant limitations: on the one hand, manual sampling is highly subjective and inefficient, making it difficult to cover large-scale storage environments; on the other hand, while traditional near-infrared spectroscopy and other technologies can reflect changes in internal components, they are not sensitive enough to the physical and mechanical changes in the microstructure of the peel during the early stages of peel dehydration (such as cell wall degradation and changes in interfacial elasticity), thus failing to provide early warning.
[0003] In recent years, hyperspectral imaging and mechanical testing technologies have been gradually applied to the quality testing of agricultural products, but these applications are mostly limited to the independent analysis of static indicators. Fruit dehydration is a dynamic evolutionary process involving the coupling of cell wall metabolism and tissue mechanical properties. Existing technologies lack effective models that integrate spectral features and mechanical responses from multiple sources, making it impossible to accurately characterize its continuous evolution. Furthermore, the fruit's condition is affected non-linearly by environmental temperature and humidity, and traditional threshold discrimination methods struggle to address this uncertainty, leading to prediction lags or misjudgments. Therefore, there is an urgent need for a prediction method and system that can integrate multi-source sensing data, quantify the dynamic evolution of fruit quality, and achieve early and accurate warnings to improve the level of intelligent management during and after the harvest of tangerines. Summary of the Invention
[0004] To address at least one of the aforementioned technical problems, this invention proposes a method and system for predicting water loss and peel saturation in tangerines based on fruit quality modeling.
[0005] The first aspect of this invention provides a method for predicting waterlogging and dehydration of mandarin oranges based on fruit quality modeling, comprising: Hyperspectral reflectance data of the peel and impulse response data of the fruit texture of mandarin oranges at different stages of the floating peel and drying process were obtained. The intensity of the characteristic peak of cellulose degradation in the peel cell wall was extracted based on the hyperspectral reflectance data of the peel. At the same time, the impulse response data of the fruit texture was enveloped and demodulated based on Hilbert transform to obtain the elastic modulus of the peel-flesh interface. Based on the characteristic peak intensity of cellulose degradation and the elastic modulus of the peel-pulp interface, fruit quality modeling was performed on mandarin oranges at different stages of the floating peel and water-drying process to determine the critical fruit quality value of mandarin oranges when they reach the floating peel and water-drying stage. According to the preset time period, the hyperspectral reflectance change data of the peel and the impulse response change data of the fruit texture of the mandarin orange fruit in the area to be predicted for waterlogging are obtained, and the evolution trajectory of the hyperspectral reflectance change data of the peel and the impulse response change data of the fruit texture is determined based on the hidden Markov model. Based on the evolutionary trajectory and the critical fruit quality value of mandarin oranges reaching the point of floating peel and water loss, the potential evolutionary probability of floating peel and water loss of mandarin oranges in the prediction area is determined. The treatment strategy for tangerines in the predicted area of floating peel and waterlogging is determined based on the potential evolution probability of floating peel and waterlogging.
[0006] In this scheme, the acquisition of hyperspectral reflectance data of the peel and impulse response data of the fruit texture of tangerines at different stages of the peel drying process, the extraction of the characteristic peak intensity of cellulose degradation in the peel cell wall based on the hyperspectral reflectance data, and the envelope demodulation of the impulse response data of the fruit texture based on Hilbert transform to obtain the elastic modulus of the peel-flesh interface are specifically as follows: Samples of mandarin oranges were collected at different stages of the floating peel and water-drying process. These different stages of the floating peel and water-drying process include the initial stage of non-floating peel and water-drying, in which the peel and pulp are tightly attached without any gaps; the transition stage from non-floating peel and water-drying to floating peel and water-drying, in which the gaps between peel cells initially expand and the pulp tissue begins to loosen; and the final stage of floating peel and water-drying, in which the peel and pulp completely separate to form obvious cavities and the pulp juice cells shrink and disintegrate. The epidermal hyperspectral reflectance data of each mandarin orange sample were obtained. The epidermal hyperspectral reflectance data were preprocessed by dark current correction, scattering correction and first derivative transformation in sequence. The characteristic band intervals related to the degradation of cellulose in the peel cell wall were determined in the preprocessed epidermal hyperspectral reflectance data. The spectral reflectance within the characteristic band interval is continuously projected and filtered to extract the reflection peak intensity value corresponding to each characteristic band. The reflection peak intensity values are arranged in band order and then Gaussian fitting is used to construct the cellulose degradation characteristic peak curve. The peak intensity of the pericarp cell wall cellulose degradation characteristic peak is determined based on the peak intensity of the cellulose degradation characteristic peak curve. Each mandarin orange sample is subjected to vibration tapping at a preset frequency to obtain fruit texture impulse response data. The fruit texture impulse response data is then subjected to zero-phase filtering. Hilbert transform is used to construct an analytical signal from the filtered fruit texture impulse response data to obtain the instantaneous amplitude envelope of the fruit texture impulse response data. The starting time of the fruit peel's stress response is determined based on the rising slope of the instantaneous amplitude envelope, and the ending time of the fruit pulp's stress response is determined based on the falling slope of the instantaneous amplitude envelope. The difference in envelope amplitude between the starting time of the fruit peel's stress response and the ending time of the fruit pulp's stress response is extracted. The stress-strain ratio at the fruit peel-pulp interface is calculated in conjunction with the magnitude of the pulse excitation force, and the stress-strain ratio is used as the elastic modulus of the fruit peel-pulp interface.
[0007] In this scheme, the process of modeling fruit quality for tangerines at different stages of the floating peel and water-drying process based on the intensity of the characteristic peak of cellulose degradation and the elastic modulus of the peel-pulp interface, and determining the critical fruit quality value for tangerines reaching the stage of floating peel and water-drying, specifically involves: Based on the characteristic peak intensity of cellulose degradation, the elastic modulus of the peel-pulp interface and the corresponding floating peel and water-drying process stages of each mandarin orange sample, a fruit quality evolution sample set was constructed. The fruit quality evolution sample set was standardized and a fruit quality state sequence was constructed according to the order of the floating peel and water-drying process stages. The Gaussian process regression algorithm was used to fit the fruit quality state sequence. The intensity of the characteristic peak of cellulose degradation and the elastic modulus of the peel-flesh interface were used as input variables, and the fruit quality state was used as the output variable to establish a continuous evolution model of fruit quality. Based on the continuous evolution model of fruit quality, the predicted mean and predicted variance of fruit quality at each stage of the floating peel and water-drying process were calculated to obtain the fruit quality evolution curve. The first derivative of the fruit mass evolution curve is calculated to obtain the fruit mass change rate curve. The second derivative of the fruit mass change rate curve is calculated to obtain the fruit mass change acceleration curve. The fruit mass evolution inflection point is determined based on the fruit mass change rate curve and the fruit mass change acceleration curve. The fruit mass evolution inflection point is mapped to the fruit mass evolution curve to obtain the position of fruit mass mutation. Centered on the location of the fruit quality mutation, a reliable interval for fruit quality is constructed by combining the prediction variance of the corresponding floating skin and water-drying process stage. Bayesian change point detection is performed on each fruit quality state within the reliable interval, and the posterior change point probability between adjacent fruit quality states is calculated. Based on the posterior change point probability, the state transition point where the fruit continuously jumps from the non-floating skin and water-drying state to the floating skin and water-drying state is determined. The predicted fruit quality value corresponding to the state inflection point is determined as the critical value for the fruit quality of tangerines when they reach the point of floating peel and dryness.
