Method and system for detecting fertilization state of early incubation duck eggs based on transmission imaging

Abstract

The application belongs to the technical field of fertilization state detection of hatching eggs, and specifically discloses a fertilization state detection method and system for hatching duck eggs in the early stage of hatching based on transmission imaging, which comprises the following steps: placing a whole tray of hatching duck eggs in a transmission imaging device to obtain a whole-tray transmission image; preprocessing the whole-tray transmission image to obtain a sub-image region, and constructing an effective detection region for the sub-image region; extracting features from the effective detection region to obtain enhanced features; performing spatial re-labeling on the enhanced features to obtain spatially focused enhanced features; inputting the spatially focused enhanced features into a fertilization state discrimination model to output a fertilization state discrimination result of the hatching duck eggs; and performing reliability evaluation on the fertilization state discrimination result to complete the fertilization state detection of the hatching duck eggs in the early stage of hatching. The application solves the problems of poor imaging consistency, difficulty in stable construction of effective detection regions, and insufficient engineering reliability of discrimination results in the application of the existing hatching egg fertilization state detection technology in whole-tray detection and the early stage of hatching.

Description

Technical Field

[0001] This invention belongs to the field of fertilization status detection technology in the early stage of hatching eggs, specifically involving a method and system for detecting the fertilization status of duck eggs in the early stage of hatching based on transmission imaging. Background Technology

[0002] In the hatching and production of fertilized eggs, the fertilization status of the eggs directly affects hatching efficiency, energy utilization, and the quality of chicks. Infertile eggs cannot develop into normal embryos during incubation, yet they enter the incubation process alongside fertilized eggs, occupying incubation space and continuously consuming incubation resources such as heating and ventilation. If not identified and removed in the early stages, infertile eggs may also rot and crack in the high-temperature, high-humidity incubation environment, adversely affecting the incubation environment of surrounding normally fertilized eggs, thereby reducing the stability and survival rate of the entire batch. Therefore, effective screening and removal of infertile eggs in the early stages of incubation is a crucial step in improving hatching efficiency, reducing production costs, and ensuring hatching safety.

[0003] In current hatching production, the screening of fertilization status of hatching eggs is mostly concentrated in the mid-to-late stages of incubation, with manual candling remaining one of the most commonly used detection methods. Because the blastodisc inside the egg has not yet formed a vascular structure that can be directly identified by the naked eye in the early stages of incubation, it is difficult to reliably determine fertilization by visual inspection. Therefore, manual candling is usually scheduled after about 7 days of incubation, when the vascular network inside the egg gradually becomes clearer, facilitating manual observation and judgment. However, in large-scale hatching scenarios, the number of hatching eggs in a single batch often reaches tens of thousands. Relying solely on manual candling is not only inefficient and labor-intensive, but the results are also highly dependent on the operator's experience, making it difficult to meet the actual needs of modern hatching production for continuous, automated, and consistent processes.

[0004] With the continuous improvement of the scale and automation level of poultry hatching production, researchers have attempted to automate the screening of fertilization status of hatching eggs through optical detection and machine vision, and have proposed a variety of detection methods and devices. For example, machine vision-based methods for detecting the fertilization status of poultry eggs, photoelectric detection schemes that integrate physical parameters of hatching eggs, and the use of visible light or near-infrared spectroscopy to detect fertilized poultry eggs.

[0005] However, the aforementioned existing technical solutions generally face certain limitations in engineering applications. On the one hand, most solutions rely on the significant presentation of embryonic structure or vascular features in the mid-to-late stages of incubation, and their effective discrimination information does not yet have stable visibility in the early stages of incubation. On the other hand, under the conditions of the early stages of incubation (e.g., about 24 hours), the embryonic structure is in a weakly visible or even invisible state, and the slight differences related to the blastodisc in the transmission image are easily interfered with by factors such as the bright edges of the eggshell, unstable background light field, and imaging noise. This makes it difficult to stably determine the effective detection area and the features used for discrimination are not robust enough, thus limiting the accuracy and consistency of early fertilization state discrimination.

[0006] Furthermore, for infertile eggs, the longer they remain in the high-temperature, high-humidity environment of the incubator, the more significant the decline in their quality and freshness. This not only reduces the reuse value of infertile eggs as food or processing raw materials but may also increase safety risks during subsequent processing. Therefore, screening infertile eggs only during the mid- or late-stages of incubation is no longer sufficient to meet the practical needs of modern incubation production for the early, efficient utilization and quality control of infertile eggs.

[0007] Based on the above problems, there is an urgent need for a solution to detect the fertilization status of hatching eggs in the early stage of incubation, which can reliably acquire and effectively identify early information related to the germinal disc, so as to promptly screen out infertile eggs in the early stage of incubation, improve the efficiency of incubation resource utilization, and reduce the adverse effects of the decline in the quality of infertile eggs. Summary of the Invention

[0008] The purpose of this invention is to solve the problems of poor imaging consistency, difficulty in stably constructing the effective detection area, and insufficient engineering reliability of the discrimination results in the application of existing egg fertilization status detection technology in whole-plate detection and early incubation stages. This invention proposes a method and system for detecting the fertilization status of duck eggs in the early incubation stage based on transmission imaging.

