Apple sugar center three-dimensional model construction method based on RIFE interpolation
Through the RIFE interpolation method, combined with BiSeNet and slice stacking three-dimensional model reduction method, the three-dimensional model of apple candy heart is reconstructed, which solves the problem that the existing technology cannot effectively evaluate the spatial structure and overall content of apple candy heart, and achieves high-precision Sugar-center reconstruction and model reliability improvement.
Patent Information
- Application Number
- CN202411498607.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-25
- Publication Date
- 2025-02-21
AI Technical Summary
The existing apple candy heart evaluation methods are mostly a single cross-cut rough evaluation, which cannot effectively evaluate the spatial structure distribution and overall content of apple candy heart, and there is a lack of complete reference for non-destructive testing.
The three-dimensional model construction method of apple candy heart based on RIFE interpolation was used to extract the sugar heart characteristics of various apple slices through BiSeNet, and combined with the slice stacking three-dimensional model reduction method and RIFE interpolation technology, the complete three-dimensional model of apple candy heart was reconstructed.
The high-precision three-dimensional reconstruction of apple candy hearts is realized, which helps to better understand and study the structure and characteristics of apple candy hearts, improves the reliability and credibility of the model, and provides richer information for subsequent research and application.
Smart Images
Figure CN119494918A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of agricultural information technology, and in particular to a method for constructing a three-dimensional model of an apple candy core based on RIFE interpolation. Background Art
[0002] In the field of modern biology and medicine, three-dimensional reconstruction technology has become an important research tool. It can convert two-dimensional image data into three-dimensional models, providing researchers with a more intuitive and comprehensive perspective. Especially in food science, the three-dimensional reconstruction of the internal structure of fruits, such as the reconstruction of the apple heart, is of great significance for understanding its growth mechanism, quality evaluation, and breeding improvement. Since the existing apple heart evaluation methods are mostly a single cross-section to roughly evaluate the heart content of the apple cross section as a reference, there is a lack of research on the spatial structure distribution and overall content evaluation of the apple heart, and the non-destructive detection of the overall heart content of the apple lacks a complete reference, and can only detect the presence or absence of the heart. Therefore, it is necessary to reconstruct the three-dimensional model of the apple heart to provide a scientific and reasonable reference for the spatial structure research and non-destructive detection of the heart.
[0003] Traditional 3D reconstruction methods mainly rely on external mapping, 3D modeling and other methods. However, this method cannot restore the spatial structure of the internal features of the object. The feature slice stacking method can restore the approximate 3D model by obtaining the feature cross-sectional image. However, due to the obvious edge step features of the feature slice stacking method, the model is less refined. Therefore, the RIFE interpolation method is used to smooth the external features of the model, which can make the restored and reconstructed 3D model of the sugar core more accurate. Summary of the invention
[0004] In view of this, the present invention aims at the deficiencies of the prior art and proposes a method for constructing a three-dimensional model of a candy apple core based on RIFE interpolation, aiming to solve at least one of the problems raised in the above background technology.
[0005] The present invention provides a method for constructing a three-dimensional model of a candy apple core based on RIFE interpolation, comprising the following steps:
[0006] S1. Slice the apple sample into equal slices;
[0007] S2, extract the core features of each layer of apple slices through BiSeNet;
[0008] S3, reconstructing the initial three-dimensional model of the sugar core using the sugar core features in each slice by means of a slice stacking three-dimensional model restoration method;
[0009] S4, use RIFE to interpolate the features in the initial model to obtain the sugar core interpolation model;
[0010] S5, integrating the initial model of the candy core and the interpolation model to form a complete three-dimensional model of the candy core, thereby achieving complete reconstruction of the apple candy core;
[0011] S6. Conduct an MIoU verification experiment on the interpolated images to verify the accuracy of the interpolation model and evaluate the accuracy of the sugar core 3D model reconstructed using the slice stacking 3D model reconstruction method.
[0012] In some embodiments, step S1 includes selecting fresh, ripe and undamaged apples, measuring and marking equidistant slicing positions of the apples, and slicing along pre-marked positions using a knife.
[0013] In some embodiments, a cross-sectional image of each slice is acquired by taking a photo.
[0014] In some embodiments, in step S2, a labeled training dataset is used to train the BiSeNet model, and an independent verification dataset is used to verify the trained BiSeNet model.
[0015] In some embodiments, the verified BiSeNet model is used to extract the features of the sugar core of each layer of apple slice image.
[0016] In some embodiments, step S3 includes setting parameters of the slice stacking 3D model restoration method according to the thickness and resolution of the apple slices.