[0008] In this scheme, the acquisition of hyperspectral reflectance change data of tangerine peel and impulse response change data of fruit texture in the predicted area of waterlogging and peel floating is carried out at a preset time period. The evolution trajectory of the hyperspectral reflectance change data and the impulse response change data of fruit texture is determined based on a hidden Markov model. Specifically: According to the preset time period, the hyperspectral reflectance data of the peel of mandarin oranges and the impulse response data of the fruit texture in the predicted area of water loss and peel floating are obtained. The intensity change value of the reflection peak in the characteristic band range related to the degradation of cellulose in the peel cell wall is extracted. The attenuation rate change value of the elastic modulus of the peel-flesh interface is extracted according to the impulse response data of the fruit texture. The observation sequence of the hidden Markov model is constructed by combining the change value of the reflection peak intensity and the change value of the attenuation rate of the elastic modulus of the peel-pulp interface. After aligning the data of each time node in the observation sequence according to the dimensions, a multidimensional observation vector is constructed. The state space of the Hidden Markov Model is defined as the set of continuous evolution states of the mandarin orange fruit from the early stage of non-floating water-drying to the late stage of floating water-drying. The state transition probability matrix, observation probability matrix and initial state distribution vector are initialized. The Baum-Welch algorithm is used to iteratively train the multidimensional observation vector. The forward and backward probabilities of the observation sequence under the Hidden Markov Model are calculated by the forward-backward algorithm. The transition probability values between adjacent states in the state transition probability matrix and the conditional probability density of the observation vector in each state in the observation probability matrix are updated until the increment of the log-likelihood function converges to a stable threshold. The Viterbi algorithm is used to decode the trained Hidden Markov Model, calculate the optimal state path corresponding to the observation sequence, and establish a mapping relationship between each state node in the optimal state path and the floating peel and drying process stage of the mandarin orange fruit. Based on the mapping relationship, the observation vectors at each time point are concatenated in the order of state to obtain the continuous trajectory of the evolution of epidermal hyperspectral reflectance change data and fruit texture impulse response change data as the floating skin dries up.
[0009] In this scheme, the step of determining the potential evolutionary probability of tangerine fruit in the area to be predicted for floating peel and water loss based on the evolutionary trajectory and the critical fruit quality value of tangerines reaching the point of floating peel and water loss specifically involves: Based on the evolutionary trajectory and the continuous evolution model of fruit quality, the predicted fruit quality value corresponding to each time node is calculated, and the quality evolution sequence of mandarin orange fruits in the area to be predicted for peeling and waterlogging is constructed. The predicted fruit quality values in the quality evolution sequence are compared with the critical fruit quality value of the mandarin orange when it reaches the point of floating peel and dryness. The quality deviation of each state node is calculated. A quality convergence sequence is constructed based on the direction and magnitude of the change in the quality deviation of adjacent state nodes. The evolutionary segment in which the quality evolution sequence continuously converges to the critical fruit quality value is determined. Based on the hidden Markov model state transition probabilities corresponding to each time node in the evolutionary segment, the cumulative transition probability of each state node in the evolutionary segment continuing to transition to the subsequent floating and low water state is calculated. Combined with the mass deviation corresponding to each state node in the mass approach sequence, the cumulative transition probability is corrected to obtain the critical approach probability corresponding to each state node. According to the time sequence of the evolution trajectory, the critical approach probabilities corresponding to each state node are recursively fused to calculate the cumulative arrival probability of the quality evolution sequence finally reaching or exceeding the critical value of the fruit quality. The cumulative arrival probability is determined as the potential evolution probability of the tangerine fruit in the area to be predicted for floating peel and water loss.
[0010] In this scheme, the step of determining the treatment strategy for tangerines in the predicted area of floating peel and waterlogging based on the potential evolution probability of waterlogging and waterlogging is specifically as follows: When the potential evolution probability of floating skin and water loss is greater than the first probability threshold, it is determined that there is a high risk of floating skin and water loss in the predicted area. The tangerine fruits in the corresponding area are marked as priority harvesting objects. At the same time, the node positions in the evolution trajectory of the fruits in the risk area with mass deviation greater than the deviation threshold are retrieved. The epidermal hyperspectral reflectance change data and fruit texture impulse response change data corresponding to the node are associated to generate a harvest warning list containing the abnormal interval of hyperspectral characteristic peaks and the elastic modulus attenuation gradient. When the potential evolution probability of floating skin and water-drying is between the first probability threshold and the second probability threshold, it is determined that there is a risk of floating skin and water-drying in the area to be predicted. The tangerine fruits in the medium-risk area are marked as dynamic monitoring objects. The collection cycle of hyperspectral reflectance data and impulse response data is shortened according to the time node interval of the evolution trajectory. The optimal state path is updated through the Viterbi algorithm based on the real-time observation vector obtained by the collection, and the potential evolution probability of floating skin and water-drying is recalculated. When the updated potential evolution probability of floating skin and water shortage shows a monotonically increasing trend for three consecutive times, the dynamic monitoring object will be upgraded to a priority harvesting object. When the potential evolution probability of floating skin and water loss is less than the second probability threshold, it is determined that there is a low risk of floating skin and water loss in the area to be predicted. The tangerines in the low-risk area are marked as regular storage objects. The data collection frequency is maintained according to the original preset time period. The real-time fruit quality prediction value is compared with the fruit quality critical value. When the difference obtained by comparison is less than the safety margin threshold, the regular storage object is adjusted to a dynamic monitoring object. Based on the spatial distribution density of priority harvesting objects, dynamic monitoring objects, and regular storage objects, differentiated treatment plans including harvesting priorities and monitoring frequencies are generated, and the treatment strategy for mandarin oranges in the predicted area of floating peel and water shortage is determined.
[0011] A second aspect of the present invention also provides a tangerine peel dehydration prediction system based on fruit quality modeling. The system includes a memory and a processor. The memory includes a tangerine peel dehydration prediction method program based on fruit quality modeling. When the tangerine peel dehydration prediction method program based on fruit quality modeling is executed by the processor, it implements the steps of the tangerine peel dehydration prediction method based on fruit quality modeling as described in any of the above claims.
