A control method and system of an intelligent coring device

Abstract

The application discloses a kind of control method and system of intelligent nuclear removal equipment, it is related to food quality control relevant field, including: the image information of fruit is obtained by visual collector, fruit external feature identification is carried out to image information, obtain fruit category and hardness pre-judgment result;Based on fruit category and hardness pre-judgment result, the constraint force analysis of clamping and nuclear removal is carried out, and target balanced configuration is carried out according to clamping constraint force combined with nuclear removal constraint force, to determine collaborative control parameter;According to the control execution of collaborative control parameter, and track execution feedback data, based on execution feedback data, adaptive analysis is carried out, and execution compensation adjustment parameter is identified;Collaborative control parameter dynamic adaptive adjustment is carried out using execution compensation adjustment parameter, and nuclear removal operation control is completed.The technical problem that the existing nuclear removal equipment control cannot adapt to different fruit characteristics, resulting in poor nuclear removal effect, is solved, to achieve accurate nuclear removal for different characteristics of fruit, and then improve the technical effect of juice quality.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of food quality control, in particular to a control method and system of an intelligent coring device. BACKGROUND

[0002] In the fruit juice production process, the precision of the coring link has a crucial influence on the quality of the juice and the overall production benefit. If the coring is improper, not only will it cause the mixing of the fruit core into the juice and affect the taste, but also it may cause the reduction of the juice yield and the loss of the nutritional ingredients due to the excessive damage to the fruit pulp. At present, the main method of fruit coring is to use the traditional mechanical coring device. This kind of device carries out the coring operation on the fruit according to the preset fixed parameters, and realizes the clamping and coring actions through simple mechanical structure. Since the fruits are various in types and have great differences in hardness, and the device lacks the precise recognition and self-adaptive adjustment capability for the characteristics of different fruits, this method, when facing fruits of different types and hardness, either causes the fruits to slide in the coring process due to the insufficient clamping force, affecting the coring accuracy, or causes the excessive damage to the fruit pulp due to the excessive coring force, reducing the juice yield and the quality of the juice, and cannot meet the diversified production needs.

[0003] In the related art at present, the control of the coring device has the technical problem of poor coring effect caused by the inability to adapt to the characteristics of different fruits. SUMMARY

[0004] The present application provides a control method and system of an intelligent coring device, which obtains the fruit type and hardness prediction result by acquiring the image of the fruit through a visual collector and recognition, analyzes the clamping and coring constraint force, determines the collaborative control parameters through the balanced configuration, executes the control according to the parameters, tracks the feedback data, analyzes the execution compensation adjustment parameters, dynamically adjusts the collaborative control parameters by using the parameters, and completes the coring operation control, etc. technical means, solves the technical problem of poor coring effect caused by the inability to adapt to the characteristics of different fruits in the existing coring device control, and achieves the technical effect of precise coring of different characteristics of fruits, and further improves the quality of the juice.

[0005] The present application provides a control method of an intelligent coring device, which comprises: acquiring the image information of the fruit through a visual collector, recognizing the external characteristics of the fruit, obtaining the fruit type and hardness prediction result; analyzing the constraint force of clamping and coring based on the fruit type and hardness prediction result, determining the collaborative control parameters through the balanced configuration of the target according to the clamping constraint force combined with the coring constraint force; controlling the execution according to the collaborative control parameters, tracking the execution feedback data, identifying the execution compensation adjustment parameters based on the execution feedback data; dynamically adjusting the collaborative control parameters by using the execution compensation adjustment parameters, and completing the coring operation control.

[0006] In a possible implementation, the fruit external feature recognition is performed on the image information to obtain a fruit category and hardness prediction result, and the following processing is performed: visual feature recognition and extraction are performed on the image information to obtain visual features including color, texture, and shape contour; the visual features are input into a pre-trained fruit category classification model to output a fruit category, the fruit category classification model being a machine learning model based on a support vector machine; and a hardness engine is called in a pre-stored knowledge base according to the fruit category to perform recognition matching and hardness prediction, and a hardness prediction result corresponding to the fruit category is obtained.

[0007] In a possible implementation, the hardness engine is called in a pre-stored knowledge base according to the fruit category to perform recognition matching and hardness prediction, and a hardness prediction result corresponding to the fruit category is obtained, and the following processing is performed: a hardness prediction model and a key appearance feature index related to the fruit category are called in the pre-stored knowledge base according to the fruit category, the key appearance features including color space component values, skin glossiness, and texture roughness; feature values of the key appearance features are quantitatively calculated from the visual features based on the key appearance feature index; and the feature values of the quantized key appearance features are input into the hardness prediction model to perform hardness prediction of the corresponding fruit category, and a hardness prediction result is output, which is a data quantization result describing a fruit category hardness level.

[0008] In a possible implementation, the feature values of the key appearance features are quantitatively calculated from the visual features based on the key appearance feature index, and the following processing is performed: the key appearance feature index is analyzed to obtain description information of a region of interest corresponding to the key appearance features, and the corresponding pixel region is located in the image information based on the description information; for each key appearance feature, a data-based feature value is calculated according to pixel data of the located pixel region according to a quantitative calculation manner defined in the index; and all calculated feature values are combined in a preset order to obtain the feature values of the key appearance features, wherein the preset order meets the input feature vector requirement of the hardness prediction model.

[0009] In a possible implementation, the following processing is performed: the region of interest includes a global region, a fixed region, and a relative region, wherein the global region is the entire fruit image, the fixed region is a region defined by an image coordinate range, and the relative region is a region defined based on morphological feature points of the fruit image, and the morphological feature points include a fruit image centroid and a circumscribed rectangle center.

[0010] In a possible implementation, based on the fruit category and hardness prediction result, constraint force analysis of clamping and core removal is performed, target balance configuration is performed according to clamping constraint force and core removal constraint force, a cooperative control parameter is determined, and the following processing is performed: according to the fruit category and hardness prediction result, a clamping force constraint range and a core removal force constraint range of the current fruit are obtained through fuzzy database mapping; a clamping target function is established based on the clamping force constraint range with fruit damage as a target, and a core removal target function is established based on the core removal force constraint range with successful core removal and equipment protection as a target; balanced optimization is performed according to the clamping target function and the core removal target function, a clamping parameter and a core removal parameter control strategy combination that satisfies the clamping target and maximizes the core removal target are obtained, and the cooperative control parameter is generated.

[0011] In a possible implementation, control is performed according to the cooperative control parameter, and execution feedback data is tracked, adaptive analysis is performed based on the execution feedback data, an execution compensation adjustment parameter is identified, and the following processing is performed: the clamping and core removal execution mechanism is controlled to act based on the cooperative control parameter, and clamping force feedback data, core removal force feedback data, and fruit visual data after clamping are collected in real time; the hardness deviation between the actual hardness of the fruit and the predicted hardness is analyzed based on comparison between the core removal force feedback data and the core removal force expected value in the cooperative control parameter; the spatial position of the core in the clamped state is recalculated based on the fruit visual data after clamping, and the spatial deviation of the core position is analyzed by comparing with the pre-established core model position of the image information; the clamping force set value and the parameter adjustment amount of the core removal executor are calculated according to the hardness deviation and the compensation adjustment parameter for execution.