[0009] The technical solution of the present invention is as follows: Firstly, a method for detecting the fertilization status of duck eggs in the early stages of incubation based on transmission imaging, comprising the following steps:

[0010] The entire tray of breeding duck eggs is placed in a transmission imaging device, and synchronous transmission imaging is performed through an independent light source unit attached to the bottom of each breeding duck egg to obtain a transmission image of the entire tray.

[0011] The entire transmission image is preprocessed and individual eggs are segmented to obtain sub-image regions that correspond one-to-one with a single breeding duck egg, and effective detection regions are constructed for the sub-image regions.

[0012] Feature extraction is performed on the effective detection area. Based on the radial optical density distribution pattern of the embryonic disc region in the early stage of incubation transmission image, radial structure prior is introduced to model and enhance the features, resulting in radial structure enhanced features, and radial gradient features are constructed.

[0013] The dynamic focus center of the embryonic disc region is adaptively determined based on the enhanced feature response distribution, and a spatial weight mapping is constructed based on the dynamic focus center to spatially recalibrate the enhanced features, thereby obtaining spatial focus enhancement features;

[0014] The spatial focusing enhancement features are input into the fertilization status discrimination model for analysis, and the fertilization status discrimination results of breeding duck eggs are output.

[0015] The reliability of the fertilization status discrimination results is evaluated. A reliability score is generated based on the consistency or confidence between the radial structural enhancement features, spatial focusing enhancement features, and radial gradient features and the fertilization status discrimination results. When the reliability score is lower than the preset threshold, the corresponding duck eggs are marked as pending re-inspection or supplementary analysis is triggered to complete the detection of the fertilization status of duck eggs in the early stage of incubation.

[0016] The beneficial effects of this invention are:

[0017] 1. High-throughput rapid detection of whole trays of breeding duck eggs in the early stage of incubation significantly improves early screening efficiency and engineering application value. This invention focuses on the parallel detection of whole trays of breeding duck eggs in the early stage of incubation. Compared to existing solutions that rely on single-egg detection or candling in the later stages of incubation, it achieves synchronous processing of the entire tray in terms of detection method and moves the detection sequence forward to the stage when the embryonic disc is barely visible. This allows infertile eggs to be identified and removed early in the incubation process, reducing resource consumption and potential risks during subsequent incubation. It has significant early application value and large-scale engineering significance.

[0018] 2. Constructing a stable transmission imaging environment with low crosstalk and low background interference improves the availability of weak germinal disc information in the early stages of incubation. This invention, at the device level, utilizes a collaborative design between the light-absorbing detection disk and the top-covered rubber pad structure of the light source to construct a relatively confined transmission light channel under whole-disc detection conditions. This effectively suppresses background stray light and crosstalk interference between adjacent eggs, improving the signal-to-noise ratio of the transmission image of a single egg and the consistency of the whole-disc imaging. This provides a reliable imaging foundation for the stable acquisition of weak germinal disc information in the early stages of incubation.

[0019] 3. Compatible with multispectral transmission conditions, enhancing the system's versatility and engineering deployment flexibility in different production scenarios. The imaging acquisition component of this invention can be matched or compatible with transmission light source units of different emission bands, enabling the entire detection device to adapt to different spectral conditions, different varieties of duck eggs, and different production line configuration requirements, thereby improving the system's versatility, scalability, and engineering deployment flexibility while ensuring detection performance.

[0020] 4. This invention achieves adaptive construction of the effective detection area for a single egg from a whole tray of images, providing standardized input for subsequent model analysis. Under the condition of a whole tray of transmission images, this invention adaptively constructs the effective detection area for a single breeding duck egg through a multi-step joint processing method. This effectively eliminates non-target interference factors such as eggshell edge highlights, background noise, and posture differences, forming a stable and uniform analysis input. This reduces the impact of individual differences and edge artifacts on the fertilization status discrimination results, improving the stability of the whole tray detection process and the consistency of batch processing.

[0021] 5. Introducing a "center-edge" radial structure prior enhances the separability of weak embryonic disc differences in feature space during the early stages of incubation. Addressing the issues of weak embryonic disc structure, low contrast, and susceptibility to edge brightness fluctuations during the early incubation stage, this invention utilizes the radial optical density distribution pattern of the embryonic disc region under transmission imaging conditions to radially model and enhance the features. This strengthens the response of the critical central region of the embryonic disc and suppresses interference from non-critical edge regions, thereby improving the distinguishability and discrimination stability between fertilized and infertile eggs in feature space.

[0022] 6. An adaptive feature focusing mechanism for embryonic discs with dynamic spatial coordinate constraints is introduced to improve the spatial discrimination stability of off-center samples. During large-scale detection of the entire embryonic disc, its spatial position in the transmission image may shift due to individual differences. This invention introduces dynamic spatial coordinate constraints to adaptively determine the dynamic focusing center of the embryonic disc based on the feature response distribution and spatially recalibrates the features. This enhances the feature response of the core region of the embryonic disc and suppresses interference from non-critical spatial regions, thereby improving the spatial robustness of off-center embryonic discs and samples with structural differences.