[0017] In some embodiments, step S4 includes inputting the initial three-dimensional model of the sugar core into the RIFE algorithm to perform an interpolation process.
[0018] In some embodiments, step S5 includes ensuring that an initial three-dimensional model of the sugar core is obtained by the slice stacking three-dimensional model restoration method, and a sugar core interpolation model is obtained by interpolation of the RIFE algorithm, setting the parameters of the integration algorithm, inputting the initial three-dimensional model of the sugar core and the interpolation model into the integration algorithm, and executing the integration process.
[0019] In some embodiments, the parameters of the integration algorithm include weight distribution, number of iterations, and learning rate.
[0020] In some embodiments, step S6 includes setting verification parameters of the MIoU algorithm, wherein the verification parameters include a threshold value and a category weight, inputting the sugar center interpolation model and the sugar center initial three-dimensional model into the MIoU algorithm respectively, executing a verification process, and comparing the MIoU values of the two models.
[0021] Compared with the prior art, the beneficial effect of the present invention is that by using BiSeNet to extract the core features of each layer of apple slices, and combining the slice stacking three-dimensional model restoration method and RIFE interpolation technology, the three-dimensional model of the apple core can be accurately reconstructed. This high-precision reconstruction helps to better understand and study the structure and characteristics of the apple core, and the accuracy of the interpolation model can be verified by conducting an interpolation image average intersection over union (MIoU) verification experiment. This method can evaluate the accuracy of the three-dimensional model of the core reconstructed using the slice stacking three-dimensional model reconstruction method, thereby improving the reliability and credibility of the model; integrating the initial core model and the interpolation model to form a complete core three-dimensional model, and realizing the complete reconstruction of the apple core. This complete reconstruction helps to more comprehensively understand the morphology and distribution of the apple core, and provides more abundant information for subsequent research and application. The method of the present application can be applied to the research of apple core, agricultural planting optimization, food processing and other fields. By accurately reconstructing the three-dimensional model of the apple core, it can provide important reference and support for research and practice in related fields.
[0022] The foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure.
[0023] Other features and aspects of the present disclosure will become more apparent from the following detailed description of exemplary embodiments with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0025] Figure 1 A flow chart of a method for constructing a three-dimensional model of a candy apple core based on RIFE interpolation provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0026] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0027] In the description of the present application, it should be understood that the terms "center", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.
[0028] The terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, unless otherwise specified, "plurality" means two or more.
[0029] In the description of this application, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.
[0030] As mentioned in the background technology, the existing apple candy core evaluation methods mostly use a single cross-section to roughly evaluate the candy core content of the apple cross section as a reference, resulting in a lack of research on the spatial structure distribution and overall content evaluation of the apple candy core. The non-destructive detection of the overall candy core content of the apple lacks a complete reference, and can only detect the presence or absence of candy core. Therefore, it is necessary to reconstruct a three-dimensional model of the apple candy core to provide a scientific and reasonable reference for the spatial structure research and non-destructive detection of the candy core.
[0031] Traditional 3D reconstruction methods mainly rely on external mapping, 3D modeling and other methods. However, this method cannot restore the spatial structure of the internal features of the object. The feature slice stacking method can restore the approximate 3D model by obtaining the feature cross-sectional image. However, due to the obvious edge step features of the feature slice stacking method, the model is less refined. Therefore, the RIFE interpolation method is used to smooth the external features of the model, which can make the restored and reconstructed 3D model of the sugar core more accurate.
[0032] In order to improve the above problems, the present application proposes a method for constructing a three-dimensional model of an apple candy core based on RIFE interpolation. By using BiSeNet to extract the candy core features of each layer of apple slices, and combining the slice stacking three-dimensional model restoration method and RIFE interpolation technology, the three-dimensional model of the apple candy core can be accurately reconstructed. This high-precision reconstruction helps to better understand and study the structure and characteristics of the apple candy core. The accuracy of the interpolation model can be verified by conducting an interpolation image average intersection over union (MIoU) verification experiment. This method can evaluate the accuracy of the three-dimensional model of the candy core reconstructed using the slice stacking three-dimensional model reconstruction method, thereby improving the reliability and credibility of the model; integrating the initial model of the candy core with the interpolation model to form a complete three-dimensional model of the candy core, and realizing the complete reconstruction of the apple candy core. This complete reconstruction helps to more comprehensively understand the morphology and distribution of the apple candy core, and provides more abundant information for subsequent research and application. The method of the present application can be applied to the research of apple candy core, agricultural planting optimization, food processing and other fields. By accurately reconstructing the three-dimensional model of the apple candy core, it can provide important reference and support for research and practice in related fields.