[0012] In a possible implementation, based on the fruit visual data after clamping, the spatial position of the core in the clamped state is recalculated, and the spatial deviation of the core position is analyzed by comparing with the pre-established core model position of the image information, and the following processing is performed: the fruit image after clamping is collected by a visual collector, wherein the visual collector includes a visible light camera and an infrared camera; the image collected by the visible light camera is used to identify the external contour of the fruit after clamping, and the clamping spatial position coordinates of the fruit in the core removal executor coordinate system are determined based on a pre-calibrated conversion relationship; the image collected by the infrared camera is used to identify the core feature, and the core spatial position coordinates of the core in the core removal executor coordinate system are determined; the clamping spatial position coordinates and the core spatial position coordinates are aligned at the center, and compared with the core model position, to obtain the spatial deviation.

[0013] In a possible implementation, the clamping force compensation correction is performed according to the hardness deviation, the clamping force set value and the parameter adjustment amount of the core removing actuator are calculated, and the following processing is performed: based on the hardness deviation, a loss tolerance coefficient of the current hardness is determined by referring to historical experience data; the pulp loss evaluation is performed on the feedback actual hardness according to the current cooperative control parameter, and a pulp loss estimation amount under the current working condition is obtained; the loss tolerance coefficient is used to calculate an adjustment target of the pulp loss estimation amount, and the clamping parameter and the core removing parameter are balanced and reconstructed according to the adjustment target, so that the clamping force set value and the parameter adjustment amount of the core removing actuator that satisfy the current hardness deviation are calculated.

[0014] The application further provides a control system of an intelligent core removing device, which comprises: a fruit external feature recognition module, which is used for acquiring image information of a fruit through a visual collector, performing fruit external feature recognition on the image information, and obtaining a fruit category and a hardness prediction result; a cooperative control parameter determination module, which is used for performing constraint force analysis of clamping and core removing based on the fruit category and the hardness prediction result, performing target balance configuration according to the clamping constraint force in combination with the core removing constraint force, and determining a cooperative control parameter; an execution feedback module, which is used for performing control execution according to the cooperative control parameter, tracking execution feedback data, performing self-adaptive analysis based on the execution feedback data, and identifying an execution compensation adjustment parameter; and a parameter dynamic adaptive adjustment module, which is used for performing dynamic adaptive adjustment of the cooperative control parameter by using the execution compensation adjustment parameter, and completing core removing operation control.

[0015] The application provides a control method and system of an intelligent core removing device, which first acquires image information of a fruit through a visual collector, performs fruit external feature recognition on the image information, and obtains a fruit category and a hardness prediction result, then performs constraint force analysis of clamping and core removing based on the fruit category and the hardness prediction result, performs target balance configuration according to the clamping constraint force in combination with the core removing constraint force, determines a cooperative control parameter, then performs control execution according to the cooperative control parameter, tracks execution feedback data, performs self-adaptive analysis based on the execution feedback data, and identifies an execution compensation adjustment parameter, and finally performs dynamic adaptive adjustment of the cooperative control parameter by using the execution compensation adjustment parameter, and completes core removing operation control. The technical effect of accurately removing cores of fruits with different characteristics and improving juice quality is achieved. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings of the embodiments of the present application will be briefly introduced below, and the flowcharts are used to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the foregoing or the following operations are not necessarily performed in sequence. On the contrary, various steps can be processed in reverse order or simultaneously according to needs. Meanwhile, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1 A flowchart of a control method of an intelligent kernel removal device provided by the embodiments of the present application.

[0018] Figure 2 A structural diagram of a control system of an intelligent kernel removal device provided by the embodiments of the present application.

[0019] Legend: fruit external feature recognition module 10, cooperative control parameter determination module 20, execution feedback module 30, parameter dynamic adaptive adjustment module 40. DETAILED DESCRIPTION

[0020] The foregoing description is only a summary of the technical solutions of the present application, in order to more clearly understand the technical means of the present application, the content of the specification can be implemented, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the following specific embodiments of the present application are described.

[0021] In order to make the purposes, technical solutions and advantages of the present application more clear, the following will combine the drawings to further describe the present application in detail, the described embodiments should not be regarded as limiting the present application, all other embodiments obtained by those skilled in the art without making creative labor, belong to the scope of protection of the present application.

[0022] In the following description, "some embodiments" are described, which describe a subset of all possible embodiments, but it can be understood that "some embodiments" can be the same subset or different subset of all possible embodiments, and can be combined with each other without conflict. The terms "include" and "have" and any variations, are intended to cover non-exclusive inclusion, for example, a process, method, system, product or server including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or modules not clearly listed or inherent to these processes, methods, products or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as understood by those skilled in the art in the technical field of the present application. The terms used herein are only for the purpose of describing the embodiments of the present application.

[0023] The embodiment of the present application provides a control method of an intelligent kernel removal device, as shown in the accompanying drawings, the method comprises the following steps: Figure 1

[0024] In step S100, image information of the fruit is acquired by a visual collector, fruit external feature recognition is performed on the image information, and a fruit category and hardness prediction result are obtained.

[0025] Specifically, the RGB image and depth information of the fruit are acquired by the visual collector, such as an industrial camera or a 3D structured light sensor, image segmentation and feature extraction are performed by using a deep learning model, and the fruit category and external features such as surface texture and color distribution are recognized. The fruit category is a large kernel fruit, including peaches, plums, lychees, apricots, olives and the like. Meanwhile, the hardness prediction is performed by combining a texture roughness or color depth regression model or by using a historical data mapping table according to the corresponding relationship between the skin color and the hardness, and the hardness prediction result is output. The hardness prediction result is a quantitative value or a classification label of the fruit hardness, such as soft, medium and hard.

[0026] In a possible implementation, the image information is subjected to fruit external feature recognition to obtain a fruit category and a hardness prediction result, and step S100 further comprises step S110 of performing visual feature recognition and extraction on the image information to obtain visual features, including color, texture and shape contour. Specifically, an image processing algorithm is used to extract three types of visual features, including color, texture and shape contour. For example, the RGB image is converted to an HSV / Lab color space, a color histogram is calculated, the H channel is divided into 8 bins, the pixel proportion of each bin is counted, and then the K-means clustering is used to determine the three colors with the highest proportion in the image as the main color tone. The contrast, correlation and energy statistics are calculated by using a gray level co-occurrence matrix, and the multi-scale texture features are extracted by using a Gabor filter bank. The Canny edge detection algorithm is used to extract the contour, the Hu invariant moment and the concave-convex degree of the contour are calculated, the fruit shape is fitted by using the minimum circumscribed rectangle, and the length-width ratio and the circularity are calculated.