[0023] 7. A deep prediction model integrating multi-scale features, radial structure priors, and dynamic spatial focusing mechanisms improves the overall discrimination stability of whole-disk detection. Based on the aforementioned radial structure prior enhancement and dynamic spatial focusing mechanisms, this invention constructs a prediction model integrating multi-scale features to jointly model blastodisk structure information at different scales. This enables the model to adaptively handle weak blastodisk features, eccentric samples, and individual difference samples while maintaining overall discrimination capabilities, thereby significantly improving the stability and consistency of fertilization status discrimination for whole-disk duck eggs in the early stages of incubation.

[0024] 8. Introducing a reliability assessment and re-inspection mechanism enhances the credibility and controllability of test results in production applications. This invention introduces a result reliability assessment mechanism during the fertilization status determination process. By comprehensively evaluating the determination confidence level and the consistency of multidimensional features, it achieves proactive identification and risk isolation of uncertain samples. When the reliability of the test results is insufficient, re-inspection or supplementary analysis is automatically triggered, thereby preventing low-confidence results from directly entering the production decision-making process and improving the engineering controllability and application safety of whole tray duck egg fertilization status detection in large-scale hatching production.

[0025] Secondly, a system for detecting the fertilization status of duck eggs in the early stages of incubation based on transmission imaging includes:

[0026] The transmission imaging acquisition module is used to place the entire tray of breeding duck eggs in the transmission imaging device and acquire the transmission image of the entire tray through synchronous transmission imaging by an independent light source unit attached to the bottom of each breeding duck egg.

[0027] The image preprocessing and effective detection region construction module is used to preprocess the whole tray of transmission images and segment individual eggs to obtain sub-image regions that correspond one-to-one with a single breeding duck egg, and to construct effective detection regions for the sub-image regions.

[0028] The feature enhancement module is used to extract features from the effective detection area. Based on the radial optical density distribution pattern of the embryonic disc region in the early stage of incubation transmission image, the radial structure prior is introduced to model and enhance the features, resulting in radial structure enhanced features, and radial gradient features are constructed.

[0029] The feature focusing module is used to introduce dynamic spatial coordinate constraints, adaptively determine the dynamic focusing center of the embryonic disc region based on the feature response distribution, and construct a spatial weight mapping based on the dynamic focusing center to spatially recalibrate the enhanced features, thereby obtaining spatially focused enhanced features.

[0030] The prediction module is used to input spatial focusing enhancement features into the fertilization status discrimination model for analysis and output the fertilization status discrimination results of breeding duck eggs;

[0031] The reliability assessment and re-inspection decision module for the discrimination results is used to assess the reliability of the fertilization status discrimination results. It generates a reliability score based on the consistency or confidence between the radial structural enhancement features, spatial focusing enhancement features, and radial gradient features and the fertilization status discrimination results. When the reliability score is lower than the preset threshold, the corresponding duck eggs are marked as pending re-inspection or supplementary analysis is triggered to complete the detection of the fertilization status of duck eggs in the early stage of incubation. Attached Figure Description

[0032] Figure 1 The diagram shows a flowchart of a method for detecting the fertilization status of duck eggs in the early stages of incubation based on transmission imaging.

[0033] Figure 2 The diagram shows a schematic of a transmission imaging device for detecting the fertilization status of eggs in a whole tray.

[0034] Figure 3 The image shows a multi-scale detection model for breeding duck eggs.

[0035] Figure 4 The diagram shows a block diagram of a system for detecting the fertilization status of duck eggs in the early stages of incubation based on transmission imaging.

[0036] Explanation of reference numerals in the attached drawings: 1. Transmitted light source assembly; 2. Light source protective sleeve or covering gasket structure; 3. Detection plate; 4. Imaging acquisition device; 5. Enclosed dark box. Detailed Implementation

[0037] Exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be understood that the embodiments shown and described in the drawings are merely exemplary and are intended to illustrate the principles and spirit of the invention, and are not intended to limit the scope of the invention.

[0038] Example 1:

[0039] like Figure 1 As shown, a method for detecting the fertilization status of duck eggs in the early stages of incubation based on transmission imaging includes the following steps:

[0040] S1: Place the entire tray of breeding duck eggs in the transmission imaging device, and perform synchronous transmission imaging through an independent light source unit attached to the bottom of each breeding duck egg to obtain a transmission image of the entire tray.

[0041] Whole-disc transmission images are used to characterize the early optical distribution features of the blastodisc region inside breeding duck eggs.

[0042] S2: Preprocess the entire transmission image and segment each egg to obtain a sub-image region corresponding to each individual duck egg, and construct an effective detection region for each sub-image region.

[0043] Effective detection region construction processing includes, but is not limited to, one or more combinations of central region constraint, threshold segmentation / edge separation, morphological denoising, geometric constraint fitting, and mask extraction, in order to suppress the influence of eggshell edge highlighting, background noise, and pose differences on subsequent analysis.

[0044] S3: Feature extraction is performed on the effective detection area. Based on the radial optical density distribution pattern of the embryonic disc region in the early stage of incubation transmission image, radial structure prior is introduced to model and enhance the features, resulting in radial structure enhanced features, and radial gradient features are constructed.

[0045] Radial structure prior enhancement includes, but is not limited to, radial weighting, radial partition statistics, radial gradient measurement, central region emphasis and edge suppression, etc., to enhance the response of the embryonic disc-related regions and suppress interference in non-critical regions.