[0033] See also Figure 1 As shown, according to an embodiment of the present application, a method for constructing a 3D model of a candy apple core based on RIFE interpolation includes the following steps:
[0034] S1. Slice the apple sample into equal slices;
[0035] S2, extract the core features of each layer of apple slices through BiSeNet;
[0036] S3, reconstructing the initial three-dimensional model of the sugar core using the sugar core features in each slice by means of a slice stacking three-dimensional model restoration method;
[0037] S4, use RIFE to interpolate the features in the initial model to obtain the sugar core interpolation model;
[0038] S5, integrating the initial model of the candy core and the interpolation model to form a complete three-dimensional model of the candy core, thereby achieving complete reconstruction of the apple candy core;
[0039] S6. Conduct an MIoU verification experiment on the interpolated images to verify the accuracy of the interpolation model and evaluate the accuracy of the sugar core 3D model reconstructed using the slice stacking 3D model reconstruction method.
[0040] Specifically, the purpose of isometric slicing is to obtain continuous two-dimensional image data of the inside of the apple, and the apple is cut into equidistant slices along its axis using precision cutting equipment. The slice thickness is usually thin (such as 1-2 mm) to ensure sufficient resolution to observe the internal structure. BiSeNet is a deep learning model specifically designed for image segmentation tasks that can accurately distinguish different regions. In this application, it is used to identify the core area in each slice and output a binary mask of the core area of each slice. By stacking the core masks of all slices, a rough three-dimensional core model is formed. This method combines modern deep learning technology and advanced interpolation algorithms to provide new possibilities for accurately reconstructing complex biological structures. It is not only suitable for the study of apple cores, but in theory it can also be extended to the internal structure analysis of other agricultural products or biological samples, and has potential application value in agricultural science research, quality inspection of the food industry, and the field of biotechnology.
[0041] In some specific embodiments, step S1 includes selecting fresh, ripe and undamaged apples, measuring and marking equidistant slicing positions of the apples, and slicing along pre-marked positions using a knife.
[0042] Specifically, fresh apples can better reflect their internal structure in their natural state and avoid changes in internal structure caused by long storage time. Selecting apples without damage can ensure the integrity of the internal structure and avoid structural damage or deformation caused by damage, which will affect the accuracy of subsequent feature extraction and model reconstruction. By accurately measuring and marking the equidistant slice positions of apples, the thickness of each slice can be ensured to be consistent. This helps to maintain data consistency and continuity in subsequent image processing and 3D reconstruction. Manual or mechanical precision cutting can minimize errors in the slicing process and ensure the flatness and uniformity of the slices, which is crucial for high-quality image acquisition and analysis. Equidistant slices can provide evenly distributed data points, which is very important for building accurate 3D models. Uniform data can reduce possible deviations and discontinuities during model reconstruction. Through accurate slicing, a detailed view of the internal structure of the apple can be obtained, which is very helpful for identifying and analyzing the characteristics of the sugar core. High-quality slice images can reduce the amount of work required in the image preprocessing stage, such as image correction, denoising, etc. For the step of feature extraction using deep learning models (such as BiSeNet), high-quality input data can improve the training effect and final performance of the model. The meticulous operation of selecting, measuring, marking, and slicing apples in step S1 is the foundation of the entire 3D model building process, which directly affects the accuracy and practicality of the final model. By ensuring the high quality of data and the accuracy of processing, a solid foundation can be laid for subsequent steps, thereby improving the effectiveness and reliability of the entire model building method.
[0043] In some specific embodiments, a cross-sectional image of each slice is acquired by taking a photo.
[0044] Specifically, by taking pictures, the characteristics of the core of each slice can be recorded in detail, and these pictures can serve as the basis for extracting and analyzing the core characteristics in subsequent steps. Taking pictures with a camera is a non-contact measurement method that does not cause any physical damage to the apple slices, ensuring the integrity of the experimental samples. Digital images are easy to store and transmit, which facilitates data sharing between different devices and locations, and is conducive to long-term data preservation and remote analysis. Using computer vision technology to automatically process images can quickly and accurately extract the characteristic information of the core, greatly improving the efficiency of data analysis. The image intuitively shows the distribution of the core of the apple, which helps researchers to more intuitively understand the structure and distribution of the core. By comparing the images of different slices, the changing trend of the core as the apple grows can be observed, and it can also be used to verify the accuracy of the model. The captured images can not only be used for feature extraction of deep learning models such as BiSeNet, but can also be applied to other image processing algorithms, such as edge detection, texture analysis, etc., providing a variety of possible analysis methods for research.
