Intelligent monitoring and evaluation system for muscovy duck embryonic development and incubation parameter optimization method

Abstract

The invention discloses an intelligent monitoring and evaluating system, method and equipment for muscovy duck embryonic development, and belongs to the technical field of poultry incubation. The system comprises: a database storing standard embryo images associated with specific incubation parameters; the image acquisition module is used for acquiring a real-time embryo image; the image intelligent comparison and diagnosis engine is configured to perform feature comparison on the real-time image and a database standard image and output a quantitative evaluation result and / or a parameter adjustment suggestion; and a result output module. According to the method, quantitative diagnosis of the embryonic development state is realized through the machine learning model, and the incubation parameters can be automatically optimized according to the diagnosis result, so that the problem that the incubation performance is unstable due to the fact that the traditional incubation technology depends on artificial experience and lacks objective standards is effectively solved, and the incubation accuracy and the production efficiency of the muscovy ducks are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to an intelligent monitoring and evaluation system for muscovy duck embryo development and a method for optimizing incubation parameters. BACKGROUND

[0002] As an excellent lean-type poultry breed, muscovy duck plays an important role in modern characteristic poultry industry. Incubation is a core link in muscovy duck production, and the incubation rate and healthy hatch rate directly determine the economic benefits of the breeding farm. However, due to its own structural characteristics (such as high yolk fat content, thick eggshell and waxy oil film), the incubation physiology of muscovy duck eggs is significantly different from that of chicken and ordinary domestic duck, and it is more sensitive to incubation environmental parameters.

[0003] In the existing muscovy duck incubation production, there are two core challenges in technical management: First, the monitoring and evaluation means of embryo development process are primitive and highly dependent on experience. At present, the industry generally uses artificial egg illumination to judge the viability and development process of the embryo. This method mainly relies on the personal experience and subjective judgment of the incubation personnel, and lacks objective and unified evaluation standards. Although some research has tried to establish a static atlas of muscovy duck embryo development to provide visual reference for technical personnel, this atlas has limitations in practical application: it can only provide qualitative and morphological rough comparison, and cannot quantify the speed of development process, the density of blood vessel network and the strength of embryo viability. Different technicians may draw completely different conclusions about the development status of the same embryo, resulting in a lack of precise and consistent basis for adjusting environmental parameters during incubation, which is one of the important reasons for the large fluctuations in incubation results and the frequent occurrence of abnormal situations.

[0004] Second, the adjustment of incubation process parameters is lagging and lacks data-driven. The existing incubation technology mainly relies on preset and fixed temperature and humidity curves (for example, setting different temperatures and humidities in three to four stages). However, in actual production, due to differences in factors such as egg source, duck age, season and climate, the actual development speed of the embryo often deviates from the ideal model. The existing technical solution cannot form an effective closed-loop control between the physiological feedback of the embryo and the parameter setting of the incubator. When egg illumination finds that the embryo is developing slowly or too fast, the technician often has to make parameter adjustments based on vague experience, and the timing, amplitude and direction of the adjustment are all uncertain, making it difficult to achieve precise optimization and often missing the best adjustment window, affecting the final hatching effect. SUMMARY

[0005] In view of the deficiencies in the prior art, the present application aims to provide a technical solution that can overcome the above-mentioned defects. Not only can a more accurate and standard embryonic development reference system be provided, but also the whole process of automatic collection, intelligent analysis, quantitative diagnosis, and automatic generation of parameter optimization suggestions can be intelligentized, thereby upgrading the incubation of Muscovy duck from an empirical skill to a precise and controllable industrial technology.

[0006] The technical solution adopted by the present application to solve its technical problems is: An intelligent monitoring and evaluation system for Muscovy duck embryonic development, comprising: a Muscovy duck embryonic development standard atlas database, which stores standard embryonic images associated with specific incubation parameters at different embryonic ages; an image acquisition module for acquiring real-time embryonic images of Muscovy duck eggs to be tested; an image intelligent comparison and diagnosis engine configured to compare the features of the real-time embryonic images with the standard images of the corresponding embryonic age in the atlas database, and output a quantitative development state evaluation result and / or incubation parameter adjustment suggestion based on the comparison result; a result output module for displaying the development state evaluation result and / or incubation parameter adjustment suggestion.

[0007] As a preferred embodiment, the image intelligent comparison and diagnosis engine is implemented through a machine learning model, which is trained using the images in the standard atlas database as a training set to identify one or more features of the embryo, such as blood vessel density, embryonic morphology, and yolk sac absorption degree, and output a development fit score according to the feature similarity.