[0046] S4: Adaptively determine the dynamic focus center of the embryonic disc region based on the enhanced feature response distribution, and construct a spatial weight mapping based on the dynamic focus center to spatially recalibrate the enhanced features, thereby obtaining spatial focus enhancement features;

[0047] Spatial weights are mapped to continuous or piecewise continuous decay functions, including but not limited to Gaussian, exponential, piecewise linear, or attention-normalized mappings; recalibration methods include but are not limited to multiplicative modulation, additive bias, gated fusion, or attention-weighted fusion.

[0048] S5: Input the spatial focusing enhancement features into the fertilization status discrimination model for analysis, and output the fertilization status discrimination results of the breeding duck eggs;

[0049] The fertilization status discrimination model is a prediction model that integrates multi-scale feature extraction, radial structure prior enhancement and dynamic spatial focusing mechanism; the prediction model can be specifically implemented as a Conv-egg multi-scale prediction model, which outputs the fertilization or afertility discrimination results of duck eggs of the corresponding species.

[0050] S6: Conduct a reliability assessment of the fertilization status discrimination results. Generate a reliability score based on the consistency or confidence level between the radial structure enhancement features, spatial focusing enhancement features, and radial gradient features and the fertilization status discrimination results. When the reliability score is lower than the preset threshold, mark the corresponding duck eggs as pending re-inspection or trigger supplementary analysis processing to complete the fertilization status detection of duck eggs in the early stage of incubation.

[0051] By marking corresponding duck eggs as awaiting re-inspection or triggering supplementary analysis, the overall reliability, controllability, and engineering applicability of the fertilization status detection results for the entire tray of duck eggs in large-scale hatching production applications are improved. The fertilization status discrimination results and reliability scores are uniformly recorded, managed, and output by the detection software system, and can be used for re-inspection queue management, result traceability, and statistical analysis of low-reliability samples.

[0052] In this embodiment, as Figure 2 As shown, the transmission imaging device includes: a transmission light source assembly 1, a light source protective sleeve or covering gasket structure 2, a detection plate 3, an imaging acquisition device 4, and a sealed dark box 5.

[0053] The transmission light source assembly 1, the light source protective sleeve or covering pad structure 2, the detection tray 3, and the imaging acquisition device 4 are located inside the enclosed dark box 5. The detection tray 3 is used to hold multiple hatching eggs arranged in a tray. The transmission light source assembly 1 is located below the detection tray 3 and contains multiple transmission light source units that correspond one-to-one with the positions of the hatching eggs in the detection tray 3. The light source protective sleeve / covering pad structure 2 is located on the top of each transmission light source unit and is used to fit or partially cover the hatching egg shell to form a relatively closed transmission light channel. The imaging acquisition device 4 is located above the detection tray 3 and is used to acquire transmission image data after being irradiated by the transmission light source and passing through the hatching eggs.

[0054] In this embodiment, step S2 specifically includes the following sub-steps:

[0055] S201: Based on the spatial arrangement characteristics of the hatching eggs in the overall transmission image, the overall transmission image is divided into regions to obtain a single egg sub-image region corresponding to each individual hatching egg. Region division can generate multiple sub-regions based on arrangement rules, annotation information, or detection tray structural parameters; the shape of the sub-regions includes, but is not limited to, rectangles, polygons, or other shapes that can cover the transmission area of ​​the hatching egg.

[0056] S202: For each individual egg image, the central region is cropped, retaining only the main transmission area located in the center of the image to reduce the impact of the edge regions on subsequent processing. The size of the central cropped region can be determined according to a certain proportion of the atomic image size, and the proportion can be set according to the spatial proportion of the egg in the imaging field of view.

[0057] S203: Convert the center-cropped single-egg image into a grayscale image, and perform binarization processing using an adaptive threshold segmentation method to distinguish the hatching egg region from the background region. The binarization processing can be expressed as:

[0058]

[0059] in, Represents pixels grayscale value, and These represent the grayscale sets of the foreground and background, respectively; the determination relationship between the foreground and background can be reversed under different imaging conditions.

[0060] S204: Perform morphological processing on the binarized image to remove non-target areas caused by eggshell edge highlights, background noise, or localized scattered light. Morphological processing includes, but is not limited to, opening operations, the expression of which is:

[0061]

[0062] in, Represents a binary image. Represents a structural element. and These represent erosion and expansion operations, respectively.

[0063] S205: Perform connected component extraction on the morphologically processed binary image, and filter based on the spatial location of the connected components, retaining only those connected components whose centroids lie within a preset center constraint range. The centroid of the connected component ( It can be determined by the following formula:

[0064]

[0065] in, Indicates a connected region. Indicates the area of ​​the region.

[0066] S206: In a connected region that satisfies the central constraint, select the connected region with the largest area as the target region, and perform geometric fitting on it to obtain geometric parameters. Geometric fitting includes, but is not limited to, minimum circumcircle fitting or ellipse fitting: when using minimum circumcircle fitting, the geometric parameters include the coordinates of the circle's center and the radius; when using ellipse fitting, the geometric parameters include the coordinates of the ellipse's center, major axis, minor axis, and rotation angle.

[0067] S207: Construct a mask based on the fitted geometric parameters and apply the mask to the cropped single-egg image to extract the effective detection area containing only the main transmission area of ​​the hatching egg. Optionally, when no target area meeting the conditions is detected, the cropped area is used as a backup effective detection area to ensure process continuity and robustness.

