An egg production performance dynamic monitoring and abnormal egg intelligent identification system

Abstract

A kind of egg production performance dynamic monitoring and abnormal egg intelligent identification system and method, including camera, image processing module, target detection module, feature extraction module, strong verification distance algorithm module and monitoring analysis module.Cameras collect egg images in egg production area, target detection module loads target detection model according to YOLO model weight path, and preliminary screening is carried out to egg target by confidence threshold;Feature extraction module extracts egg size, contour, color and surface state features to form verification feature vector;Strong verification distance algorithm module carries out distance calculation to verification feature vector and normal egg feature sample, carries out secondary verification according to minimum feature distance, determines normal egg or abnormal egg;Monitoring analysis module obtains egg production performance change result according to the number of eggs, abnormal egg proportion and egg size distribution in continuous acquisition period, and outputs abnormal early warning, improves the reliability of abnormal egg identification, realizes the dynamic monitoring of egg production performance.

Description

Technical Field

[0001] This invention belongs to the field of intelligent monitoring technology for poultry farming, specifically referring to a dynamic monitoring system for egg production performance and an intelligent identification system for abnormal eggs. Background Technology

[0002] With the development of large-scale, intensive chicken farming, egg production and quality have become crucial monitoring indicators in poultry production. Currently, farms typically rely on manual inspections, scheduled collection, and manual selection to count eggs and screen for abnormal eggs. This approach suffers from high labor intensity, low efficiency, and the significant influence of human experience on results. Especially in large-scale farming environments, with a large number of eggs, it is difficult for humans to continuously and accurately count egg production, size distribution, and the proportion of abnormal eggs over different time periods, making it challenging to detect changes in egg production performance in a timely manner.

[0003] Some existing visual inspection devices can capture images of eggs using cameras and identify them using object detection models. However, in a real chicken coop environment, eggs are easily affected by factors such as changes in lighting, conveyor belt vibration, egg occlusion, surface dirt, and changes in camera angle, causing variations in the confidence level output by the object detection model. If eggs are judged as normal or abnormal solely based on the confidence level of the object detection model, false negatives or missed detections are likely to occur.

[0004] Furthermore, existing visual detection methods for eggs typically focus on the number or location of eggs, failing to adequately utilize features such as actual size, shape, color, and surface condition. Even when eggs are identified, reliable verification of their condition is difficult. Additionally, there is a lack of effective correlation between single egg identification results and the overall egg production performance of the flock. It cannot generate dynamic results on changes in egg production performance based on the number of eggs, the proportion of abnormal eggs, and changes in egg size over a continuous collection period, making it difficult to detect abnormalities in egg production performance in a timely manner. Summary of the Invention

[0005] In order to overcome the shortcomings of the prior art, the present invention provides a dynamic monitoring system for egg production performance and intelligent identification of abnormal eggs, so as to at least partially solve the above-mentioned technical problems.

[0006] The technical solution adopted in this invention is as follows: This invention proposes a dynamic monitoring system for egg production performance and an intelligent identification system for abnormal eggs, including a camera, an image processing module, a target detection module, a feature extraction module, a strong verification distance algorithm module, and a monitoring and analysis module; the camera is used to acquire images of eggs within the egg-laying area and transmit the egg images to the image processing module; the image processing module is used to preprocess the egg images and transmit the processed egg images to the target detection module; the target detection module is configured with a YOLO model weight path, used to load the target detection model according to the YOLO model weight path, and perform target detection on the processed egg images, outputting the egg target bounding box, target confidence score, and egg size parameters; the target detection module is set with a confidence threshold, used to determine targets whose confidence scores reach the confidence threshold as candidate egg targets, and to classify the targets... Targets with a confidence level below the confidence threshold are identified as targets to be verified. The feature extraction module extracts size, contour, color, and surface state features from the bounding boxes corresponding to the candidate egg targets and the targets to be verified, forming a feature vector to be verified. The strong verification distance algorithm module calculates the feature distance between the feature vector to be verified and pre-established normal egg feature samples to obtain the corresponding minimum feature distance. Based on the comparison between the minimum feature distance and the preset distance threshold, the candidate egg targets and the targets to be verified are verified a second time to determine whether the corresponding eggs are normal or abnormal. The monitoring and analysis module generates dynamic monitoring data on egg production, the proportion of abnormal eggs, and the distribution of egg size based on the number of normal eggs, the number of abnormal eggs, and egg size parameters identified within a continuous collection period. Based on the dynamic monitoring data, the changes in egg production performance are obtained.