[0027] ​Step S120, input the visual features into a pre-trained fruit category classification model to output a fruit category, the fruit category classification model being a machine learning model based on a support vector machine. Specifically, the visual features extracted in step S110 are input into the pre-trained support vector machine classification model, the features are mapped to a high-dimensional space through a kernel function, and an optimal classification hyperplane is found. The model training process is as follows: 1000 fruit images are collected, including 5 categories such as peaches and plums, each category is labeled with 200 images, and the training set and the test set are divided according to the ratio of 7:3. Extract color, texture, and shape feature vectors for each image. Use the LibSVM toolkit to train a multi-class support vector machine, optimize the hyperparameters through grid search, and make the test set accuracy reach a preset accuracy threshold to obtain a trained fruit category classification model. After inputting the visual features extracted in step S110 into the model, the model outputs the probability distribution of each category, and the category corresponding to the maximum probability is taken as the classification result.

[0028] Step S130, according to the fruit category, call the hardness engine in the pre-stored knowledge base to perform identification matching and hardness prediction, and obtain a hardness prediction result corresponding to the fruit category. Specifically, according to the fruit category output in step S120, the corresponding hardness prediction engine in the pre-stored knowledge base is called, such as a rule base or a regression model, and the hardness is inferred in combination with the visual features to obtain a hardness prediction result. The pre-stored knowledge base stores the hardness-feature mapping relationship of different fruit categories, for example, a regression model of the hardness of peaches and the texture contrast, color saturation; a rule base of the hardness of plums and the roundness, texture energy, etc.

[0029] In one possible implementation, according to the fruit category, the hardness engine in the pre-stored knowledge base is called to perform identification matching and hardness prediction, and a hardness prediction result corresponding to the fruit category is obtained, and step S130 further includes step S131, according to the fruit category, the hardness prediction model and the key appearance feature index related to the fruit category in the pre-stored knowledge base are called, wherein the key appearance features include color space component values, skin glossiness, and texture roughness. Specifically, the pre-stored knowledge base stores the mapping relationship of the category and the hardness prediction model and the key appearance feature index in JSON / XML format, takes the category name as the key, and takes the hardness prediction model and the key appearance feature index as the value, and realizes fast retrieval through a hash table. According to the fruit category output in step S120, the corresponding item is queried through the hash table, the hardness prediction model and the key appearance feature index are loaded, including color, glossiness, and roughness. By calling the category-specific model, the error of the general model is avoided.

[0030] Step S132, based on the key appearance feature index, quantitatively calculate the feature value of the key appearance feature from the visual feature. Specifically, according to the key appearance feature index returned in step S131, the target feature is filtered and quantified from the visual feature extracted in step S110. For example, the mean value of the saturation channel in the histogram converted from the HSV color space is extracted as the color space component value; the energy value calculated from the gray level co-occurrence matrix is taken as the skin glossiness, wherein the energy value reflects the uniformity of the texture, and the higher the glossiness, the lower the energy value; the contrast value calculated from the gray level co-occurrence matrix is taken as the texture roughness, wherein the contrast value reflects the texture definition, and the larger the contrast value, the rougher the texture.

[0031] Step S133, input the quantified feature value of the key appearance feature into the hardness prediction model to predict the hardness of the corresponding category, and output the hardness prediction result as the data quantization result of the fruit category hardness grade. Specifically, the quantified feature value is input into the hardness prediction model loaded in step S131 to perform hardness quantification calculation, and a numerical result is obtained, which is an absolute value of hardness expressed in physical units. Then, the numerical result is converted into a grade label according to a predefined threshold, such as soft, medium, hard, etc.

[0032] In one possible implementation, based on the key appearance feature index, the feature value of the key appearance feature is quantitatively calculated from the visual feature, and step S132 further includes step S1321, analyzing the key appearance feature index to obtain the description information of the region of interest corresponding to the key appearance feature, and based on the description information, locating the corresponding pixel region in the image information, wherein the region of interest includes a global region, a fixed region, and a relative region, wherein the global region is the entire fruit image, the fixed region is a region defined by an image coordinate range, and the relative region is a region defined based on the morphological feature points of the fruit image, and the morphological feature points include the fruit image centroid and the circumscribed rectangle center. Specifically, the key appearance feature index is analyzed to obtain the description information of the region of interest corresponding to each key appearance feature. The region of interest includes three types, which are the global region, the fixed region, and the relative region. The global region refers to the entire fruit image, which contains all parts of the fruit; the fixed region is a region defined by an image coordinate range, for example, a certain rectangular region from the top left corner to the bottom right corner of the image can be set as the fixed region; the relative region is a region defined based on the morphological feature points of the fruit image, and the morphological feature points include the fruit image centroid and the circumscribed rectangle center. The fruit image centroid is the mass center of the fruit image, and the circumscribed rectangle center is the center point of the circumscribed rectangle of the fruit image. For example, a region extending outward by a certain radius with the fruit image centroid as the center is taken as the relative region.

[0033] According to the above description information, the corresponding pixel region is located in the image information of the fruit. For example, if the key appearance feature corresponds to a fixed region, the region is found according to the set image coordinate range; if it corresponds to a relative region, the shape feature point is first determined, and then the corresponding region is found according to the definition rule.

[0034] In step S1322, for each key appearance feature, the data-based feature value is calculated according to the quantization calculation method defined in the index according to the pixel data of the located pixel region. Specifically, for each key appearance feature, the pixel data of the located pixel region is used to calculate the data-based feature value according to the quantization calculation method defined in the index. For example, if the key appearance feature is the color space component value, and the index defines that the average value of the saturation channel in the HSV color space is calculated, then the HSV color space information is extracted from the located pixel region, and the average value of the saturation channel is calculated as the feature value of the key appearance feature. If the key appearance feature is skin gloss, and the index defines that the energy value calculated by the gray level co-occurrence matrix is used to represent the skin gloss, then the gray level information is extracted from the pixel region, the gray level co-occurrence matrix is constructed, and the energy value is calculated as the feature value.

[0035] In step S1323, all the calculated feature values are combined in a preset order to obtain the feature values of the key appearance features, wherein the preset order meets the input feature vector requirement of the hardness prediction model. Specifically, all the calculated feature values of the key appearance features are combined in a preset order to obtain the feature values of the key appearance features that meet the input feature vector requirement of the hardness prediction model. The preset order is determined according to the arrangement requirement of the hardness prediction model for the input feature vector. Different hardness prediction models can have different input feature vector order requirements. Only when the feature values are combined in the correct order, the hardness prediction model can accurately receive and process these feature values, thereby performing hardness prediction.

[0036] In step S200, the constraint force analysis of clamping and core removal is performed based on the fruit category and the hardness prediction result, the target balance configuration is performed according to the clamping constraint force combined with the core removal constraint force, and the cooperative control parameter is determined.

[0037] Specifically, the clamping constraint force refers to the force exerted by the mechanical clamping jaw on the fruit, which needs to balance the fixing stability and the flesh protection. The coring constraint force refers to the force exerted by the coring needle or rotating knife head on the fruit core, which needs to overcome the adhesion force between the fruit core and the flesh. According to the fruit category and hardness prediction result, the clamping force and coring force are queried from the preset constraint force database or dynamically calculated through a mechanical model, to ensure that the clamping force does not cause flesh bruising and the coring force can overcome the adhesion force of the fruit core. A multi-objective optimization algorithm is used to balance the clamping force and coring force, for example, the optimization objectives are to minimize the flesh damage rate and maximize the coring success rate, a comprehensive score is generated by a weighted summation method, an optimal parameter combination is selected, and a synergistic control parameter is obtained, including force size, action time, acceleration, etc., which is used to coordinate the clamping and coring actions.