[0045] In some specific embodiments, in step S2, a labeled training dataset is used to train the BiSeNet model, and an independent verification dataset is used to verify the trained BiSeNet model.
[0046] Specifically, by using a pre-labeled training dataset, the BiSeNet model was able to learn how to accurately identify and segment the candy core area in apple slices. This supervised learning method helps the model capture key visual features and patterns. Using an independent dataset that was not involved in the training process for validation can objectively evaluate the performance of the model on new data, which is an important means of testing the generalization ability of the model. If the model can perform well on an independent dataset, it means that it has good generalization ability and can handle unseen data. Separating the training set and the validation set is a standard practice in machine learning, which helps prevent the model from overfitting to noise or specific samples in the training data, a phenomenon known as overfitting. By testing on an independent validation set, it can be ensured that the model learns the general laws of the data rather than accidental ones. Training the BiSeNet model with a labeled training dataset and validating it with an independent validation dataset are key steps to ensure the accuracy, generalization ability, and practicality of deep learning models. This not only helps improve model performance, but also promotes the rigor of scientific research and the reliability of technology.
[0047] In some specific embodiments, the verified BiSeNet model is used to extract the features of the sugar core of each layer of apple slice image.
[0048] Specifically, BiSeNet is a proven deep learning model that has shown good performance on multiple datasets, so it can accurately extract the features of the candy core from apple slice images. The BiSeNet model can quickly process large amounts of image data, which is very important for experiments that require the analysis of a large number of samples, and can significantly improve the efficiency of data processing and analysis. The BiSeNet model can be used to automate the feature extraction process, reduce the need for manual operations, reduce the possibility of human errors, and improve the consistency of the workflow. Once the BiSeNet model is trained and validated for the features of the apple candy core, it can be easily applied to new datasets, supporting the expansion of future research and the analysis of new samples. The BiSeNet model designs spatial paths and contextual paths to capture feature information at different scales, which helps to understand the structure and distribution of the apple candy core from different levels.
[0049] In some specific embodiments, step S3 includes setting parameters of the slice stacking 3D model restoration method according to the thickness and resolution of the apple slices.
[0050] Specifically, by taking the actual thickness of the slices into account, the three-dimensional structure inside the apple can be restored more accurately. This is because each slice represents a layer of a certain thickness inside the apple, and the correct thickness information helps to accurately locate the position of each feature in space. Adjusting the parameters of the model according to the resolution of the image ensures that the level of detail of the model matches the resolution of the original data, thereby retaining more detail information. Reasonable setting of parameters can reduce unnecessary calculations. For example, if the slices are thin and the resolution is high, the complexity of the interpolation or reconstruction algorithm can be appropriately reduced to save computing resources. By adjusting the parameters, the best balance can be found between the accuracy of the model and the computational efficiency, which is particularly important when dealing with large amounts of data. Since the size, shape and internal structure of apples may vary, methods that can adjust parameters according to specific circumstances are more flexible and applicable. Setting the parameters of the slice stacking three-dimensional model restoration method according to the thickness and resolution of the apple slices is a key step to ensure model accuracy, optimize computational efficiency, enhance model adaptability, improve model quality, and promote subsequent analysis and application.
[0051] In some specific embodiments, step S4 includes inputting the initial three-dimensional model of the sugar core into the RIFE algorithm to perform an interpolation process.
[0052] Specifically, the RIFE algorithm generates a more continuous animation or image sequence by inserting additional virtual frames between adjacent slices, which helps to eliminate gaps in the original data and make the model smoother. Smooth transitions and continuity are critical to visual effects, especially in 3D reconstruction and visualization applications, providing a more natural and realistic view. Through the interpolation process, the features of the candy apple heart can be further refined, revealing more details that may not be directly observed, which is very important for scientific research and quality assessment. Interpolation can also indirectly improve the spatial resolution of the model, so that the model still maintains clear edges and structures when it is enlarged or rotated. The use of the RIFE algorithm ensures the integrity of the data during the conversion from two-dimensional slices to three-dimensional models, reducing the risk of information loss.
[0053] In some specific embodiments, step S5 includes ensuring that an initial three-dimensional model of the sugar core obtained by the slice stacking three-dimensional model restoration method and an interpolation model of the sugar core obtained by interpolation through the RIFE algorithm are already available, setting parameters of the integration algorithm, inputting the initial three-dimensional model of the sugar core and the interpolation model into the integration algorithm, and executing the integration process.