[0007] Furthermore, the camera is fixedly installed above the egg-laying and conveying area, and the camera's shooting area covers the visual detection area through which the eggs pass. The camera is connected to a supplementary lighting component, which is used to provide stable illumination to the visual detection area to improve the image contrast between the target area of ​​the egg and the background area.

[0008] Furthermore, the image processing module includes an image cropping unit, an image resizing unit, and an image enhancement unit; the image cropping unit is used to extract the visual detection region from the egg image, the image resizing unit is used to resize the visual detection region to an image size that matches the input size of the target detection model, and the image enhancement unit is used to process the brightness, contrast, and noise of the resized image.

[0009] Furthermore, the target detection module includes a model loading unit, a target recognition unit, and a confidence level filtering unit. The model loading unit is used to read the model weight file according to the YOLO model weight path and load the corresponding target detection model. The target recognition unit is used to use the target detection model to perform target recognition on the processed egg image and output the egg target box and the target confidence level corresponding to the egg target box. The confidence level filtering unit is used to compare the target confidence level with the confidence level threshold and determine the target box that reaches the confidence level threshold as the candidate egg target.

[0010] Furthermore, the confidence screening unit is also used to mark target boxes below the confidence threshold as targets to be verified, and transmit the targets to be verified to the feature extraction module. The strong verification distance algorithm module performs secondary verification based on the size features, contour features, color features and surface state features of the targets to be verified, so as to reduce the probability that low confidence targets are directly misjudged as abnormal eggs.

[0011] Furthermore, the feature extraction module is used to obtain the major axis pixel size, minor axis pixel size, target area area, and major-minor axis ratio of the egg based on the egg target bounding box, and convert the pixel size into the actual size according to the pre-set camera calibration parameters to form the egg size feature; the feature extraction module is also used to extract the color mean, color dispersion, texture features, contour integrity, and surface anomaly features based on the pixel information of the egg target area to form a feature vector to be verified.

[0012] Furthermore, the strong verification distance algorithm module establishes a normal egg feature sample library, which includes standard feature vectors corresponding to multiple normal eggs. The strong verification distance algorithm module calculates the weighted feature distance between the feature vector to be verified and each standard feature vector according to the preset feature weights, and determines the minimum value among the weighted feature distances as the minimum feature distance corresponding to the feature vector to be verified. The weighted feature distance is determined by the square root of the sum of the products of the squares of the differences of each feature parameter and the corresponding feature weights.

[0013] Furthermore, the strong verification distance algorithm module compares the minimum feature distance with a preset distance threshold; when the minimum feature distance is less than or equal to the preset distance threshold, the corresponding egg target is identified as a normal egg; when the minimum feature distance is greater than the preset distance threshold, the corresponding egg target is identified as an abnormal egg, and the abnormality type of the abnormal egg is determined based on the size characteristics and surface condition characteristics.

[0014] Furthermore, the monitoring and analysis module is used to determine the abnormality score based on the target confidence level, minimum feature distance, and degree of egg size abnormality, and to classify the abnormal eggs according to the comparison results of the abnormality score and the preset abnormality score threshold; the abnormality types include size abnormality, shape abnormality, color abnormality, surface abnormality, and breakage abnormality.

[0015] Furthermore, the monitoring and analysis module counts the number of eggs laid, the number of normal eggs, the number of abnormal eggs, the proportion of abnormal eggs, the average egg size, and the number of eggs in different size ranges according to the preset collection cycle. It also determines the rate of change in egg production performance based on the number of eggs laid in adjacent collection cycles. When the rate of change in egg production performance is lower than the preset rate of change threshold and the proportion of abnormal eggs reaches the preset abnormal proportion threshold, the monitoring and analysis module outputs an early warning message for abnormal egg production performance.

[0016] Compared with the prior art, the present invention has the following advantages:

[0017] The system continuously captures images of eggs within the egg-laying area using a camera, and uses a target detection module configured with YOLO model weight paths for rapid target recognition. At the same time, a confidence threshold is set to perform preliminary screening of the detection results, enabling the system to automatically locate and count the number of eggs, reducing the workload of manual inspection and counting.

[0018] After target detection, the size, outline, color, and surface condition features of the egg are further extracted. A strong validation distance algorithm is then used to compare the distance between the egg to be verified and a normal egg feature sample, thus forming a two-level identification mechanism that combines target detection and feature distance verification. When the target confidence is low, further verification can still be performed based on the comprehensive characteristics of the egg. At the same time, secondary identification can be performed on eggs with high target confidence but abnormal actual size or surface condition, which helps to improve the reliability of abnormal egg identification.