[0038] In one possible implementation, based on the fruit category and hardness prediction result, the clamping and coring constraint forces are analyzed, the target balance configuration is performed according to the clamping constraint force combined with the coring constraint force, and the synergistic control parameter is determined. Step S200 further includes step S210 of obtaining the clamping force constraint range and the coring force constraint range of the current fruit through fuzzy database mapping according to the fruit category and hardness prediction result. Specifically, the fuzzy database is constructed based on a large amount of experimental data and expert experience. In the construction process, the relevant data of the clamping force and the coring force of different fruit categories under different hardness conditions are collected, and the fuzzy mathematical method is used to process and store these data to form the fuzzy database. That is, the clamping force constraint range and the coring force constraint range corresponding to different fruit categories and hardness levels are preset in the fuzzy database. For example, for a peach with a hardness prediction result of hard, the fuzzy database specifies that the clamping force constraint range is 10-15 N and the coring force constraint range is 20-25 N. According to the fruit category and hardness prediction result obtained in step S100, the fuzzy database is queried and mapped. Through mapping, the reasonable interval of the clamping force and the coring force allowed by the current fruit during operation is quickly obtained.

[0039] Step S220, based on the clamping force constraint range, a clamping target function is established for the fruit damage, and based on the coring force constraint range, a coring target function is established for the successful coring and equipment protection. Specifically, fruit damage is a factor that needs to be considered in the clamping process. If the clamping force is too small, the fruit may slip during coring, affecting the coring effect. If the clamping force is too large, it is easy to cause indentation, damage and other damage on the surface of the fruit. Therefore, when establishing the clamping target function, the degree of fruit damage is taken as the core index. For example, a function related to the clamping force and the stress on the surface of the fruit can be defined, and the specific form of the function is determined through experiments and theoretical analysis, so that the function can accurately reflect the influence of the clamping force on the fruit damage. For example, if the stress on the surface of the fruit is proportional to the clamping force, and the degree of fruit damage is proportional to the area of the surface stress exceeding a certain threshold, then the clamping target function can be expressed as minimizing the function value related to the degree of fruit damage under the premise of meeting the clamping force constraint range.

[0040] Successful coring is the core goal of coring operation, and the equipment also needs to be protected from damage. If the coring force is too small, the fruit core may not be completely removed; if the coring force is too large, it may cause damage to the coring tool or excessive impact on other parts of the equipment. Therefore, the coring target function needs to consider factors such as the probability of successful coring and the impact force on the equipment. For example, a function related to the coring force, the success rate of coring and the impact force on the equipment can be defined, and the specific form of the function is determined through experiments and simulation analysis, so that the function can accurately reflect the influence of the coring force on the successful coring and equipment protection. For example, if the success rate of coring is proportional to the coring force within a certain range, and the impact force on the equipment is proportional to the square of the coring force, then the coring target function can be expressed as maximizing the success rate of coring while minimizing the function value related to the impact force on the equipment under the premise of meeting the coring force constraint range.

[0041] Step S230, according to the clamping target function and the coring target function, balanced optimization is performed to obtain a combination of clamping parameter and coring parameter control strategies that maximizes the clamping target and the coring target, and generate the cooperative control parameters. Specifically, according to the clamping target function and the coring target function established in step S220, an optimization algorithm is used for balanced optimization, and the multiple targets are weighed and optimized to find the optimal solution that meets the requirements of multiple targets. In the optimization process, the clamping target and the coring target are considered simultaneously, and the values of the clamping parameters and the coring parameters are adjusted continuously to make the values of the clamping target function and the coring target function as optimal as possible. For example, when using a multi-objective genetic algorithm, the clamping parameters and the coring parameters are encoded as chromosomes, and the population is continuously evolved through selection, crossover and mutation operations. Excellent individuals are selected according to the fitness values of the clamping target function and the coring target function, and after multiple generations of evolution, an approximate optimal solution that meets the requirements is obtained.

[0042] After the balanced optimization, the clamping parameters and the core removal parameter control strategy combination that satisfy the clamping target and the core removal target maximization are obtained. These parameter combinations are the cooperative control parameters, which are used to control the clamping mechanism and the core removal mechanism of the intelligent core removal equipment. For example, the cooperative control parameters include the clamping force size, the clamping time of the clamping mechanism, the core removal force size, the core removal speed of the core removal mechanism, etc. By inputting these cooperative control parameters into the control system of the equipment, the clamping and core removal operations can be accurately controlled, and the performance and reliability of the core removal equipment are improved.

[0043] In step S300, the control execution is performed according to the cooperative control parameters, the execution feedback data is tracked, the adaptive analysis is performed based on the execution feedback data, and the execution compensation adjustment parameters are identified.

[0044] Specifically, the clamping force and the core removal force are monitored in real time by a force sensor such as a strain gauge sensor, and the mechanical arm displacement data fed back by an encoder are combined to construct an execution feedback data set to evaluate the control effect. The force change rate is calculated by using a sliding window, and abnormal events such as meat sliding and core sticking are identified by threshold comparison and other methods. According to the specific type of the abnormal event, the control parameters are adjusted to adapt to the environmental changes, and the execution compensation adjustment parameters are generated.

[0045] In one possible implementation, the control execution is performed according to the cooperative control parameters, the execution feedback data is tracked, the adaptive analysis is performed based on the execution feedback data, and the execution compensation adjustment parameters are identified, and step S300 further includes step S310 of controlling the clamping and core removal execution mechanisms to act based on the cooperative control parameters, and collecting clamping force feedback data, core removal force feedback data, and fruit visual data after clamping in real time. Specifically, the cooperative control parameters are transmitted to the control system of the intelligent core removal equipment, and the control system controls the clamping execution mechanism and the core removal execution mechanism according to these parameters. The clamping execution mechanism clamps the fruit stably according to the clamping parameters such as the clamping force size and the clamping time, and the core removal execution mechanism performs the core removal operation according to the core removal parameters such as the core removal force size and the core removal speed.

[0046] During the action of the execution mechanism, key feedback data are collected in real time by various sensors. Among them, the force sensors are installed on the clamping mechanism and the core removal mechanism to collect clamping force feedback data and core removal force feedback data, respectively. These data reflect the size and change of the actual force acting on the fruit. At the same time, the visual sensor is used to collect the fruit visual data after clamping, which contains the appearance information of the fruit, including the shape, size, core position, etc.