[0054] In some specific embodiments, the parameters of the integration algorithm are set including weight distribution, number of iterations, and learning rate.
[0055] In some specific embodiments, step S6 includes setting verification parameters of the MIoU algorithm, the verification parameters include thresholds and category weights, inputting the sugar center interpolation model and the sugar center initial three-dimensional model into the MIoU algorithm respectively, executing the verification process, and comparing the MIoU values of the two models.
[0056] Specifically, by calculating the MIoU value, the performance of the two models can be objectively quantified, providing a basis for selecting the best model. MIoU is a commonly used metric for evaluating the accuracy of pixel classification, especially in image segmentation tasks. Using MIoU as an evaluation criterion provides a unified metric for comparison between different models, which helps to fairly evaluate the pros and cons of the models. By comparing the MIoU values, researchers or engineers can select the model with better performance for subsequent analysis and application, ensuring that the research or product development is based on the best data. The sugar core interpolation model and the sugar core initial three-dimensional model are respectively input into the MIoU algorithm to perform the verification process, and the MIoU values of the two models are compared, which not only helps to quantify and optimize the model performance, but also promotes the rigor, transparency and multi-angle analysis of the research, providing a solid foundation for scientific research and industrial applications.
[0057] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.
Claims
1. A method for constructing a three-dimensional model of a candy apple core based on RIFE interpolation, characterized in that: The following steps are involved: S1. Slice the apple sample into equal slices; S2, extract the core features of each layer of apple slices through BiSeNet; S3, reconstructing an initial three-dimensional model of the sugar core using the sugar core features in each slice by means of a slice stacking three-dimensional model restoration method; S4, use RIFE to interpolate the features in the initial model to obtain the sugar core interpolation model; S5, integrating the initial model of the candy core and the interpolation model to form a complete three-dimensional model of the candy core, thereby achieving complete reconstruction of the apple candy core; S6. Conduct an MIoU verification experiment on the interpolated images to verify the accuracy of the interpolation model and evaluate the accuracy of the sugar core 3D model reconstructed using the slice stacking 3D model reconstruction method.
2. The method for constructing a 3D model of a candy apple core based on RIFE interpolation according to claim 1, characterized in that: The step S1 includes selecting fresh, ripe and undamaged apples, measuring and marking equidistant slicing positions of the apples, and slicing along the pre-marked positions using a knife.
3. The method for constructing a three-dimensional model of a candy apple core based on RIFE interpolation according to claim 2, characterized in that: Take a picture to obtain a cross-sectional image of each slice.
4. The method for constructing a 3D model of a candy apple core based on RIFE interpolation according to claim 1, characterized in that: In step S2, the BiSeNet model is trained using a labeled training data set, and the trained BiSeNet model is verified using an independent verification data set.
5. The method for constructing a 3D model of a candy apple core based on RIFE interpolation according to claim 4, characterized in that: The verified BiSeNet model is used to extract the features of the sugar core of each layer of apple slice images.
6. The method for constructing a 3D model of a candy apple core based on RIFE interpolation according to claim 1, characterized in that: The step S3 includes setting the parameters of the slice stacking 3D model restoration method according to the thickness and resolution of the apple slices.
7. The method for constructing a 3D model of a candy apple core based on RIFE interpolation according to claim 1, characterized in that: The step S4 includes inputting the initial three-dimensional model of the sugar core into the RIFE algorithm to perform an interpolation process.
8. The method for constructing a 3D model of a candy apple core based on RIFE interpolation according to claim 1, characterized in that: The step S5 includes ensuring that the initial three-dimensional model of the sugar core obtained by the slice stacking three-dimensional model restoration method and the sugar core interpolation model obtained by the RIFE algorithm interpolation are already available, setting the parameters of the integration algorithm, inputting the initial three-dimensional model of the sugar core and the interpolation model into the integration algorithm, and executing the integration process.
9. The method for constructing a 3D model of a candy apple core based on RIFE interpolation according to claim 8, characterized in that: The parameters of the integration algorithm include weight distribution, number of iterations, and learning rate.
10. The method for constructing a 3D model of a candy apple core based on RIFE interpolation according to claim 1, characterized in that: The step S6 includes setting verification parameters of the MIoU algorithm, wherein the verification parameters include a threshold value and a category weight, inputting the sugar core interpolation model and the sugar core initial three-dimensional model into the MIoU algorithm respectively, executing a verification process, and comparing the MIoU values of the two models.