[0012] This invention discloses a method and system for predicting waterlogging and stunted growth of mandarin oranges based on fruit quality modeling. The method first collects hyperspectral reflectance data of the peel and impulse response data of the fruit texture of mandarin oranges at different stages of waterlogging and stunted growth. The intensity of characteristic peaks related to cellulose degradation in the peel is extracted, and the elastic modulus of the peel-flesh interface is obtained through Hilbert transform demodulation. Based on these parameters, a fruit quality model is constructed to determine the critical value for fruit quality during waterlogging and stunted growth. Hyperspectral and impulse response change data of mandarin oranges in the test area are collected at preset intervals. A hidden Markov model is used to characterize the data evolution trajectory, and the potential evolution probability of waterlogging and stunted growth is calculated based on the critical value, ultimately generating targeted treatment strategies. This invention integrates spectral and mechanical characteristics, achieving quantitative characterization and early warning of the waterlogging and stunted growth process with high prediction accuracy, providing a scientific basis for mandarin orange fruit quality control and post-harvest storage decisions. Attached Figure Description
[0013] Figure 1 A flowchart of a method for predicting water loss and peel saturation in tangerines based on fruit quality modeling, according to the present invention, is shown. Figure 2 The flowchart of the present invention is shown to determine the critical value of fruit quality of mandarin oranges that have reached the point of floating peel and dryness. Figure 3 The flowchart illustrating the present invention for determining the potential evolution probability of waterlogging and dehydration in mandarin orange fruits is shown. Figure 4 A block diagram of a tangerine peel dehydration prediction system based on fruit quality modeling is shown in this invention. Detailed Implementation
[0014] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.
[0015] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.
[0016] Figure 1The flowchart of a method for predicting water loss and peel stagnation in tangerines based on fruit quality modeling according to the present invention is shown.
[0017] like Figure 1 As shown, the first aspect of the present invention provides a method for predicting water loss and peel stagnation in tangerines based on fruit quality modeling, comprising: Hyperspectral reflectance data of the peel and impulse response data of the fruit texture of mandarin oranges at different stages of the floating peel and drying process were obtained. The intensity of the characteristic peak of cellulose degradation in the peel cell wall was extracted based on the hyperspectral reflectance data of the peel. At the same time, the impulse response data of the fruit texture was enveloped and demodulated based on Hilbert transform to obtain the elastic modulus of the peel-flesh interface. Based on the characteristic peak intensity of cellulose degradation and the elastic modulus of the peel-pulp interface, fruit quality modeling was performed on mandarin oranges at different stages of the floating peel and water-drying process to determine the critical fruit quality value of mandarin oranges when they reach the floating peel and water-drying stage. According to the preset time period, the hyperspectral reflectance change data of the peel and the impulse response change data of the fruit texture of the mandarin orange fruit in the area to be predicted for waterlogging are obtained, and the evolution trajectory of the hyperspectral reflectance change data of the peel and the impulse response change data of the fruit texture is determined based on the hidden Markov model. Based on the evolutionary trajectory and the critical fruit quality value of mandarin oranges reaching the point of floating peel and water loss, the potential evolutionary probability of floating peel and water loss of mandarin oranges in the prediction area is determined. The treatment strategy for tangerines in the predicted area of floating peel and waterlogging is determined based on the potential evolution probability of floating peel and waterlogging.
[0018] It should be noted that by acquiring hyperspectral reflectance data of the peel and impulse response data of fruit texture of tangerines at different stages of the floating peel and water-drying process, the intensity of the characteristic peak of cellulose degradation in the peel cell wall and the elastic modulus of the peel-pulp interface were extracted respectively. This enabled the synchronous quantitative characterization of changes in the chemical composition of the peel and the mechanical properties of the peel-pulp interface, providing a multidimensional quality evaluation basis for the formation process of floating peel and water-drying. Furthermore, a fruit quality model was established based on the intensity of the characteristic peak of cellulose degradation and the elastic modulus of the peel-pulp interface, and the critical fruit quality value of tangerines reaching the floating peel and water-drying stage was determined. This transformed the boundary of floating peel and water-drying from empirical judgment to a vectorized quality index, improving the objectivity and accuracy of the critical state determination. Subsequently, according to a preset time... This study periodically collects hyperspectral reflectance and texture impulse response data of mandarin oranges in the area to be predicted. Combined with a hidden Markov model, it determines the dynamic evolution trajectory of fruit quality, enabling continuous tracking and state identification of the development process of mandarin orange peel dehydration, avoiding misjudgments caused by single detections. Based on this, the potential evolution probability of mandarin orange peel dehydration is calculated according to the evolution trajectory and critical fruit quality values, allowing for early prediction of the trend and likelihood of dehydration, thus improving the timeliness of risk warnings. Finally, corresponding mandarin orange treatment strategies are formulated based on the potential evolution probability of dehydration, enabling differentiated harvesting and dynamic monitoring of mandarin oranges in different risk areas, improving the scientific nature of harvesting decisions, and reducing economic losses caused by dehydration.
[0019] According to an embodiment of the present invention, the acquisition of hyperspectral reflectance data of the peel and impulse response data of the fruit texture of tangerines at different stages of the peel-drying process, the extraction of the characteristic peak intensity of cellulose degradation in the peel cell wall based on the hyperspectral reflectance data, and the envelope demodulation of the impulse response data of the fruit texture based on Hilbert transform to obtain the elastic modulus of the peel-flesh interface are specifically as follows: Samples of mandarin oranges were collected at different stages of the floating peel and water-drying process. These different stages of the floating peel and water-drying process include the initial stage of non-floating peel and water-drying, in which the peel and pulp are tightly attached without any gaps; the transition stage from non-floating peel and water-drying to floating peel and water-drying, in which the gaps between peel cells initially expand and the pulp tissue begins to loosen; and the final stage of floating peel and water-drying, in which the peel and pulp completely separate to form obvious cavities and the pulp juice cells shrink and disintegrate. The epidermal hyperspectral reflectance data of each mandarin orange sample were obtained. The epidermal hyperspectral reflectance data were preprocessed by dark current correction, scattering correction and first derivative transformation in sequence. The characteristic band intervals related to the degradation of cellulose in the peel cell wall were determined in the preprocessed epidermal hyperspectral reflectance data. The spectral reflectance within the characteristic band interval is continuously projected and filtered to extract the reflection peak intensity value corresponding to each characteristic band. The reflection peak intensity values are arranged in band order and then Gaussian fitting is used to construct the cellulose degradation characteristic peak curve. The peak intensity of the pericarp cell wall cellulose degradation characteristic peak is determined based on the peak intensity of the cellulose degradation characteristic peak curve. Each mandarin orange sample is subjected to vibration tapping at a preset frequency to obtain fruit texture impulse response data. The fruit texture impulse response data is then subjected to zero-phase filtering. Hilbert transform is used to construct an analytical signal from the filtered fruit texture impulse response data to obtain the instantaneous amplitude envelope of the fruit texture impulse response data. The starting time of the fruit peel's stress response is determined based on the rising slope of the instantaneous amplitude envelope, and the ending time of the fruit pulp's stress response is determined based on the falling slope of the instantaneous amplitude envelope. The difference in envelope amplitude between the starting time of the fruit peel's stress response and the ending time of the fruit pulp's stress response is extracted. The stress-strain ratio at the fruit peel-pulp interface is calculated in conjunction with the magnitude of the pulse excitation force, and the stress-strain ratio is used as the elastic modulus of the fruit peel-pulp interface.