[0047] Step S320, based on the pitting force feedback data and the pitting force expected value in the cooperative control parameter, the hardness deviation between the actual hardness of the fruit and the predicted hardness is analyzed. Specifically, the pitting force feedback data collected from the pitting execution mechanism is compared with the pitting force expected value preset in the cooperative control parameter. The pitting force expected value is determined based on the fruit category and the hardness prediction result, combined with the pitting target function and other factors, which represents the ideal pitting force required to complete the pitting operation under the current predicted hardness. If there is a difference between the pitting force feedback data and the expected value, it means that the actual hardness of the fruit is not consistent with the predicted hardness. According to this difference, the hardness deviation between the actual hardness of the fruit and the predicted hardness is analyzed. For example, if the actual pitting force is greater than the expected value, it means that the actual hardness of the fruit is higher than the predicted hardness; on the contrary, if the actual pitting force is less than the expected value, the actual hardness of the fruit is lower than the predicted hardness.

[0048] Step S330, based on the visual data of the fruit after clamping, the spatial position of the kernel in the clamped state is recalculated and compared with the kernel model position pre-established by the image information, and the spatial deviation of the kernel position is obtained. Specifically, the spatial position of the kernel in the clamped state is recalculated by using the collected visual data of the fruit after clamping, combined with image processing algorithm. The image processing algorithm can analyze the visual data, identify the contour, feature points and other information of the fruit, and then determine the coordinate position of the kernel in the current state through geometric calculation and model matching. The spatial position of the kernel recalculated is compared with the kernel model position pre-established by the image information. The kernel model position is obtained by analyzing and modeling a large number of sample fruit image data before the device runs, which represents the position of the kernel in the fruit in the ideal state. By comparing the difference between the two, the spatial deviation of the kernel position is analyzed, which reflects the deviation of the kernel position in the actual pitting operation.

[0049] Step S340, according to the hardness deviation, the clamping force compensation correction is carried out, and the clamping force set value and the parameter adjustment amount of the pitting executor are calculated. Specifically, according to the size and direction of the hardness deviation analyzed in step S320, the actual hardness of the fruit is calculated, and the target balance configuration is re-performed according to the actual hardness of the fruit by using the similar method as steps S210-230 to determine the new cooperative control parameter. The original clamping force set value is compensated and corrected according to the new cooperative control parameter, so that the clamping force can better adapt to the actual hardness of the fruit. At the same time, the parameters of the pitting executor are adjusted according to the new cooperative control parameter, to ensure that the pitting operation can be carried out smoothly.

[0050] Step S350, according to the spatial deviation, calculate the position correction of the kernel remover motion path, based on the parameter adjustment and position correction constitute the execution compensation adjustment parameter. Specifically, according to the kernel position spatial deviation obtained in step S330, calculate the position correction of the kernel remover motion path. If the fruit kernel position deviates in a certain direction, adjust the motion trajectory of the kernel remover, so that it can accurately reach the fruit kernel position for kernel removal operation. Through geometric calculation and motion control algorithm, determine the position correction of the kernel remover in each coordinate axis direction, ensure that the kernel remover can move according to the corrected path. The clamping force set value calculated in step S340, the parameter adjustment of the kernel remover, and the position correction calculated in step S350 are integrated together to constitute the complete execution compensation adjustment parameter. These parameters are used for further control of the execution mechanism, realize the dynamic adjustment of clamping and kernel removal operation, ensure that the whole kernel removal process can be optimized according to the actual situation, improve the precision and success rate of kernel removal.

[0051] In a possible implementation, based on the visual data of the clamped fruit, the spatial position of the core in the clamped state is recalculated and compared with the pre-established core model position in the image information, and the spatial deviation of the core position is obtained. Step S330 further includes step S331 of acquiring the image of the clamped fruit by a visual collector, wherein the visual collector includes a visible light camera and an infrared camera. The image acquired by the visible light camera is used to identify the external contour of the clamped fruit, and based on the pre-calibrated conversion relationship, the spatial position coordinates of the fruit in the core remover coordinate system are determined. The image acquired by the infrared camera is used to identify the core feature, and the core spatial position coordinates in the core remover coordinate system are determined. Specifically, the visible light camera in the visual collector acquires the image of the clamped fruit. The visible light camera can capture the color, texture and other visual information of the fruit surface. The acquired visible light image is analyzed by using an image processing algorithm to identify the external contour of the fruit. The image processing algorithm can use an edge detection algorithm, which can accurately detect the pixel points of the object edge in the image, thereby outlining the contour shape of the fruit. In addition, morphological processing algorithms such as dilation and erosion operations can be combined to optimize and repair the contour, remove noise interference, and make the contour smoother and more accurate. After identifying the external contour of the fruit, based on the pre-calibrated conversion relationship, the contour information in the image is converted into the clamped spatial position coordinates of the fruit in the core remover coordinate system. The pre-calibration is a step performed during the installation and debugging of the device, and the corresponding relationship between the image coordinate system and the core remover coordinate system is established by using a standard calibration object such as a checkerboard calibration board. For example, the corner point coordinates of the checkerboard calibration board in the image and the actual coordinates in the core remover coordinate system are known, and the conversion matrix between the two can be obtained by calculation. By using this conversion matrix, the coordinates of the fruit contour in the image can be converted into the actual coordinates in the core remover coordinate system, thereby determining the clamped spatial position of the fruit.

[0052] The infrared camera captures the infrared image of the fruit after clamping by taking advantage of the difference in thermal radiation characteristics between the core and the flesh of the fruit. Different substances emit different intensities and wavelengths of infrared radiation due to their molecular structure and temperature. The core has different thermal characteristics from the flesh and will show unique brightness or texture features in the infrared image, which allows the infrared camera to capture information about the core. Image recognition algorithms are used to analyze the infrared image to identify the characteristics of the core. For example, a convolutional neural network can be used to automatically learn the feature patterns of the core in the infrared image by training a large number of sample data, so that the infrared image can be feature extracted and classified to identify the position and shape of the core. Similar to the visible light camera, after identifying the core features, the position information of the core in the infrared image is converted to the core space position coordinates in the core executor coordinate system according to the pre-calibrated conversion relationship.

[0053] In step S332, the clamping space position coordinates and the core space position coordinates are centered and compared with the core model position to obtain the spatial deviation. Specifically, the clamping space position coordinates of the fruit and the core space position coordinates of the core are centered to eliminate the influence of the overall position and attitude of the fruit on the analysis of the core position. Since the fruit may be tilted or offset during clamping, directly comparing the position coordinates of the two will result in errors. By centering, the position of the core relative to the center of the fruit is standardized. Specifically, the center points of the clamping space position coordinates of the fruit and the core space position coordinates of the core are calculated. For a two-dimensional coordinate system, the center point can be obtained by calculating the average of the coordinates; for a three-dimensional coordinate system, the average of the three coordinate axes is calculated respectively. Then, the core center point is translated relative to the fruit center point to make them coincide, completing the centering operation.

[0054] The core position coordinates after centering are compared with the pre-established core model position of the image information. When comparing, various methods can be used, such as calculating the Euclidean distance, Manhattan distance, etc. between the two to quantify the difference between the actual position of the core and the model position. According to the comparison result, the spatial deviation of the core position is determined, including the component deviation in each coordinate axis direction and the total spatial distance deviation. For example, if the core is offset by 2 mm in the X-axis direction, 1 mm in the Y-axis direction, and 0 mm in the Z-axis direction compared to the model position, the component deviations can be recorded, and the total spatial distance deviation is calculated to be 2.24 mm.