[0068] In this embodiment, step S3 specifically includes the following sub-steps:

[0069] S301: Establish a radial coordinate system using the geometric center of the effective detection area as the radial reference center, and calculate the normalized radial distance for pixels or feature points within the effective detection area. This is used to characterize its positional relationship relative to the center of the egg body:

[0070]

[0071] in,( () represents the center coordinates of the effective detection area. This represents the equivalent radius of the effective detection area.

[0072] S302: Based on the spatial distribution of the main egg structure in the effective detection area, the maximum range of radial modeling is limited to avoid background areas from participating in the modeling and introducing noise interference. The radial modeling range can be determined according to the coverage ratio of the main structure in the transmission image, so that the modeling range covers the central key area and suppresses invalid edge areas.

[0073] S303: Within the defined radial modeling range, the effective detection area is divided into multiple concentric annular regions along the radial direction. Each annular region corresponds to a different radial position interval, which is used to characterize the optical density change structure from the center to the edge.

[0074] S304: Statistically or aggregate the feature responses within each annulus to obtain the corresponding radial response values, and perform radial weighting enhancement on the features based on the response distribution relationship between the annulus. This is used to strengthen the key structural features of the central annulus and relatively suppress the non-key features of the edge annulus, thereby highlighting the radial differences in the blastodisc region.

[0075] S305: Construct radial gradient features based on the difference in radial response between the central and outer annular bands. Used to quantify the degree of structural difference between the center and the peripheral regions:

[0076]

[0077] in, This represents the radial response value of the central ring. This represents the radial response value of the outer ring.

[0078] S306: The radially weighted enhancement features are used as radial structure enhancement features and input into the subsequent spatial focusing processing step.

[0079] In this embodiment, step S4 specifically includes the following sub-steps:

[0080] S401: Based on the feature response distribution after radial structure enhancement, the significant response locations within the effective detection area are analyzed, and the dynamic attention center location for spatial focusing is adaptively estimated to characterize the actual spatial location of the blastodisc region in the current image. The dynamic attention center is not fixed at the geometric center of the image and can change with the sample to adapt to blastodisc eccentricity.

[0081] S402: Focusing on Dynamic Attention Center For reference, the relative normalized radial distance is calculated for each pixel or feature point within the effective detection area. :

[0082]

[0083] in, Represents pixel or feature point coordinates.

[0084] S403: Based on Constructing spatial weight mapping The weight mapping is a continuously decaying function, the form of which includes, but is not limited to, a Gaussian function:

[0085]

[0086] in, Parameters to control the rate of weight decay. Weight mapping is applied to the feature map to achieve spatial dimension recalibration. Recalibration methods include, but are not limited to:

[0087] Recalibrated by element-wise multiplication:

[0088]

[0089] Or additive recalibration:

[0090]

[0091] in, For the input feature map, The focused feature map, This indicates element-wise multiplication. To recalibrate the strength coefficient.

[0092] S404: The radial focusing weight is used to spatially modulate the features, enhance the feature response in the neighborhood of the dynamic attention center, and relatively suppress the features in the peripheral region, so that the feature space forms a focusing distribution with the embryonic disc region as the core, thereby reducing the focusing shift caused by the fixed center / fixed direction assumption.

[0093] S405: The features processed by dynamic spatial focusing are used as spatial focusing enhancement features for subsequent fertilization status determination steps.

[0094] In this embodiment, the fertilization status discrimination model used in step S5 is the Conv-egg multi-scale prediction model. This model, based on a convolutional neural network structure, integrates a radial structure prior feature enhancement mechanism and a feature focusing mechanism based on dynamic spatial coordinate constraints to discriminate the fertilization status of a whole tray of duck eggs in the early stages of incubation. Figure 3 As shown, step S5 specifically includes the following sub-steps:

[0095] S501: Multi-scale Feature Extraction

[0096] The features of a single duck egg, after effective detection region construction, radial structure enhancement, and dynamic spatial focusing, are input into a multi-scale feature extraction network. Local detail features and overall structural features of the germinal disc region are extracted in parallel through convolutional layers or feature branches at different scales, resulting in a multi-scale feature representation. It is used to characterize the weak optical differences and structural changes of the embryonic disc at different spatial resolutions.

[0097] S502: Radial Structure Prior-Guided Multi-Scale Feature Fusion

[0098] In the multi-scale feature fusion process, a feature weighting mechanism based on the "center-edge" radial structure prior is introduced to jointly model features at different scales. Let the radial structure weight mapping after processing in Example 3 be... Then the feature fusion result guided by radial prior can be expressed as:

[0099]

[0100] in, Indicates the first Feature representation at each scale This indicates element-wise multiplication.

[0101] By using the above fusion method, features located in the central region of the blastodisc receive higher weight during the fusion process, thereby strengthening key regional information related to the fertilization state and suppressing the interference of edge and non-key regional features on the discrimination results.

[0102] S503: Feature Recalibration under Dynamic Spatial Coordinate Constraints

[0103] Radial structure prior enhancement features Based on this, the adaptive feature focusing mechanism for embryonic discs based on dynamic spatial coordinate constraints described in Example 4 is embedded into the multi-scale feature fusion process. Let the spatial weight mapping constructed based on the dynamic attention center be... Then the features after spatial focusing and recalibration can be expressed as:

[0104]

[0105] Through the above dynamic spatial recalibration, the model can adaptively focus on the actual position of the blastodisc in the feature space, reducing the feature response shift caused by blastodisc eccentricity or individual differences, and improving the discrimination stability of eccentric samples.