[0020] It should be noted that the characteristic peak intensity of cellulose degradation in the peel cell wall refers to the magnitude of the characteristic band reflection peak value extracted from epidermal hyperspectral reflectance data, corresponding to the degree of cellulose degradation in the peel cell wall. It reflects the integrity of structural materials such as cellulose in the peel cell wall and the level of cell wall degradation. When the peel cell wall degrades, the cell wall structure gradually loosens, the mechanical support capacity of the peel decreases, and the spectral reflectance characteristics of the corresponding characteristic band change. Therefore, the characteristic peak intensity of cellulose degradation in the peel cell wall can characterize the aging degree of the mandarin orange peel tissue and the process of peel structural degradation during the occurrence of peel dehydration. The elastic modulus of the peel-pulp interface is an equivalent mechanical parameter of the peel-pulp interface region calculated based on fruit texture impulse response data. It represents the peel-pulp interface's ability to resist deformation under stress and reflects the tightness of the bond between the peel and pulp and the stability of the interface structure. As waterlogging and dehydration gradually occur, the connecting tissue between the peel and pulp continuously degenerates, the interfacial bonding force decreases, and the interfacial elastic modulus decreases accordingly. Therefore, the elastic modulus of the peel-pulp interface can characterize the mechanical integrity of the peel-pulp interface and the degree of separation between the two. Combining the characteristic peak intensity of cellulose degradation in the peel cell wall with the elastic modulus of the peel-pulp interface can comprehensively reflect the evolution of mandarin orange fruit quality from two dimensions: changes in the chemical structure of the peel tissue and changes in the mechanical properties of the peel-pulp interface. This provides a more comprehensive and accurate characterization basis for modeling and risk prediction of the waterlogging and dehydration process. The mandarin orange samples were mandarin orange samples obtained at different stages of the waterlogging and dehydration process according to a preset cycle.
[0021] Figure 2The flowchart of the present invention is shown to determine the critical value of fruit quality for tangerines that have reached the stage of floating peel and dryness.
[0022] According to an embodiment of the present invention, the step of modeling the fruit quality of mandarin oranges at different stages of the floating peel and water-drying process based on the intensity of the characteristic peak of cellulose degradation and the elastic modulus of the peel-pulp interface, and determining the critical fruit quality value of mandarin oranges reaching the stage of floating peel and water-drying, specifically involves: Based on the characteristic peak intensity of cellulose degradation, the elastic modulus of the peel-pulp interface and the corresponding floating peel and water-drying process stages of each mandarin orange sample, a fruit quality evolution sample set was constructed. The fruit quality evolution sample set was standardized and a fruit quality state sequence was constructed according to the order of the floating peel and water-drying process stages. The Gaussian process regression algorithm was used to fit the fruit quality state sequence. The intensity of the characteristic peak of cellulose degradation and the elastic modulus of the peel-flesh interface were used as input variables, and the fruit quality state was used as the output variable to establish a continuous evolution model of fruit quality. Based on the continuous evolution model of fruit quality, the predicted mean and predicted variance of fruit quality at each stage of the floating peel and water-drying process were calculated to obtain the fruit quality evolution curve. The first derivative of the fruit mass evolution curve is calculated to obtain the fruit mass change rate curve. The second derivative of the fruit mass change rate curve is calculated to obtain the fruit mass change acceleration curve. The fruit mass evolution inflection point is determined based on the fruit mass change rate curve and the fruit mass change acceleration curve. The fruit mass evolution inflection point is mapped to the fruit mass evolution curve to obtain the position of fruit mass mutation. Centered on the location of the fruit quality mutation, a reliable interval for fruit quality is constructed by combining the prediction variance of the corresponding floating skin and water-drying process stage. Bayesian change point detection is performed on each fruit quality state within the reliable interval, and the posterior change point probability between adjacent fruit quality states is calculated. Based on the posterior change point probability, the state transition point where the fruit continuously jumps from the non-floating skin and water-drying state to the floating skin and water-drying state is determined. The predicted fruit quality value corresponding to the state inflection point is determined as the critical value for the fruit quality of tangerines when they reach the point of floating peel and dryness.
[0023] It should be noted that the occurrence of watery peel and shriveled state in tangerines is a continuous evolutionary process involving the gradual aging of the peel tissue, the continuous loosening of the interface between the peel and the pulp, and eventual clear separation between the two. In actual production, there is no clearly visible abrupt boundary; rather, it is a dynamic process of gradual transition from a normal state to a watery peel and shriveled state. If judgment is made solely based on human experience or a single testing indicator, it is easy to fail to accurately identify the critical point at which tangerines transition from a non-watery peel and shriveled state due to individual fruit differences, varietal differences, and different developmental stages. This could lead to premature harvesting, resulting in decreased fruit quality, or delayed warnings, allowing watery peel and shriveled state to have already formed, causing economic losses. Therefore, it is necessary to determine a critical fruit quality value that can objectively characterize the onset of watery peel and shriveled state in tangerines. A continuous evolution model of fruit quality was established by utilizing the characteristic peak intensity of cellulose degradation and the elastic modulus of the peel-pulp interface. This model allows fruit quality to be continuously expressed as the peel becomes waterlogged. By combining the rate of change, acceleration of change, and Bayesian variable point detection, the model identifies the turning point where fruit quality changes from a steady state to a continuous transition. This accurately determines the critical position where the fruit quality undergoes a fundamental change, and the corresponding predicted fruit quality value is determined as the critical value for the fruit quality of mandarin oranges when they reach the point of peel becoming waterlogged. This realizes the transformation from empirical judgment to data-driven quantitative determination of the threshold for peel becoming waterlogged, improving the objectivity of the predicted threshold for peel becoming waterlogged.
[0024] According to an embodiment of the present invention, the step of acquiring hyperspectral reflectance change data of tangerine peel and impulse response change data of fruit texture in the predicted area of waterlogging according to a preset time period, and determining the evolution trajectory of the hyperspectral reflectance change data of peel and impulse response change data of fruit texture based on a hidden Markov model, specifically includes: According to the preset time period, the hyperspectral reflectance data of the peel of mandarin oranges and the impulse response data of the fruit texture in the predicted area of water loss and peel floating are obtained. The intensity change value of the reflection peak in the characteristic band range related to the degradation of cellulose in the peel cell wall is extracted. The attenuation rate change value of the elastic modulus of the peel-flesh interface is extracted according to the impulse response data of the fruit texture. The observation sequence of the hidden Markov model is constructed by combining the change value of the reflection peak intensity and the change value of the attenuation rate of the elastic modulus of the peel-pulp interface. After aligning the data of each time node in the observation sequence according to the dimensions, a multidimensional observation vector is constructed. The state space of the Hidden Markov Model is defined as the set of continuous evolution states of the mandarin orange fruit from the early stage of non-floating water-drying to the late stage of floating water-drying. The state transition probability matrix, observation probability matrix and initial state distribution vector are initialized. The Baum-Welch algorithm is used to iteratively train the multidimensional observation vector. The forward and backward probabilities of the observation sequence under the Hidden Markov Model are calculated by the forward-backward algorithm. The transition probability values between adjacent states in the state transition probability matrix and the conditional probability density of the observation vector in each state in the observation probability matrix are updated until the increment of the log-likelihood function converges to a stable threshold. The Viterbi algorithm is used to decode the trained Hidden Markov Model, calculate the optimal state path corresponding to the observation sequence, and establish a mapping relationship between each state node in the optimal state path and the floating peel and drying process stage of the mandarin orange fruit. Based on the mapping relationship, the observation vectors at each time point are concatenated in the order of state to obtain the continuous trajectory of the evolution of epidermal hyperspectral reflectance change data and fruit texture impulse response change data as the floating skin dries up.