[0008] As a preferred embodiment, the system further comprises a feedback learning module; after the user confirms or corrects the evaluation result output by the diagnosis engine, the feedback learning module uses the real-time embryonic image and the confirmed result as new training data to update and optimize the machine learning model.

[0009] As a preferred embodiment, the specific incubation parameters are incubation parameters that do not use water cooling in a small box incubator, specifically including: 1-18 embryonic age temperature 37.9℃, humidity 57%, 19-23 embryonic age temperature 37.6℃, humidity 60%, 24-28 embryonic age temperature 37.3℃, humidity 63%, 29 embryonic age to hatching temperature 37.0℃, humidity 66%.

[0010] As a preferred embodiment, the images in the standard atlas database are constructed through the following standardization process: Imaging in a standardized light box to ensure constant color temperature of the light source and no environmental light interference; When imaging the blastula, a scale is placed beside and all images are processed to the same resolution and color space; Each standard image is associated with its corresponding embryonic age, hatching parameter record and textual description of key developmental features.

[0011] Another technical problem to be solved by the present application is to provide an image recognition-based muscovy duck hatching parameter optimization method, which applies the intelligent monitoring and evaluation system of any one of the above embodiments and comprises the following steps: At multiple key embryonic age points in the hatching process, real-time embryo images of muscovy duck eggs are obtained; Through the image intelligent comparison and diagnosis engine, a quantitative development state evaluation result is obtained; If the development state evaluation result is lower than a preset threshold, the system automatically generates an adjustment suggestion for the current hatching parameter, and the adjustment suggestion includes adjusting at least one of temperature, humidity or ventilation volume by a specific amplitude.

[0012] Another technical problem to be solved by the present application is to provide an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the muscovy duck hatching parameter optimization method as described above when executing the program.

[0013] Another technical problem to be solved by the present application is to provide a computer-readable storage medium having a computer program stored thereon, wherein the program implements the steps of the muscovy duck hatching parameter optimization method as described above when executed by a processor.

[0014] Another technical problem to be solved by the present application is to provide an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps as described above when executing the program.

[0015] The beneficial effects of the present application are: Using the molecular marker identified by the present application, genotyping can be performed on the ear tissue sample of a young boar (such as at weaning) without waiting for sexual maturity (after about 8 months) to perform time-consuming freezing experiments, thereby advancing the selection time by at least 6 months or more; frequent semen collection, expensive freezing reagents and complex laboratory operations are avoided, and the evaluation cost of each candidate boar can be reduced by more than 90%; due to the significant reduction in cost, a larger scale of candidate population can be screened, the selection intensity of excellent genetic individuals is significantly improved, and genetic progress is accelerated.

[0016] The application provides a specific SNP marker group verified by a large population and directly related to freezing resistance. Breeders can select according to genotypes, which completely changes the blindness of relying on indirect indicators and experience selection; the DNA marker is not disturbed by environmental, seasonal, feeding management and other phenotype influencing factors, and the evaluation result is more objective, stable and reliable; the application provides specific marker combinations for different breeds such as Duroc, Large White and crossbreeding, and realizes precise personalized breeding.

[0017] By selecting a boar with high genetic potential for freezing resistance, the quality of frozen semen produced can be ensured from the source, the post-thaw motility can be improved, and the pregnancy rate can be improved; stable and high-quality sperm sources are the premise of large-scale industrialization of frozen semen technology. The application provides technical support for establishing a high-performance frozen semen supply core group, promotes the healthy development of the pig frozen semen market, and can more effectively screen and preserve boar genetic resources with excellent freezing resistance characteristics, which has important significance for the protection and utilization of local pig germplasm resources.

[0018] The application first systematically reveals the important role of key genes such as SOX9, PARN and MAP2K6 in regulating the freezing resistance of pig sperm, and clarifies the potential mechanism of its function through cAMP signaling pathway, MAPK signaling pathway and antioxidant stress pathway; these findings not only have breeding value, but also provide clear molecular targets and theoretical basis for subsequent development of new cryoprotectants and optimization of freezing procedures, for example, designing protective agent components for the antioxidant damage pathway related to the SOX9 gene. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 It is a kind of principle block diagram of intelligent monitoring and evaluation system for embryonic development of Muscovy duck. DETAILED DESCRIPTION

[0020] The application will be further described below in combination with the drawings and specific embodiments, so that those skilled in the art can better understand the application and implement it, but the embodiments are not limiting the application.