[0106] S504: Discriminant Feature Convergence

[0107] Features after multi-scale fusion, radial structure prior enhancement, and dynamic spatial focusing Feature aggregation is performed to form a highly discriminative feature representation for fertilization status determination. The feature aggregation methods include, but are not limited to, global average pooling, weighted pooling, or attention-guided feature aggregation.

[0108] S505: Fertilization status detection output

[0109] The discriminative features are input into the classification layer or discriminative unit, and the fertilization status of the corresponding duck egg is output. The prediction process can be expressed as follows:

[0110]

[0111] in, This indicates the result of fertilization or azoospermia. and These are the classification layer parameters. The discrimination result also outputs the corresponding discrimination confidence level, which is used for subsequent reliability assessment and re-inspection control.

[0112] In this embodiment, step S6 specifically includes the following sub-steps:

[0113] S601: Based on radial structure enhancement features and dynamic spatial focusing features, calculate the confidence level or consistency index of the fertilization state discrimination result. The consistency index is used to reflect the degree of support of different feature dimensions for the discrimination result.

[0114] S602: Evaluate the consistency between radial gradient features, spatial focusing weight distribution, or discrimination output results. When all feature dimensions show high consistency in fertilization / absence discrimination, the current discrimination result is considered to have high reliability.

[0115] S603: Mapping the decision confidence or consistency index to a reliability score The result is then compared with a preset reliability threshold; when the reliability score is higher than or equal to the threshold, the current fertilization status is confirmed. The reliability score can be obtained by weighting multiple consistency indicators:

[0116]

[0117] in, Indicates the first Consistency metrics corresponding to class features This represents the corresponding weighting coefficient.

[0118] S604: When the reliability score is below the threshold, the corresponding hatching egg will be marked as pending re-inspection, or supplementary analysis processing will be triggered. Supplementary analysis processing includes, but is not limited to, re-acquiring images, delaying the detection time, or introducing other auxiliary discrimination strategies to improve the controllability and engineering usability of the overall output results.

[0119] Example 2:

[0120] Based on Example 1, this embodiment of the invention provides a system for detecting the fertilization status of duck eggs in the early stage of incubation based on transmission imaging. This system can be used to implement the transmission imaging-based method for detecting the fertilization status of duck eggs in the early stage of incubation as described in the previous embodiments. Figure 4 As shown, the system includes:

[0121] The transmission imaging acquisition module is used to place the entire tray of breeding duck eggs in the transmission imaging device and acquire the transmission image of the entire tray through synchronous transmission imaging by an independent light source unit attached to the bottom of each breeding duck egg.

[0122] The image preprocessing and effective detection region construction module is used to preprocess the whole tray of transmission images and segment individual eggs to obtain sub-image regions that correspond one-to-one with a single breeding duck egg, and to construct effective detection regions for the sub-image regions.

[0123] The feature enhancement module is used to extract features from the effective detection area. Based on the radial optical density distribution pattern of the embryonic disc region in the early stage of incubation transmission image, the radial structure prior is introduced to model and enhance the features, resulting in radial structure enhanced features, and radial gradient features are constructed.

[0124] The feature focusing module is used to introduce dynamic spatial coordinate constraints, adaptively determine the dynamic focusing center of the embryonic disc region based on the feature response distribution, and construct a spatial weight mapping based on the dynamic focusing center to spatially recalibrate the enhanced features, thereby obtaining spatially focused enhanced features.

[0125] The prediction module is used to input spatial focusing enhancement features into the fertilization status discrimination model for analysis and output the fertilization status discrimination results of breeding duck eggs;

[0126] The reliability assessment and re-inspection decision module for the discrimination results is used to assess the reliability of the fertilization status discrimination results. It generates a reliability score based on the consistency or confidence between the radial structural enhancement features, spatial focusing enhancement features, and radial gradient features and the fertilization status discrimination results. When the reliability score is lower than the preset threshold, the corresponding duck eggs are marked as pending re-inspection or supplementary analysis is triggered to complete the detection of the fertilization status of duck eggs in the early stage of incubation.

[0127] According to embodiments of the present invention, the present invention also provides an electronic device, a readable storage medium, and a computer program product.

[0128] In an exemplary embodiment, the electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method for detecting the fertilization status of duck eggs in the early stage of incubation based on transmission imaging as described in Embodiment 1 above.

[0129] In an exemplary embodiment, the readable storage medium may be a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the method for detecting the fertilization status of duck eggs in the early stage of incubation based on transmission imaging as described in Embodiment 1 above.

[0130] In an exemplary embodiment, the computer program product includes a computer program that, when executed by a processor, implements the method for detecting the fertilization status of duck eggs in the early stage of incubation based on transmission imaging as described in Embodiment 1 above.