[0025] It should be noted that by acquiring hyperspectral reflectance data of the peel and impulse response data of the fruit texture of tangerines in the area to be predicted according to a preset time period, and extracting the reflection peak intensity change value reflecting the degree of cellulose degradation in the peel cell wall and the interfacial elastic modulus decay rate change value reflecting the change in the bonding state between the peel and the pulp, the synchronous perception of the tissue chemical changes of the peel and the internal mechanical changes of the fruit during the floating peel and water-drying process of tangerines is achieved, avoiding the problem that a single indicator cannot fully reflect the evolution state of floating peel and water-drying. By constructing a multidimensional observation vector containing the reflection peak intensity change value and the elastic modulus decay rate change value, and using a hidden Markov model to analyze the implicit state of tangerines from the early stage of non-floating peel and water-drying to the end of floating peel and water-drying, the underlying state of the fruit is analyzed. The model is designed to model the state of the fruit, enabling the development of internal peel dehydration that cannot be directly observed to be probabilistically represented by externally detectable parameters, thus achieving a dynamic description of the continuous changes in the fruit's state. The Baum-Welch algorithm is used to optimize the state transition probability and observation probability, improving the model's adaptability to data characteristics of different mandarin orange individuals and different evolutionary stages. Then, the Viterbi algorithm is used to determine the optimal state evolution path corresponding to the observation sequence, achieving accurate localization of the stage of peel dehydration in mandarin oranges. Finally, based on the mapping relationship between state nodes, a continuous evolution trajectory of changes in epidermal hyperspectral reflectance and fruit texture impulse response is constructed, transforming the fruit quality change process from discrete detection results into a continuous dynamic evolution process.
[0026] Figure 3 The flowchart illustrating the present invention for determining the potential evolutionary probability of waterlogging and dehydration in mandarin orange fruits is shown.
[0027] According to an embodiment of the present invention, the step of determining the potential evolutionary probability of tangerine fruit in the area to be predicted for floating peel and water loss based on the evolutionary trajectory and the critical fruit quality value of tangerines reaching the point of floating peel and water loss specifically involves: Based on the evolutionary trajectory and the continuous evolution model of fruit quality, the predicted fruit quality value corresponding to each time node is calculated, and the quality evolution sequence of mandarin orange fruits in the area to be predicted for peeling and waterlogging is constructed. The predicted fruit quality values in the quality evolution sequence are compared with the critical fruit quality value of the mandarin orange when it reaches the point of floating peel and dryness. The quality deviation of each state node is calculated. A quality convergence sequence is constructed based on the direction and magnitude of the change in the quality deviation of adjacent state nodes. The evolutionary segment in which the quality evolution sequence continuously converges to the critical fruit quality value is determined. Based on the hidden Markov model state transition probabilities corresponding to each time node in the evolutionary segment, the cumulative transition probability of each state node in the evolutionary segment continuing to transition to the subsequent floating and low water state is calculated. Combined with the mass deviation corresponding to each state node in the mass approach sequence, the cumulative transition probability is corrected to obtain the critical approach probability corresponding to each state node. According to the time sequence of the evolution trajectory, the critical approach probabilities corresponding to each state node are recursively fused to calculate the cumulative arrival probability of the quality evolution sequence finally reaching or exceeding the critical value of the fruit quality. The cumulative arrival probability is determined as the potential evolution probability of the tangerine fruit in the area to be predicted for floating peel and water loss.
[0028] It should be noted that the formation of the floating peel and water-drying state of tangerines is not caused by a sudden change in state at a single point in time, but is a gradual evolutionary process formed by the gradual degradation of the cell wall of the peel, the gradual decrease in the elasticity of the peel-pulp interface, and the continuous changes in the internal tissue structure of the fruit. Therefore, it is difficult to determine whether the fruit will enter the floating peel and water-drying state in the future based solely on the current test results. By utilizing the evolutionary trajectory of epidermal hyperspectral reflectance changes and fruit texture impulse response changes obtained from a Hidden Markov Model (HMM), the state transition trend of mandarin oranges at different time points can be reflected. Combined with the critical value of waterlogged peel determined by a continuous fruit quality evolution model, the current fruit quality state is correlated with the boundary of waterlogged peel occurrence. By calculating the degree of quality deviation between each state node and the critical value, it is determined whether the fruit quality is in a continuous evolutionary process converging towards the critical state of waterlogged peel. Simultaneously, the state transition probability in the HMM describes the possible path of the fruit's development from the current state to the subsequent waterlogged peel state, and the state transition probability is corrected by incorporating the quality convergence trend. This ensures that the calculation results simultaneously consider the current quality level and future evolutionary direction, thereby obtaining a more accurate probability of waterlogged peel evolution that better reflects the actual physiological changes. The potential evolution probability of waterlogged peel represents the likelihood that mandarin oranges in the predicted area will reach or exceed the critical value of waterlogged peel after continuous physiological evolution from their current state, reflecting the risk level and future trend of mandarin oranges developing from a normal state to a waterlogged peel state.
[0029] According to an embodiment of the present invention, the step of determining the treatment strategy for tangerines in the predicted area of floating peel and waterlogging based on the potential evolution probability of waterlogging and waterlogging specifically includes: When the potential evolution probability of floating skin and water loss is greater than the first probability threshold, it is determined that there is a high risk of floating skin and water loss in the predicted area. The tangerine fruits in the corresponding area are marked as priority harvesting objects. At the same time, the node positions in the evolution trajectory of the fruits in the risk area with mass deviation greater than the deviation threshold are retrieved. The epidermal hyperspectral reflectance change data and fruit texture impulse response change data corresponding to the node are associated to generate a harvest warning list containing the abnormal interval of hyperspectral characteristic peaks and the elastic modulus attenuation gradient. When the potential evolution probability of floating skin and water-drying is between the first probability threshold and the second probability threshold, it is determined that there is a risk of floating skin and water-drying in the area to be predicted. The tangerine fruits in the medium-risk area are marked as dynamic monitoring objects. The collection cycle of hyperspectral reflectance data and impulse response data is shortened according to the time node interval of the evolution trajectory. The optimal state path is updated through the Viterbi algorithm based on the real-time observation vector obtained by the collection, and the potential evolution probability of floating skin and water-drying is recalculated. When the updated potential evolution probability of floating skin and water shortage shows a monotonically increasing trend for three consecutive times, the dynamic monitoring object will be upgraded to a priority harvesting object. When the potential evolution probability of floating skin and water loss is less than the second probability threshold, it is determined that there is a low risk of floating skin and water loss in the area to be predicted. The tangerines in the low-risk area are marked as regular storage objects. The data collection frequency is maintained according to the original preset time period. The real-time fruit quality prediction value is compared with the fruit quality critical value. When the difference obtained by comparison is less than the safety margin threshold, the regular storage object is adjusted to a dynamic monitoring object. Based on the spatial distribution density of priority harvesting objects, dynamic monitoring objects, and regular storage objects, differentiated treatment plans including harvesting priorities and monitoring frequencies are generated, and the treatment strategy for mandarin oranges in the predicted area of floating peel and water shortage is determined.