[0021] EMBODIMENT REFERENCE Figure 1 As shown in the figure, an intelligent monitoring and evaluation system for embryonic development of Muscovy duck includes: The hardware architecture of the system mainly includes four units: data acquisition unit, calculation and storage unit, incubation execution unit and user interaction unit.

[0022] The data acquisition unit includes: Standardized egg incubator: the internal dimensions of the box are 40 cm x 40 cm x 50 cm, the inner wall is sprayed with matte black paint to eliminate reflections. A 15 cm diameter circular high-transparency light uniformity plate is installed in the center of the bottom of the box, and 36 LED lamp beads with a color temperature of 5500K±100K and a color rendering index (CRI)>95 are arranged below to ensure uniform and non-mottled light during egg illumination. A 5 cm diameter slidable closed shooting hole is provided on the front and both sides of the box, and a shooting hole of the same size is provided at the top to facilitate the industrial camera to capture egg images from multiple angles.

[0023] Embryo image shooting platform: a 50 cm x 50 cm still photography studio, two ring-shaped LED fill light lamps are arranged on the top and both sides, with the same requirement of color temperature 5500K and CRI>95. The shooting platform background is pure white, and a set of transparent scales from 1 mm to 50 mm and 24 color standard cards are always ready for use for later image proportioning and color correction.

[0024] Image sensor: an industrial camera with Sony IMX series sensor is preferred, with a resolution of not less than 12 million pixels, equipped with a 35mm or 50mm fixed focus macro lens to ensure small image distortion and clear details. The camera parameters (aperture, shutter, ISO) should be preset and locked to ensure consistent shooting conditions.

[0025] Computing and storage unit: the core algorithm of the system runs on a high-performance workstation equipped with an Intel i7 or equivalent processor, 32 GB of memory, and an NVIDIA RTX 3080 or equivalent GPU. The server hard disk capacity for storing standard atlas database and user data should not be less than 10 TB, and it is recommended to configure a RAID 1 disk array for data backup.

[0026] Incubation execution unit: the present application is preferably adapted to a small box incubator on the market (capacity 100-500 eggs). Through modification or its own communication interface (such as RS485, Wi-Fi or 4G module), data communication with the system of the present application is realized. The system can send instructions to the incubator to adjust the opening and closing degree of its heater, humidifier and air door.

[0027] User interaction unit: the user interacts with the system through a 10-inch tablet computer or desktop computer. The interactive interface is developed by Web front-end or local application program, which intuitively displays real-time images, development curves, diagnosis reports and adjustment suggestions of embryos.

[0028] Sample selection and pretreatment: Fresh eggs were selected from a healthy, laying-peak population of Muscovy ducks. The egg weight was controlled at 75 ± 5 g. The eggs were collected within 4 hours of laying and disinfected by soaking in a 1‰ concentration of chlorhexidine solution at 40 °C for 3 min. After drying, the blunt end of each egg was uniquely numbered with a pencil. The eggs were then stored in an egg storage room at 18 °C and 75% relative humidity for no more than 5 days.

[0029] Precise incubation and process recording: Before incubation, the eggs were pre-warmed in an environment of 25 °C for 12 hours. The incubation process was performed in a precise small cabinet incubator without water spraying to cool the eggs throughout the process. The core incubation parameters used are shown in the following table, which were optimized by a large number of experiments and are specific environmental conditions for generating standard graphs: Table 1: Incubation parameters for Muscovy duck standard graph construction The temperature and humidity fluctuations of the incubator, egg turning, and any abnormal events were recorded daily.

[0030] The steps for multi-modal image data collection are as follows: Egg phase collection: Starting from the 3rd day of incubation, 20 eggs were randomly selected at a fixed time each day (e.g., 10:00 am) for egg imaging. Each egg was photographed from the top, front, and side, generating a set of sequence images.

[0031] Embryo phase collection: At key embryonic age points (7, 14, 21, 25, 28, 30, 32 days), 3-5 eggs were sampled each time, and the shell was broken in accordance with animal ethics guidelines. The embryo was carefully placed next to the ruler on the shooting platform, and high-definition images were taken from the back, belly, and side. For later-stage embryos, special attention was paid to close-up shots of the yolk sac, allantoic blood vessels, and beak shape.

[0032] Image preprocessing and deep annotation steps are as follows: All images were first corrected for optical distortion. Then, automatic white balance and color correction were performed using the photographed color card to ensure the authenticity of color restoration. Finally, the sizes of the embryo and egg in the image were uniformly scaled to the real physical size (e.g., 0.1 mm per pixel) according to the ruler.