[0131] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0132] In the context of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0133] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0134] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0135] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0136] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. A method for detecting the fertilization status of duck eggs in the early stage of incubation based on transmission imaging, characterized in that, Includes the following steps: The entire tray of breeding duck eggs is placed in a transmission imaging device, and synchronous transmission imaging is performed through an independent light source unit attached to the bottom of each breeding duck egg to obtain a transmission image of the entire tray. The entire transmission image is preprocessed and individual eggs are segmented to obtain sub-image regions that correspond one-to-one with a single breeding duck egg, and effective detection regions are constructed for the sub-image regions. Feature extraction is performed on the effective detection area. Based on the radial optical density distribution pattern of the embryonic disc region in the early stage of incubation transmission image, radial structure prior is introduced to model and enhance the features, resulting in radial structure enhanced features, and radial gradient features are constructed. The dynamic focus center of the embryonic disc region is adaptively determined based on the radial structure enhancement feature response distribution, and a spatial weight mapping is constructed based on the dynamic focus center to spatially recalibrate the radial structure enhancement features, resulting in spatial focus enhancement features, specifically: Based on radial structure enhancement features, significant response locations within the effective detection area are analyzed, and the location of the dynamic attention center used for spatial focusing is adaptively estimated to characterize the actual spatial location of the embryonic disc region in the current image. Using the dynamic attention center as a reference, calculate the relative normalized radial distance of each pixel within the effective detection area; A spatial weight mapping is constructed based on the relative normalized radial distance, and the spatial weight mapping is a continuously decaying function. Spatial focusing enhancement features are obtained by spatially modulating the enhanced features using spatial weight mapping, which enhances the feature response in the neighborhood of the dynamic attention center and relatively suppresses the features in the peripheral region, so that the feature space forms a focused distribution with the embryonic disc region as the core. The spatial focusing enhancement features are input into the fertilization status discrimination model for analysis, and the fertilization status discrimination results of breeding duck eggs are output. The reliability of the fertilization status discrimination results is evaluated. A reliability score is generated based on the consistency or confidence between the radial structural enhancement features, spatial focusing enhancement features, and radial gradient features and the fertilization status discrimination results. When the reliability score is lower than the preset threshold, the corresponding duck eggs are marked as pending re-inspection or supplementary analysis is triggered to complete the detection of the fertilization status of duck eggs in the early stage of incubation.

2. The method for detecting the fertilization status of duck eggs in the early stage of incubation based on transmission imaging according to claim 1, characterized in that, The transmission imaging device includes a transmission light source assembly (1), a light source protective sleeve or a covered rubber pad structure (2), a detection plate (3), an imaging acquisition device (4), and a sealed dark box (5). The transmission light source assembly (1), the light source protective sleeve or covering pad structure (2), the detection plate (3) and the imaging acquisition device (4) are located inside the closed dark box (5); the detection plate (3) is used to carry multiple breeding duck eggs arranged in a whole plate; the transmission light source assembly (1) is set below the detection plate (3), and the transmission light source assembly (1) contains multiple transmission light source units that correspond one-to-one with the positions of the breeding eggs in the detection plate (3); the light source protective sleeve or covering pad structure (2) is set on the top of each transmission light source unit, and is used to fit or partially cover the shell of the breeding duck egg to form a relatively closed transmission light channel; the imaging acquisition device (4) is set above the detection plate (3) and is used to acquire transmission image data after being irradiated by the transmission light source and passing through the breeding egg.

3. The method for detecting the fertilization status of duck eggs in the early stage of incubation based on transmission imaging according to claim 1, characterized in that, The method for preprocessing the entire transmission image and segmenting each egg to obtain a sub-image region corresponding to a single breeding duck egg, and then constructing an effective detection region from the sub-image regions, is as follows: Based on the spatial arrangement characteristics of the hatching eggs, the entire transmission image is divided into regions to obtain sub-image regions that correspond one-to-one with each individual hatching egg. The center region of each sub-image region is cropped to retain the main transmission area located in the center of the image, so as to reduce the influence of the edge region; The single egg image after cropping the central region is converted into a grayscale image and binarized using an adaptive threshold segmentation method to distinguish the hatching egg region from the background region. Morphological processing is performed on the binarized image to remove non-target areas caused by eggshell edge highlights, background noise, or local scattered light; Connected regions are extracted from the morphologically processed binary image, and the connected regions are filtered based on their spatial location, retaining connected regions whose centroids are located within a preset center constraint range; The connected region with the largest area is selected as the target region, and geometric parameters are obtained by geometric fitting of the target region. A mask is constructed based on the fitted geometric parameters, and the mask is applied to the cropped single-egg image of the central region to extract the effective detection area containing the main transmission area of ​​the hatching egg.

4. The method for detecting the fertilization status of duck eggs in the early stage of incubation based on transmission imaging according to claim 1, characterized in that, Feature extraction is performed on the effective detection area. Based on the radial optical density distribution pattern of the embryonic disc region in the early stage of incubation transmission images, a radial structure prior is introduced to model and enhance the features, resulting in radial structure enhanced features. The specific method for constructing radial gradient features is as follows: A radial coordinate system is established using the geometric center of the effective detection area as the radial reference center, and the normalized radial distance is calculated for the pixels or feature points within the effective detection area. The normalized radial distance is used to characterize the positional relationship of the pixels or feature points within the effective detection area relative to the center of the egg body. Based on the spatial distribution of the main structure of the hatching egg in the effective detection area, the maximum range of radial modeling is limited to avoid background areas from participating in the modeling and introducing noise interference. Within the defined radial modeling range, the effective detection area is divided into multiple concentric annular regions along the radial direction. Each annular region corresponds to a different radial position interval, which is used to characterize the optical density change structure from the center to the edge. The feature responses within each annulus are statistically aggregated to obtain the radial response values ​​corresponding to each annulus. Based on the response distribution relationship between the annulus zones, the features are radially weighted and enhanced to obtain radially structure-enhanced features. Radial gradient features are constructed based on the difference in radial response between the central and outer annular zones to quantify the degree of structural difference between the central and peripheral regions.