[0055] In one possible implementation, clamping force compensation correction is performed based on the hardness deviation, and the clamping force setpoint and parameter adjustment amount of the pitting actuator are calculated. Step 340 further includes step S341, determining the current hardness loss tolerance coefficient based on the hardness deviation and referring to historical experience data. Specifically, during long-term operation of the equipment, a large amount of relevant data on fruits with different hardness during the pitting process is collected and organized. This data includes information such as fruit hardness, clamping force and pitting parameters used during pitting, and the final fruit pulp loss. Through classification and statistical analysis of this data, a database of correspondences between different hardness ranges and fruit pulp loss is established. When the current hardness deviation of the fruit is obtained, the range in which the current hardness of the fruit falls is determined based on this deviation. Then, the acceptable proportion of fruit pulp loss within this hardness range, under the premise of ensuring the pitting effect, is searched from the historical experience data correspondence database and defined as the loss tolerance coefficient. For example, if the current fruit hardness deviation shows that its actual hardness is 2 N / cm higher than the predicted hardness, the tolerance coefficient is defined as follows: 2 7-9 N / cm 2 Based on the hardness range, a database query revealed that the acceptable percentage of pulp loss within this range is 5%-10%. Therefore, the loss tolerance coefficient can be determined to be a value between 0.05 and 0.1. This coefficient is used to adjust the estimated amount of pulp loss in subsequent adjustments to ensure that pulp loss is not excessively increased when compensating for clamping force.

[0056] Step S342: Evaluate the pulp loss based on the actual hardness feedback according to the current collaborative control parameters to obtain the estimated pulp loss under the current operating conditions. Specifically, a pulp loss evaluation model based on collaborative control parameters is established by combining the physical characteristics of the fruit, such as elastic modulus and density, and the mechanical principles of the pitting process. For example, through experiments and theoretical analysis, it is found that the clamping force is directly proportional to the degree of pulp deformation under compression; the faster the pitting speed, the greater the impact force on the pulp, leading to a corresponding increase in pulp loss. These relationships are quantified and integrated into the evaluation model, enabling it to calculate the estimated pulp loss based on the input collaborative control parameters and the actual hardness value. The currently used collaborative control parameters, including the clamping force setpoint, the expected pitting force, and the pitting speed, are obtained from the equipment's control system. Simultaneously, the actual hardness value of the fruit is calculated based on the hardness deviation and predicted hardness obtained in step S320. The current collaborative control parameters and the actual hardness value are input into the pulp loss evaluation model, and the estimated pulp loss under the current operating conditions is obtained through the model's calculation.

[0057] Step S343, the loss tolerance coefficient is used to adjust the target calculation of the pulp loss estimation, and the clamping parameters and the parameters of the coring operation are balanced and reconstructed according to the adjustment target, and the clamping force setting value and the parameter adjustment amount of the coring executor satisfying the current hardness deviation are calculated. Specifically, the loss tolerance coefficient determined in step S341 is used to adjust the target calculation of the pulp loss estimation obtained in step S342. The setting of the adjustment target is to control the pulp loss in the range allowed by the loss tolerance coefficient on the premise of ensuring the success rate of coring. For example, if the loss tolerance coefficient is 0.08, the pulp loss estimation is 9 grams, and the total pulp weight of the fruit is assumed to be 100 grams, the adjustment target is to control the pulp loss within 100 x 0.08 = 8 grams, and the current pulp loss estimation is 9 grams, which exceeds the range allowed by the loss tolerance coefficient, and the clamping force and the coring parameters need to be adjusted to reduce the pulp loss.

[0058] According to the adjustment target, an optimization algorithm is used to balance and reconstruct the clamping parameters and the coring parameters. After balancing and reconstructing, new clamping force setting values and coring executor parameters are obtained. The new parameters are compared with the original cooperative control parameters to calculate the adjustment amount of the clamping force setting value and the adjustment amount of each parameter of the coring executor. For example, if the original clamping force setting value is 25N and the new clamping force setting value is 32N, the adjustment amount of the clamping force setting value is 32-25=7N; if the original coring speed is 120mm / s and the new coring speed is 90mm / s, the adjustment amount of the coring speed is 90-120=-30mm / s. These parameter adjustment amounts are used to adjust the control parameters of the equipment in real time to adapt to the changes of the actual hardness of the fruit and realize accurate coring.

[0059] Step S400, the cooperative control parameters are dynamically adjusted by using the execution compensation adjustment parameters to complete the coring operation control.

[0060] Specifically, the execution compensation adjustment parameters are input to the PLC or embedded controller, and the cooperative control parameters are dynamically updated by the PID control algorithm or fuzzy control. For example, if the compensation parameters require to increase the clamping force, the controller adjusts the PWM signal duty cycle to increase the output torque of the servo motor to realize real-time adjustment of the clamping force. Finally, the coring operation is completed through the cooperative movement of the execution mechanism such as the mechanical arm, the clamping jaw and the coring needle.

[0061] The embodiment of the application adopts the technical means of obtaining fruit images by a visual collector and identifying, obtaining fruit categories and hardness prediction results, analyzing clamping and coring constraint forces, balancing configuration to determine cooperative control parameters, performing control according to the parameters and tracking feedback data, analyzing execution compensation adjustment parameters, dynamically adjusting the cooperative control parameters by using the parameters, and completing coring operation control, to solve the technical problem that existing coring equipment control cannot adapt to different fruit characteristics, resulting in poor coring effect, and achieve the technical effect of accurately coring fruits with different characteristics and improving juice quality.

[0062] In the foregoing, with reference to Figure 1 The control method of the intelligent coring equipment according to the embodiment of the application is described in detail. Next, the control system of the intelligent coring equipment according to the embodiment of the application will be described with reference to Figure 2 The control system of the intelligent coring equipment according to the embodiment of the application is described in detail. Next, the control system of the intelligent coring equipment according to the embodiment of the application will be described with reference to

[0063] The control system of the intelligent coring equipment according to the embodiment of the application is used to solve the technical problem that existing coring equipment control cannot adapt to different fruit characteristics, resulting in poor coring effect, and achieve the technical effect of accurately coring fruits with different characteristics and improving juice quality. The control system of the intelligent coring equipment includes a fruit external feature identification module 10, a cooperative control parameter determination module 20, an execution feedback module 30, and a parameter dynamic adaptive adjustment module 40.

[0064] The fruit external feature identification module 10 is configured to obtain image information of fruits by a visual collector, identify fruit external features based on the image information, and obtain fruit categories and hardness prediction results. The cooperative control parameter determination module 20 is configured to analyze clamping and coring constraint forces based on the fruit categories and hardness prediction results, perform target balancing configuration according to the clamping constraint force and the coring constraint force, and determine cooperative control parameters. The execution feedback module 30 is configured to perform control according to the cooperative control parameters, track execution feedback data, perform self-adaptive analysis based on the execution feedback data, and identify execution compensation adjustment parameters. The parameter dynamic adaptive adjustment module 40 is configured to dynamically adjust the cooperative control parameters by using the execution compensation adjustment parameters, and complete coring operation control.