[0030] It should be noted that when the potential evolution probability of floating peel and waterlogging is greater than the first probability threshold, it indicates that the current quality evolution state of the mandarin orange fruit is already highly close to the critical state of floating peel and waterlogging, and based on the evolution trajectory, the possibility of it continuing to transition to the floating peel and waterlogging state is high. Therefore, the mandarin oranges in the corresponding area are marked as priority harvesting targets, and further abnormal evolution nodes with large quality deviations are extracted and associated with the corresponding abnormal intervals of hyperspectral characteristic peaks and elastic modulus attenuation gradients. This allows harvesters to quickly locate fruit areas where structural degradation has occurred, enabling early intervention for high-risk fruits and preventing quality decline after the full formation of floating peel and waterlogging. When the potential evolution probability of floating peel and waterlogging is between the first and second probability thresholds, it indicates that the mandarin orange fruit has not yet reached a high-risk state, but its quality evolution trend already shows the possibility of developing towards the critical state of floating peel and waterlogging. Therefore, it is marked as a dynamic monitoring target. By shortening the data collection cycle and using real-time observation data to update the state path of the hidden Markov model, continuous tracking of the evolution trend of medium-risk fruits is achieved, improving the ability to capture risk changes. When the updated potential evolution probability of floating peel and waterlogging is continuously... When the three occurrences show a monotonically increasing trend, it indicates that the fruit quality is steadily converging towards the critical value of waterlogging and dehydration, showing a strong trend of continuous deterioration. Therefore, it is upgraded to a priority harvesting target to avoid misjudgment due to short-term fluctuations. When the potential evolution probability of waterlogging and dehydration is less than the second probability threshold, it indicates that the current quality status of the mandarin oranges still has a large safety margin before reaching the critical value, and the possibility of waterlogging and dehydration occurring in the near future is low. Therefore, it is marked as a regular storage target, maintaining the original monitoring cycle to reduce management costs. At the same time, the safety margin is judged by the difference between the real-time quality prediction value and the critical value. When the difference is less than the safety margin threshold, it indicates that the fruit quality status is gradually approaching the risk boundary, and the monitoring level needs to be increased. Therefore, it is adjusted to a dynamic monitoring target to achieve timely conversion from low-risk to potential-risk status. Finally, based on the spatial distribution density of different risk targets, a differentiated treatment plan including harvesting priority and monitoring frequency is generated, transforming the mandarin orange management process from a uniform treatment method to precise control based on risk level, improving the efficiency of handling after waterlogging and dehydration warnings, reducing post-harvest losses, and enhancing the quality assurance capability of mandarin oranges.
[0031] According to an embodiment of the present invention, it further includes: The fruit texture impulse response data is pre-gain amplified, and the input gain of the sampling circuit is dynamically adjusted according to the envelope amplitude of the impulse response signal after pre-gain amplification until the peak value of the impulse response signal falls within the preset effective sampling range. Zero-phase filtering is performed on the pulse response data collected by the adjusted sampling circuit. The filtered data is then analyzed and a signal is constructed based on the Hilbert transform to obtain the instantaneous amplitude envelope. The starting time of the fruit peel force response is identified based on the waveform morphology characteristics of the instantaneous amplitude envelope. When the amplitude attenuation rate of the instantaneous amplitude envelope within a preset time window exceeds a preset dynamic threshold, it is determined that there is a fruit peel-flesh cavity in the current fruit. The slope of the falling edge of the instantaneous amplitude envelope is compensated and corrected according to the attenuation rate. The ending time of the fruit pulp force response is re-determined based on the corrected falling edge slope. The difference in envelope amplitude between the start time of the peel stress response and the end time of the corrected pulp stress response is extracted, and the stress-strain ratio is calculated in combination with the magnitude of the pulse excitation force. The stress-strain ratio is used as the optimized peel-pulp interface elastic modulus to solve the problem of elastic modulus calculation failure caused by excessive signal attenuation.
[0032] According to an embodiment of the present invention, the slope compensation correction of the falling edge of the instantaneous amplitude envelope based on the attenuation rate specifically includes: Based on the position of the first trough of the instantaneous amplitude envelope after the initial moment of the fruit peel's stress response, the starting point of the cavity resonance attenuation segment is determined. Extract the linear attenuation interval from the starting point of the cavity resonance attenuation segment to the point before the instantaneous amplitude envelope decreases to its minimum value. Based on the slope of the linear attenuation interval and the peak attenuation rate of the impulse response signal, construct an exponential attenuation compensation function. The exponential decay compensation function is superimposed on the linear decay interval to reconstruct the falling edge trajectory of the instantaneous amplitude envelope, so that the reconstructed falling edge trajectory can extend to the background noise baseline position of the effective sampling range, thereby determining the time corresponding to the background noise baseline position as the corrected fruit pulp force response termination time.
[0033] It should be noted that in the detection scenario of the late stage of tangerine peel dehydration, the obvious cavity formed between the peel and pulp causes severe energy dissipation during the propagation of the pulse excitation force. This results in extremely weak amplitude of the collected fruit texture pulse response signal, even below the minimum trigger threshold of the sampling circuit. Traditional Hilbert transform demodulation methods fail to accurately capture the termination time of the pulp force response, leading to elastic modulus calculation failures and severely affecting the accuracy of determining the degree of peel dehydration. By pre-gain amplifying the pulse response data and dynamically adjusting the input gain of the sampling circuit, the signal peak is ensured to fall within the effective sampling range, providing a reliable data foundation for subsequent waveform analysis. The starting time of the peel force response is identified based on the waveform morphology characteristics of the instantaneous amplitude envelope, and the existence of the cavity is dynamically determined by the amplitude attenuation rate within a preset time window. The slope of the falling edge is specifically compensated and corrected. Specifically, an exponential attenuation compensation function is constructed using the linear attenuation interval characteristics from the starting point of the cavity resonance attenuation segment to the minimum value of the envelope, reconstructing the falling edge trajectory to extend it to the background noise baseline position, thereby accurately determining the termination time of the corrected pulp force response. It effectively overcomes the problem of excessive signal attenuation caused by cavities, significantly improves the accuracy of calculating the elastic modulus of the peel-pulp interface, ensures the reliability of fruit quality assessment at the end of the floating peel and water-drying stage, and provides high-quality elastic modulus data support for subsequent evolutionary trajectory analysis based on hidden Markov models and prediction of potential evolutionary probability of floating peel and water-drying, avoiding misjudgments caused by signal loss and ensuring the scientific nature of harvesting and disposal strategies.
[0034] Figure 4 A block diagram of a tangerine peel dehydration prediction system based on fruit quality modeling is shown in this invention.