5. The method for detecting the fertilization status of duck eggs in the early stage of incubation based on transmission imaging according to claim 4, characterized in that, The formula for calculating the normalized radial distance is: in, Represents the normalized radial distance, ( () represents the center coordinates of the effective detection area, () represents the coordinates of pixels or feature points within the effective detection area. Indicates the equivalent radius of the effective detection area; The formula for calculating the radial gradient feature is: in, Represents radial gradient characteristics. This represents the radial response value of the central ring. This represents the radial response value of the outer ring.

6. The method for detecting the fertilization status of duck eggs in the early stage of incubation based on transmission imaging according to claim 1, characterized in that, The formula for calculating the relative normalized radial distance is: in, Represents pixels The relative normalized radial distance, Indicates the coordinates of the dynamic attention center.

7. The method for detecting the fertilization status of duck eggs in the early stage of incubation based on transmission imaging according to claim 1, characterized in that, The method for inputting spatial focusing enhancement features into the fertilization status discrimination model for analysis and outputting the fertilization status discrimination results of breeding duck eggs is as follows: Spatial focusing enhancement features are input into a multi-scale feature extraction network to extract local detail features and overall structural features of the blastodisc region in parallel, resulting in a multi-scale feature representation. Multi-scale feature representation is used to characterize the weak optical differences and structural changes of the embryonic disc at different spatial resolutions; A feature weighting mechanism based on the "center-edge" radial structure prior is introduced to jointly model multi-scale feature representations and obtain fused features; A spatial weight mapping based on a dynamic attention center is set up to perform dynamic spatial recalibration on the fused features, resulting in spatially focused recalibrated features. Feature aggregation is performed on the spatially focused and recalibrated features to obtain a highly discriminative feature representation for fertilization state determination; The highly discriminative feature representation is input into the classification layer, and the fertilization status of the corresponding duck egg is output.

8. The method for detecting the fertilization status of duck eggs in the early stage of incubation based on transmission imaging according to claim 1, characterized in that, The reliability of the fertilization status determination results is assessed. A reliability score is generated based on the consistency or confidence level between the radial structural enhancement features, spatial focusing enhancement features, and radial gradient features and the fertilization status determination results. When the reliability score is lower than a preset threshold, the corresponding duck egg is marked as awaiting re-inspection or supplementary analysis is triggered. The specific method for detecting the fertilization status of duck eggs in the early stage of incubation is as follows: Based on radial structural enhancement features and spatial focusing enhancement features, the confidence or consistency index of fertilization status discrimination results is calculated. Evaluate the consistency between radial gradient characteristics, spatial focusing weight distribution, or discrimination output results; The confidence level or consistency index is mapped to a reliability score and compared with a preset reliability threshold. When the reliability score is higher than or equal to the threshold, the current fertilization status is confirmed. When the reliability score is lower than the threshold, the corresponding eggs are marked as pending re-inspection or supplementary analysis is triggered.

9. A system for detecting the fertilization status of duck eggs in the early stage of incubation, based on transmission imaging according to any one of claims 1-8, characterized in that, include: The transmission imaging acquisition module is used to place the entire tray of breeding duck eggs in the transmission imaging device and acquire the transmission image of the entire tray through synchronous transmission imaging by an independent light source unit attached to the bottom of each breeding duck egg. The image preprocessing and effective detection region construction module is used to preprocess the whole tray of transmission images and segment individual eggs to obtain sub-image regions that correspond one-to-one with a single breeding duck egg, and to construct effective detection regions for the sub-image regions. The feature enhancement module is used to extract features from the effective detection area. Based on the radial optical density distribution pattern of the embryonic disc region in the early stage of incubation transmission image, the radial structure prior is introduced to model and enhance the features, resulting in radial structure enhanced features, and radial gradient features are constructed. The feature focusing module is used to introduce dynamic spatial coordinate constraints, adaptively determine the dynamic focusing center of the embryonic disc region based on the feature response distribution, and construct a spatial weight mapping based on the dynamic focusing center to spatially recalibrate the enhanced features, thereby obtaining spatially focused enhanced features. The prediction module is used to input spatial focusing enhancement features into the fertilization status discrimination model for analysis and output the fertilization status discrimination results of breeding duck eggs; The reliability assessment and re-inspection decision module for the discrimination results is used to assess the reliability of the fertilization status discrimination results. It generates a reliability score based on the consistency or confidence level between the radial structural enhancement features, spatial focusing enhancement features, and radial gradient features and the fertilization status discrimination results. When the reliability score is lower than the preset threshold, the corresponding duck eggs are marked as pending re-inspection or supplementary analysis is triggered to complete the detection of the fertilization status of duck eggs in the early stage of incubation.

Images