[0065] The detailed description of the specific configuration of the fruit external feature recognition module 10 is explained as follows: as described above, the fruit external feature recognition is performed on the image information to obtain the fruit category and hardness prediction result, and the fruit external feature recognition module 10 can further include: a visual feature recognition extraction unit for performing visual feature recognition extraction on the image information to obtain visual features including color, texture, shape contour; a fruit category classification unit for inputting the visual features into a pre-trained fruit category classification model to output a fruit category, the fruit category classification model being a machine learning model based on a support vector machine; and a hardness prediction unit for calling a hardness engine in a pre-stored knowledge base according to the fruit category to perform identification matching and hardness prediction and obtain a hardness prediction result corresponding to the fruit category.

[0066] The hardness prediction unit can further include: a hardness prediction model calling subunit for calling a hardness prediction model and a key appearance feature index related to the fruit category in the pre-stored knowledge base according to the fruit category, wherein the key appearance features include color space component values, skin glossiness and texture roughness; a feature value calculation subunit for quantitatively calculating feature values of the key appearance features from the visual features based on the key appearance feature index; and a hardness prediction subunit for inputting the quantified feature values of the key appearance features into the hardness prediction model to perform hardness prediction of the corresponding fruit category and output a hardness prediction result, which is a data quantization result describing the hardness grade of the fruit category.

[0067] The feature value calculation subunit can further include: a pixel region positioning component for analyzing the key appearance feature index to obtain description information of a region of interest corresponding to the key appearance features, and positioning the corresponding pixel region in the image information based on the description information; a quantitative calculation component for calculating data-based feature values according to pixel data of the positioned pixel region according to a quantitative calculation method defined in the index for each key appearance feature; and a feature value combination component for combining all calculated feature values in a preset order to obtain the feature values of the key appearance features, wherein the preset order meets the input feature vector requirement of the hardness prediction model.

[0068] The pixel region positioning component can further include: the region of interest includes a global region, a fixed region and a relative region, wherein the global region is the entire fruit image, the fixed region is a region defined by an image coordinate range, and the relative region is a region defined based on morphological feature points of the fruit image, and the morphological feature points include a fruit image centroid and a circumscribed rectangle center.

[0069] The detailed description of the specific configuration of the cooperative control parameter determination module 20 is as follows: as described above, based on the fruit category and hardness prediction results, the clamping and core removal constraint force is analyzed, the target balance configuration is performed according to the clamping constraint force combined with the core removal constraint force, the cooperative control parameter is determined, and the cooperative control parameter determination module 20 can further include: a force constraint range acquisition unit for obtaining the clamping force constraint range and the core removal force constraint range of the current fruit according to the fruit category and hardness prediction results through fuzzy database mapping; a target function establishment unit for establishing a clamping target function based on the clamping force constraint range with fruit damage as the target, and establishing a core removal target function based on the core removal force constraint range with successful core removal and equipment protection as the target; a balance optimization unit for performing balance optimization according to the clamping target function and the core removal target function, obtaining a clamping parameter and core removal parameter control strategy combination that satisfies the clamping target and maximizes the core removal target, and generating the cooperative control parameter.

[0070] The detailed description of the specific configuration of the execution feedback module 30 is as follows: as described above, the execution is controlled according to the cooperative control parameter, and the execution feedback data is tracked, the execution compensation adjustment parameter is identified based on the execution feedback data, and the execution feedback module 30 can further include: a feedback data acquisition unit for controlling the clamping and core removal execution mechanism to act based on the cooperative control parameter, and acquiring clamping force feedback data, core removal force feedback data and fruit visual data after clamping in real time; a hardness deviation analysis unit for comparing the core removal force feedback data with the core removal force expected value in the cooperative control parameter, and analyzing the hardness deviation between the actual hardness of the fruit and the predicted hardness; a core position space deviation analysis unit for recalculating the spatial position of the core in the clamped state based on the fruit visual data after clamping, and comparing it with the core model position pre-established by the image information, and analyzing the spatial deviation of the core position; a clamping force compensation correction unit for performing clamping force compensation correction according to the hardness deviation, calculating the clamping force set value and the parameter adjustment amount of the core removal executor; a position correction amount calculation unit for calculating the position correction amount of the core removal executor movement path according to the spatial deviation, and constructing the execution compensation adjustment parameter based on the parameter adjustment amount and the position correction amount.

[0071] The spatial position of the core in the clamped state is recalculated based on the visual data of the clamped fruit, and compared with the pre-established core model position of the image information, and the spatial deviation of the core position is analyzed. The spatial position coordinate determination subunit is used to collect the image of the clamped fruit through the visual collector, wherein the visual collector includes a visible light camera and an infrared camera. The image collected by the visible light camera is used to identify the external contour of the fruit after clamping, and the clamping spatial position coordinates of the fruit in the core removal executor coordinate system are determined based on the pre-calibrated conversion relationship. The image collected by the infrared camera is used to identify the core feature, and the core spatial position coordinates in the core removal executor coordinate system are determined. The spatial deviation acquisition subunit is used to center align the clamping spatial position coordinates and the core spatial position coordinates, and compare them with the core model position to obtain the spatial deviation.

[0072] The clamping force compensation correction unit can further include: a loss tolerance coefficient determination subunit for determining the loss tolerance coefficient of the current hardness based on the hardness deviation and referring to historical experience data; a pulp loss evaluation subunit for evaluating the pulp loss based on the actual hardness feedback according to the current cooperative control parameter to obtain the pulp loss estimation under the current working condition; and a balanced reconstruction subunit for adjusting the target calculation of the pulp loss estimation using the loss tolerance coefficient, and balanced reconstructing the clamping parameters and the core removal parameters according to the adjustment target to calculate the clamping force set value and the parameter adjustment amount of the core removal executor that satisfies the current hardness deviation.

[0073] The control system of the intelligent core removal equipment provided by the embodiment of the application can execute the control method of the intelligent core removal equipment provided by any embodiment of the application, and has the corresponding function modules and beneficial effects of the execution method.

[0074] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, however, any number of different modules can be used and run on the user terminal and / or server, and each unit and module included is only divided according to the functional logic, but is not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific names of each functional unit are only for easy differentiation, and do not limit the protection scope of the present application.

[0075] The foregoing DETAILED DESCRIPTION, including the above section titled "Detailed Description," is not to be taken as limiting the scope of the application. Various modifications, combinations, and equivalents can be apparent to those skilled in the art and can be made once the nature of the application is understood. Any modification, combination, or equivalent, which falls within the principles and the scope of the present application, is intended to be included in the present application. In some instances, the actions or steps can be performed in different order from those described herein, and still achieve desirable results. Additionally, the process depicted in the figures can not necessarily require the particular order shown or sequential order to achieve the desired results. In certain implementations, multitasking and parallel processing can be advantageous.