[0035] A second aspect of the present invention also provides a tangerine peel dehydration prediction system based on fruit quality modeling. The system includes: a memory 401, a processor 402, and a communication interface 403. The memory includes a tangerine peel dehydration prediction method program based on fruit quality modeling. The communication interface is used for data connection and communication between the memory and the processor. When the tangerine peel dehydration prediction method program based on fruit quality modeling is executed by the processor, it implements the steps of the tangerine peel dehydration prediction method based on fruit quality modeling as described in any of the above claims.
[0036] This invention discloses a method and system for predicting waterlogging and stunted growth of mandarin oranges based on fruit quality modeling. The method first collects hyperspectral reflectance data of the peel and impulse response data of the fruit texture of mandarin oranges at different stages of waterlogging and stunted growth. The intensity of characteristic peaks related to cellulose degradation in the peel is extracted, and the elastic modulus of the peel-flesh interface is obtained through Hilbert transform demodulation. Based on these parameters, a fruit quality model is constructed to determine the critical value for fruit quality during waterlogging and stunted growth. Hyperspectral and impulse response change data of mandarin oranges in the test area are collected at preset intervals. A hidden Markov model is used to characterize the data evolution trajectory, and the potential evolution probability of waterlogging and stunted growth is calculated based on the critical value, ultimately generating targeted treatment strategies. This invention integrates spectral and mechanical characteristics, achieving quantitative characterization and early warning of the waterlogging and stunted growth process with high prediction accuracy, providing a scientific basis for mandarin orange fruit quality control and post-harvest storage decisions.
[0037] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0038] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium 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 methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.
[0039] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
Claims
1. A method for predicting water loss and peel saturation in tangerines based on fruit quality modeling, characterized in that, Includes the following steps: Hyperspectral reflectance data of the peel and impulse response data of the fruit texture of mandarin oranges at different stages of the floating peel and drying process were obtained. The intensity of the characteristic peak of cellulose degradation in the peel cell wall was extracted based on the hyperspectral reflectance data of the peel. At the same time, the impulse response data of the fruit texture was enveloped and demodulated based on Hilbert transform to obtain the elastic modulus of the peel-flesh interface. Based on the characteristic peak intensity of cellulose degradation and the elastic modulus of the peel-pulp interface, fruit quality modeling was performed on mandarin oranges at different stages of the floating peel and water-drying process to determine the critical fruit quality value of mandarin oranges when they reach the floating peel and water-drying stage. According to the preset time period, the hyperspectral reflectance change data of the peel and the impulse response change data of the fruit texture of the mandarin orange fruit in the area to be predicted for waterlogging are obtained, and the evolution trajectory of the hyperspectral reflectance change data of the peel and the impulse response change data of the fruit texture is determined based on the hidden Markov model. Based on the evolutionary trajectory and the critical fruit quality value of mandarin oranges reaching the point of floating peel and water loss, the potential evolutionary probability of floating peel and water loss of mandarin oranges in the prediction area is determined. The treatment strategy for tangerines in the predicted area of floating peel and waterlogging is determined based on the potential evolution probability of floating peel and waterlogging.
2. The method for predicting waterlogging and dehydration of mandarin oranges based on fruit quality modeling according to claim 1, characterized in that, The process involves acquiring hyperspectral reflectance data of the peel and impulse response data of the fruit texture of tangerines at different stages of the peel-drying process. Based on the hyperspectral reflectance data, the intensity of characteristic peaks related to cellulose degradation in the peel cell wall is extracted. Simultaneously, envelope demodulation of the impulse response data of the fruit texture is performed using Hilbert transform to obtain the elastic modulus of the peel-flesh interface. Specifically: Samples of mandarin oranges were collected at different stages of the floating peel and water-drying process. These different stages of the floating peel and water-drying process include the initial stage of non-floating peel and water-drying, in which the peel and pulp are tightly attached without any gaps; the transition stage from non-floating peel and water-drying to floating peel and water-drying, in which the gaps between peel cells initially expand and the pulp tissue begins to loosen; and the final stage of floating peel and water-drying, in which the peel and pulp completely separate to form obvious cavities and the pulp juice cells shrink and disintegrate. The epidermal hyperspectral reflectance data of each mandarin orange sample were obtained. The epidermal hyperspectral reflectance data were preprocessed by dark current correction, scattering correction and first derivative transformation in sequence. The characteristic band intervals related to the degradation of cellulose in the peel cell wall were determined in the preprocessed epidermal hyperspectral reflectance data. The spectral reflectance within the characteristic band interval is continuously projected and filtered to extract the reflection peak intensity value corresponding to each characteristic band. The reflection peak intensity values are arranged in band order and then Gaussian fitting is used to construct the cellulose degradation characteristic peak curve. The peak intensity of the pericarp cell wall cellulose degradation characteristic peak is determined based on the peak intensity of the cellulose degradation characteristic peak curve. Each mandarin orange sample is subjected to vibration tapping at a preset frequency to obtain fruit texture impulse response data. The fruit texture impulse response data is then subjected to zero-phase filtering. Hilbert transform is used to construct an analytical signal from the filtered fruit texture impulse response data to obtain the instantaneous amplitude envelope of the fruit texture impulse response data. The starting time of the fruit peel's stress response is determined based on the rising slope of the instantaneous amplitude envelope, and the ending time of the fruit pulp's stress response is determined based on the falling slope of the instantaneous amplitude envelope. The difference in envelope amplitude between the starting time of the fruit peel's stress response and the ending time of the fruit pulp's stress response is extracted. The stress-strain ratio at the fruit peel-pulp interface is calculated in conjunction with the magnitude of the pulse excitation force, and the stress-strain ratio is used as the elastic modulus of the fruit peel-pulp interface.
3. The method for predicting waterlogging and dehydration of mandarin oranges based on fruit quality modeling according to claim 1, characterized in that, The method involves modeling the fruit quality of tangerines at different stages of the floating peel and water-drying process based on the characteristic peak intensity of cellulose degradation and the elastic modulus of the peel-pulp interface, and determining the critical fruit quality value for tangerines reaching the stage of floating peel and water-drying. Specifically: Based on the characteristic peak intensity of cellulose degradation, the elastic modulus of the peel-pulp interface and the corresponding floating peel and water-drying process stages of each mandarin orange sample, a fruit quality evolution sample set was constructed. The fruit quality evolution sample set was standardized and a fruit quality state sequence was constructed according to the order of the floating peel and water-drying process stages. The Gaussian process regression algorithm was used to fit the fruit quality state sequence. The intensity of the characteristic peak of cellulose degradation and the elastic modulus of the peel-flesh interface were used as input variables, and the fruit quality state was used as the output variable to establish a continuous evolution model of fruit quality. Based on the continuous evolution model of fruit quality, the predicted mean and predicted variance of fruit quality at each stage of the floating peel and water-drying process were calculated to obtain the fruit quality evolution curve. The first derivative of the fruit mass evolution curve is calculated to obtain the fruit mass change rate curve. The second derivative of the fruit mass change rate curve is calculated to obtain the fruit mass change acceleration curve. The fruit mass evolution inflection point is determined based on the fruit mass change rate curve and the fruit mass change acceleration curve. The fruit mass evolution inflection point is mapped to the fruit mass evolution curve to obtain the position of fruit mass mutation. Centered on the location of the fruit quality mutation, a reliable interval for fruit quality is constructed by combining the prediction variance of the corresponding floating skin and water-drying process stage. Bayesian change point detection is performed on each fruit quality state within the reliable interval, and the posterior change point probability between adjacent fruit quality states is calculated. Based on the posterior change point probability, the state transition point where the fruit continuously jumps from the non-floating skin and water-drying state to the floating skin and water-drying state is determined. The predicted fruit quality value corresponding to the state inflection point is determined as the critical value for the fruit quality of tangerines when they reach the point of floating peel and dryness.