[0033] Annotation work was completed by at least two experienced embryologists back-to-back. The annotation content not only included the embryonic age and description, but more importantly, pixel-level semantic segmentation, i.e., using polygons to outline the embryo body, blood vessel network, eyes, yolk sac, etc. in the image. These segmented regions can be used to calculate accurate quantitative indicators such as blood vessel density index (VDI) and embryo area. All images and annotation information were stored in a structured form (e.g., JSON format) in the database.

[0034] The present application preferably uses a Two-Stream CNN as the core model. One stream is a spatial information stream, taking ResNet-50 as the backbone, responsible for learning the global morphology and texture features of the embryo. The other stream is a blood vessel feature enhancement stream. First, a pre-trained U-Net network is used to segment the blood vessels in the input image, and then the segmented binary blood vessel map is input into a lightweight CNN to learn the distribution patterns and density features of the blood vessels.

[0035] In the later stage of the network, the feature vectors of the two streams are concatenated and then fused and decided by a fully connected layer.

[0036] After training, all images in the standard atlas database are extracted through the two-stream network to obtain high-dimensional feature vectors S_i, which are stored in the feature library.

[0037] For the real-time embryo image T to be tested, its feature vector is also extracted.

[0038] The system calculates the weighted cosine similarity between T and all standard feature vectors S_i corresponding to the embryo age as the final developmental concordance score (DSS).

[0039] Weighted developmental concordance score DSS(T,S)=α CosSim(F_global_T,F_global_S)+β CosSim(F_vessel_T,F_vessel_S) Where: F_global and F_vessel represent the feature vectors extracted by the spatial information stream and the blood vessel feature enhancement stream, respectively, CosSim is the cosine similarity calculation function, and α and β are weight coefficients, and α+β=1. Through grid search optimization, we found that in the critical period of blood vessel development in the middle of incubation, α=0.4, β=0.6; while in the early incubation and pre-hatching period, α=0.7, β=0.3.

[0040] The blood vessel density index (VDI) is calculated as an independent and highly interpretable indicator: VDI=(Number of Vessel Pixels) / (Total Area of Egg Visible Region) The system presets a dynamic diagnostic threshold DSS_threshold, which can be slightly floated according to the embryo age (e.g. 0.80 for 1-10 days, 0.85 for 11-20 days, and 0.82 for 21 days and later).

[0041] The final diagnosis result is divided into four levels: Excellent (DSS ≥ DSS_threshold): normal development, no adjustment needed.

[0042] Good (DSS_threshold-0.1 ≤ DSS < DSS_threshold): slight developmental lag, system prompts attention.

[0043] Medium (DSS_threshold-0.2 ≤ DSS < DSS_threshold-0.1): developmental lag, system generates parameter adjustment suggestions.

[0044] Poor (DSS < DSS_threshold-0.2): severe developmental lag or suspected dead embryo, system alarms and suggests isolation.

[0045] When the diagnosis result is "medium" or "poor", the system starts the parameter optimization module.

[0046] Define the developmental deviation degree D_dev = 1-DSS; To avoid oscillation of single proportional adjustment, the invention introduces incremental PID (proportional-integral-derivative) control idea to calculate the adjustment amount. Take temperature adjustment as an example: Algorithm formula 2: temperature adjustment amount (incremental PID) ΔT(k) = K_p [e(k)-e(k-1)] + K_i e(k) + K_d [e(k)-2e(k-1)+e(k-2)] Where: ΔT(k) is the recommended temperature adjustment amount (℃) this time.

[0047] e(k) = D_dev(k) is the developmental deviation degree calculated this time.

[0048] e(k-1) and e(k-2) are the developmental deviation degrees of the last time and the time before last, respectively.

[0049] K_p, K_i, K_d are the proportional, integral and derivative coefficients of the PID controller, which need to be set through historical data. For example, they can be initially set to K_p = 0.5, K_i = 0.01, K_d = 0.05 through trial and error method.

[0050] This algorithm can give smoother and more accurate adjustment suggestions according to the trend of deviation (derivative term) and cumulative effect (integral term).

[0051] In addition to the PID algorithm, the system also has an expert rule base. For example, the rule base may contain: "If VDI is significantly low but embryo morphology is normal, prioritize the recommendation to appropriately increase humidity to promote vascular expansion"; "If in the late incubation period, the embryo is found to be unable to hatch and the DSS is low, recommend reducing humidity and increasing ventilation". The output of the PID algorithm will be combined with the expert rules to generate an explanatory recommendation for the user, such as: "The embryo's vascular development is lagging by about 15%, according to the current trend, it is recommended to increase the incubation temperature by 0.08°C, and to re-evaluate after 2 days."