Claims

1. A control method of an intelligent de-coring apparatus, characterized by, The method comprises the following steps: Obtain image information of the fruit through a visual collector, and identify external characteristics of the fruit based on the image information to obtain a fruit category and a hardness prediction result; Perform constraint force analysis on clamping and pitting based on the fruit category and the hardness prediction result, perform target balanced configuration based on clamping constraint force and pitting constraint force, and determine a cooperative control parameter; Control is performed according to the cooperative control parameter, and feedback data of the execution is tracked, adaptive analysis is performed based on the feedback data of the execution, and an execution compensation adjustment parameter is identified; The cooperative control parameter is dynamically adjusted by using the execution compensation adjustment parameter, and the pitting operation control is completed.

2. The control method of the intelligent coring apparatus according to claim 1, wherein The fruit category and the hardness prediction result are obtained by identifying external characteristics of the fruit based on the image information, and the method comprises the following steps: Visual feature identification and extraction are performed on the image information to obtain visual features, including color, texture, and shape contour; The visual features are input into a pre-trained fruit category classification model to output a fruit category, and the fruit category classification model is a machine learning model based on a support vector machine; A hardness engine is called in a pre-stored knowledge base according to the fruit category to perform identification matching and hardness prediction, and a hardness prediction result corresponding to the fruit category is obtained.

3. The control method of the intelligent coring apparatus according to claim 2, wherein The hardness engine is called in the pre-stored knowledge base according to the fruit category to perform identification matching and hardness prediction, and a hardness prediction result corresponding to the fruit category is obtained, which comprises the following steps: A hardness prediction model and a key appearance feature index related to the fruit category are called in the pre-stored knowledge base according to the fruit category, wherein the key appearance feature includes color space component value, skin glossiness, and texture roughness; Based on the key appearance feature index, the feature values of the key appearance features are quantitatively calculated from the visual features; The feature values of the quantized key appearance features are input into the hardness prediction model to perform hardness prediction of the corresponding category, and a hardness prediction result is output, which is a data quantization result for describing the hardness grade of the fruit category.

4. The control method of the intelligent coring apparatus according to claim 3, wherein Based on the key appearance feature index, the feature values of the key appearance features are quantitatively calculated from the visual features, which comprises the following steps: The description information of the region of interest corresponding to the key appearance feature is obtained by analyzing the key appearance feature index, and the corresponding pixel region is located in the image information based on the description information; For each key appearance feature, the data-based feature value is calculated according to the pixel data of the located pixel region according to the quantitative calculation method defined in the index; All calculated feature values are combined in a preset order to obtain the feature values of the key appearance features, wherein the preset order meets the input feature vector requirement of the hardness prediction model.

5. The control method of the intelligent coring apparatus according to claim 4, wherein The region of interest includes a global region, a fixed region, and a relative region, wherein the global region is the entire fruit image, the fixed region is a region defined by an image coordinate range, and the relative region is a region defined based on the morphological feature points of the fruit image, and the morphological feature points include the image center of the fruit and the center of the circumscribed rectangle. 6.The control method of the intelligent nucleo-free device according to claim 1, characterized in that, ​ According to the fruit category and hardness prediction result, the clamping force constraint range and the core removal force constraint range of the current fruit are obtained through fuzzy database mapping; With fruit damage as the target, a clamping target function is established based on the clamping force constraint range, and with successful core removal and equipment protection as the target, a core removal target function is established based on the core removal force constraint range; According to the clamping target function and the core removal target function, balanced optimization is performed to obtain a clamping parameter and core removal parameter control strategy combination that maximizes the clamping target and the core removal target, and the cooperative control parameters are generated.

7. The control method of the intelligent nucleo-removing apparatus according to claim 1, characterized in that, According to the cooperative control parameters, control execution is performed, and execution feedback data is tracked, and based on the execution feedback data, adaptive analysis is performed to identify execution compensation adjustment parameters, including: The clamping and core removal execution mechanism is controlled to act based on the cooperative control parameters, and clamping force feedback data, core removal force feedback data, and post-clamping fruit visual data are collected in real time; Based on the comparison of the core removal force feedback data and the core removal force expected value in the cooperative control parameters, the hardness deviation between the actual hardness of the fruit and the predicted hardness is analyzed; Based on the post-clamping fruit visual data, the spatial position of the core in the clamped state is recalculated and compared with the pre-established core model position based on image information to analyze the spatial deviation of the core position; According to the hardness deviation, the clamping force set value and the parameter adjustment amount of the core removal executor are calculated; According to the spatial deviation, the position correction amount of the core removal executor motion path is calculated, and the parameter adjustment amount and the position correction amount form the execution compensation adjustment parameters. 8.The control method of the intelligent coring apparatus according to claim 7, wherein, Based on the post-clamping fruit visual data, the spatial position of the core in the clamped state is recalculated and compared with the pre-established core model position based on image information to analyze the spatial deviation of the core position, including: The post-clamping fruit image is collected by a visual collector, which includes a visible light camera and an infrared camera. The image collected by the visible light camera is used for external contour recognition of the clamped fruit, and based on the pre-calibrated conversion relationship, the clamping spatial position coordinates of the fruit in the core removal executor coordinate system are determined. The image collected by the infrared camera is used for core feature recognition to determine the core spatial position coordinates in the core removal executor coordinate system; The clamping spatial position coordinates and the core spatial position coordinates are centered and compared with the core model position to obtain the spatial deviation. 9.The control method of the intelligent coring apparatus according to claim 7, wherein, According to the hardness deviation, the clamping force set value and the parameter adjustment amount of the core removal executor are calculated, including: Based on the hardness deviation, the loss tolerance coefficient of the current hardness is determined based on historical experience data; The actual hardness is evaluated for pulp loss based on the current cooperative control parameters, and the pulp loss estimate under the current working condition is obtained; The loss tolerance coefficient is used to adjust the pulp loss estimate to calculate the adjustment target, and the clamping parameter and core removal parameter are balanced and reconstructed based on the adjustment target to calculate the clamping force set value and the parameter adjustment amount of the core removal executor that satisfies the current hardness deviation.

10. A control system for an intelligent de-coring apparatus, characterized by, The system is used for implementing the control method of the intelligent coring device according to any one of claims 1-9, and the system comprises: a fruit external feature recognition module, configured to acquire image information of the fruit through a visual collector, recognize the fruit external features of the image information, and obtain a fruit category and a hardness pre-judgment result; a cooperative control parameter determination module, configured to analyze clamping and coring constraint forces based on the fruit category and the hardness pre-judgment result, perform target balancing configuration according to the clamping constraint force in combination with the coring constraint force, and determine cooperative control parameters; an execution feedback module, configured to perform control according to the cooperative control parameters, track execution feedback data, perform adaptive analysis based on the execution feedback data, and identify execution compensation adjustment parameters; a parameter dynamic adaptive adjustment module, configured to perform dynamic adaptive adjustment of the cooperative control parameters by using the execution compensation adjustment parameters, and complete coring operation control.

Images