4. The method for predicting waterlogging and dehydration of mandarin oranges based on fruit quality modeling according to claim 1, characterized in that, The process involves acquiring hyperspectral reflectance data of the tangerine peel and impulse response data of the fruit texture in the predicted area of waterlogging according to a preset time period, and determining the evolution trajectory of the hyperspectral reflectance data and impulse response data of the fruit texture based on a hidden Markov model. Specifically: According to the preset time period, the hyperspectral reflectance data of the peel of mandarin oranges and the impulse response data of the fruit texture in the predicted area of water loss and peel floating are obtained. The intensity change value of the reflection peak in the characteristic band range related to the degradation of cellulose in the peel cell wall is extracted. The attenuation rate change value of the elastic modulus of the peel-flesh interface is extracted according to the impulse response data of the fruit texture. The observation sequence of the hidden Markov model is constructed by combining the change value of the reflection peak intensity and the change value of the attenuation rate of the elastic modulus of the peel-pulp interface. After aligning the data of each time node in the observation sequence according to the dimensions, a multidimensional observation vector is constructed. The state space of the Hidden Markov Model is defined as the set of continuous evolution states of the mandarin orange fruit from the early stage of non-floating water-drying to the late stage of floating water-drying. The state transition probability matrix, observation probability matrix and initial state distribution vector are initialized. The Baum-Welch algorithm is used to iteratively train the multidimensional observation vector. The forward and backward probabilities of the observation sequence under the Hidden Markov Model are calculated by the forward-backward algorithm. The transition probability values between adjacent states in the state transition probability matrix and the conditional probability density of the observation vector in each state in the observation probability matrix are updated until the increment of the log-likelihood function converges to a stable threshold. The Viterbi algorithm is used to decode the trained Hidden Markov Model, calculate the optimal state path corresponding to the observation sequence, and establish a mapping relationship between each state node in the optimal state path and the floating peel and drying process stage of the mandarin orange fruit. Based on the mapping relationship, the observation vectors at each time point are concatenated in the order of state to obtain the continuous trajectory of the evolution of epidermal hyperspectral reflectance change data and fruit texture impulse response change data as the floating skin dries up.
5. The method for predicting waterlogging and dehydration of mandarin oranges based on fruit quality modeling according to claim 4, characterized in that, The process of determining the potential evolutionary probability of tangerine fruit in the predicted area for floating peel and water loss based on the evolutionary trajectory and the critical fruit quality value of tangerines reaching the point of floating peel and water loss is as follows: Based on the evolutionary trajectory and the continuous evolution model of fruit quality, the predicted fruit quality value corresponding to each time node is calculated, and the quality evolution sequence of mandarin orange fruits in the area to be predicted for peeling and waterlogging is constructed. The predicted fruit quality values in the quality evolution sequence are compared with the critical fruit quality value of the mandarin orange when it reaches the point of floating peel and dryness. The quality deviation of each state node is calculated. A quality convergence sequence is constructed based on the direction and magnitude of the change in the quality deviation of adjacent state nodes. The evolutionary segment in which the quality evolution sequence continuously converges to the critical fruit quality value is determined. Based on the hidden Markov model state transition probabilities corresponding to each time node in the evolutionary segment, the cumulative transition probability of each state node in the evolutionary segment continuing to transition to the subsequent floating and low water state is calculated. Combined with the mass deviation corresponding to each state node in the mass approach sequence, the cumulative transition probability is corrected to obtain the critical approach probability corresponding to each state node. According to the time sequence of the evolution trajectory, the critical approach probabilities corresponding to each state node are recursively fused to calculate the cumulative arrival probability of the quality evolution sequence finally reaching or exceeding the critical value of the fruit quality. The cumulative arrival probability is determined as the potential evolution probability of the tangerine fruit in the area to be predicted for floating peel and water loss.
6. The method for predicting waterlogging and dehydration of mandarin oranges based on fruit quality modeling according to claim 1, characterized in that, The specific steps for determining the treatment strategy for tangerines in the predicted area of floating peel and waterlogging based on the potential evolution probability of waterlogging are as follows: When the potential evolution probability of floating skin and water loss is greater than the first probability threshold, it is determined that there is a high risk of floating skin and water loss in the predicted area. The tangerine fruits in the corresponding area are marked as priority harvesting objects. At the same time, the node positions in the evolution trajectory of the fruits in the risk area with mass deviation greater than the deviation threshold are retrieved. The epidermal hyperspectral reflectance change data and fruit texture impulse response change data corresponding to the node are associated to generate a harvest warning list containing the abnormal interval of hyperspectral characteristic peaks and the elastic modulus attenuation gradient. When the potential evolution probability of floating skin and water-drying is between the first probability threshold and the second probability threshold, it is determined that there is a risk of floating skin and water-drying in the area to be predicted. The tangerine fruits in the medium-risk area are marked as dynamic monitoring objects. The collection cycle of hyperspectral reflectance data and impulse response data is shortened according to the time node interval of the evolution trajectory. The optimal state path is updated through the Viterbi algorithm based on the real-time observation vector obtained by the collection, and the potential evolution probability of floating skin and water-drying is recalculated. When the updated potential evolution probability of floating skin and water shortage shows a monotonically increasing trend for three consecutive times, the dynamic monitoring object will be upgraded to a priority harvesting object. When the potential evolution probability of floating skin and water loss is less than the second probability threshold, it is determined that there is a low risk of floating skin and water loss in the area to be predicted. The tangerines in the low-risk area are marked as regular storage objects. The data collection frequency is maintained according to the original preset time period. The real-time fruit quality prediction value is compared with the fruit quality critical value. When the difference obtained by comparison is less than the safety margin threshold, the regular storage object is adjusted to a dynamic monitoring object. Based on the spatial distribution density of priority harvesting objects, dynamic monitoring objects, and regular storage objects, differentiated treatment plans including harvesting priorities and monitoring frequencies are generated, and the treatment strategy for mandarin oranges in the predicted area of floating peel and water shortage is determined.
7. A fruit quality modeling-based system for predicting waterlogging and dehydration in tangerines, characterized in that, The tangerine peel dehydration prediction system based on fruit quality modeling includes a memory and a processor. The memory includes a tangerine peel dehydration prediction method program based on fruit quality modeling. When the tangerine peel dehydration prediction method program based on fruit quality modeling is executed by the processor, it implements the steps of the tangerine peel dehydration prediction method based on fruit quality modeling as described in any one of claims 1 to 6.