[0052] When the user (a senior technician) modifies the system's diagnosis or recommendations, the interaction record (real-time images, system original output, user modification results) is marked as a high-value feedback sample.

[0053] The feedback learning module starts a model fine-tuning process once a week. It randomly samples a small batch of data from all feedback samples, and uses the loss gradient of these samples to make small updates to the original model parameters related to the key comparison and diagnosis engine.

[0054] To prevent the model from experiencing "catastrophic forgetting" on new data, the fine-tuning process will mix feedback samples with a portion of data from the original standard atlas database. This mechanism allows the system to continuously adapt to new genetic backgrounds and seasonal changes, achieving the effect of becoming smarter over time.

[0055] The present application can also be implemented as an electronic device comprising a memory and a processor. The memory (such as an NVMe solid state drive) stores a computer program that contains instructions for implementing all the steps described above, including image preprocessing, dual-stream network model inference, PID optimization calculation, and feedback learning. When the processor (CPU working in conjunction with GPU) executes the program, it can implement all the functions of the present application. Similarly, a computer readable storage medium (such as a read-only optical disc) that stores the program is also within the scope of protection of the present application.

[0056] The above embodiments of the present application are not intended to limit the scope of protection of the present application, and the embodiments of the present application are not limited thereto. According to the above content of the present application, according to the ordinary technical knowledge and common practice in the art, other modifications, replacements or changes to the above structure of the present application can be made without departing from the basic technical idea of the present application, and all such modifications, replacements or changes shall fall within the scope of protection of the present application.

Claims

1. An intelligent monitoring and evaluation system for embryonic development of Muscovy duck, characterized in that, The system comprises: a standard atlas database of muscovy duck embryo development, which stores standard embryo images associated with specific incubation parameters at different embryonic ages; an image acquisition module for obtaining real-time embryo images of muscovy duck eggs to be tested; an image intelligent comparison and diagnosis engine configured to compare the real-time embryo images with standard images of corresponding embryonic ages in the atlas database, and output a quantitative development state evaluation result and / or incubation parameter adjustment suggestion based on the comparison result; a result output module for displaying the development state evaluation result and / or incubation parameter adjustment suggestion.

2. The intelligent monitoring and evaluation system of claim 1, wherein, The image intelligent comparison and diagnosis engine is implemented through a machine learning model trained using images in the standard atlas database as a training set to identify one or more features of embryo blood vessel density, embryo morphology, and yolk sac absorption degree, and output a development fitness score according to feature similarity.

3. The intelligent monitoring and evaluation system of claim 2, wherein, The system further comprises a feedback learning module; after the user confirms or corrects the evaluation result output by the diagnosis engine, the feedback learning module uses the real-time embryo image and the confirmed result as new training data to update and optimize the machine learning model.

4. The intelligent monitoring and evaluation system of claim 1, wherein, The specific incubation parameters are incubation parameters for full-time non-spraying egg cooling in a small box incubator, specifically including: 1-18 embryonic age temperature 37.9℃, humidity 57%, 19-23 embryonic age temperature 37.6℃, humidity 60%, 24-28 embryonic age temperature 37.3℃, humidity 63%, 29 embryonic age to hatching temperature 37.0℃, humidity 66%.

5. The intelligent monitoring and evaluation system of claim 1, wherein, The images in the standard atlas database are constructed through the following standardization process: Perform egg imaging in a standardized light box to ensure constant color temperature of the light source and no environmental light interference; When imaging hatched embryos, place a ruler beside them, and process all images to the same resolution and color space; Each standard image is associated with its corresponding embryonic age, incubation parameter record, and textual description of key development features.

6. An image recognition-based parameter optimization method for incubation of Muscovy duck, characterized in that, The intelligent monitoring and evaluation system of any one of claims 1-5 is applied, and comprises the following steps: At multiple key embryonic age points during the incubation process, obtain real-time embryo images of muscovy duck eggs; Obtain a quantitative development state evaluation result through the image intelligent comparison and diagnosis engine; If the development state evaluation result is below a preset threshold, the system automatically generates an adjustment suggestion for the current incubation parameters, which includes adjusting at least one of temperature, humidity, or ventilation volume by a specific amount.

7. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to implement the steps of the muscovy duck incubation parameter optimization method of claim 6.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the steps of the muscovy duck incubation parameter optimization method of claim 6.

Images