[0019] By further correlating the identification results of a single egg with the number of eggs laid, the proportion of abnormal eggs, and the egg size distribution within a continuous collection period, dynamic monitoring data on egg production performance is formed. Based on the rate of change in egg production performance and the proportion of abnormal eggs in adjacent collection periods, a joint judgment is made, thereby enabling timely detection of abnormal changes in egg production performance and realizing the expansion from single egg identification to dynamic monitoring of flock egg production performance. Attached Figure Description

[0020] Figure 1 This is a system architecture diagram of the dynamic monitoring of egg production performance and intelligent identification of abnormal eggs proposed in an embodiment of the present invention;

[0021] Figure 2 This is a flowchart illustrating the workflow of the dynamic monitoring system for egg production performance and intelligent identification of abnormal eggs proposed in this embodiment of the invention.

[0022] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0023] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0024] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0025] like Figures 1-2 As shown, this embodiment provides a dynamic monitoring system for egg production performance and an intelligent identification system for abnormal eggs. This system is used for continuous visual acquisition, size detection, status identification, and dynamic analysis of egg production performance in the egg-laying area of ​​a chicken farm. The system includes a camera, an image processing module, a target detection module, a feature extraction module, a strong verification distance algorithm module, and a monitoring and analysis module. Data transmission between these modules occurs via a wired or wireless communication network. The camera is installed above the egg-laying conveyor area through which eggs are concentrated, with its shooting direction facing the egg conveyor surface. It continuously acquires egg images as the eggs enter the visual detection area and transmits the acquired images to the image processing module.

[0026] In this embodiment, a fixed-focal-length industrial camera is used. The camera's installation height is adjusted according to the actual conveyor belt width and the egg's imaging size, ensuring that the length of a single egg in the image meets the pixel requirements for target recognition by the target detection model. The camera's shooting area covers the visual detection area on the conveyor belt. A background surface with a stable color is placed below the visual detection area to reduce the impact of conveyor belt texture, equipment structure, and environmental background on egg contour extraction. A supplementary lighting component is placed next to the camera, providing stable illumination to the visual detection area, ensuring relatively consistent brightness and color conditions in egg images acquired at different times. As eggs enter the visual detection area with the conveyor belt, the camera continuously acquires egg images containing one or more egg targets.

[0027] The image processing module receives the raw egg image captured by the camera and first crops the image according to a pre-defined detection area, removing edges of the conveyor equipment, supports, and background areas irrelevant to egg detection. Subsequently, the image processing module resizes the cropped image according to the input size required by the target detection model, and performs brightness adjustment, contrast enhancement, and noise suppression. When local brightness differences exist in the visual detection area, brightness normalization is used to reduce the impact of illumination changes on the egg's color and surface characteristics. The processed egg image is then transmitted to the target detection module.

[0028] The object detection module includes a model loading unit configured with a YOLO model weight path. This YOLO model weight path points to a pre-trained egg object detection model weight file. For example, the model weight path parameter can be set in the system configuration file. Upon system startup, the model weight path is read and the corresponding object detection model is loaded. When it's necessary to change the model trained for different farming scenarios, different cameras, or different egg samples, the corresponding model can be called by modifying the YOLO model weight path without altering the camera or other hardware structure.

[0029] The object detection module uses the loaded object detection model to identify objects in the processed egg image, and outputs a bounding box, object category, and object confidence score for each detected object in the image. The bounding box determines the position and area occupied by the egg in the image, and the object confidence score indicates the degree to which the object detection model determines that the object belongs to the egg category. The object detection module has a confidence threshold, which can be set to a value between 0.70 and 0.90, with the specific value calibrated based on the training model accuracy, camera resolution, and the surrounding environment. In this embodiment, the confidence threshold can be set to 0.80.

[0030] When the target detection model outputs a target confidence score greater than or equal to 0.80, the target is identified as a candidate egg target. When the target confidence score is lower than 0.80, the target is not directly identified as an abnormal egg, but rather marked as a target to be verified. In this way, the YOLO model primarily performs the function of rapid localization and initial screening of egg targets. For targets with insufficient confidence, further judgment is made through subsequent feature extraction and strong validation distance algorithms. This avoids the direct exclusion of low-confidence targets caused by eggs occluding each other, changes in lighting, dirt on the egg surface, or changes in the camera's shooting angle.

[0031] The feature extraction module receives candidate egg targets and targets to be verified, and extracts egg size features, contour features, color features, and surface condition features based on the corresponding bounding boxes. During egg size feature extraction, the module obtains the egg's major axis pixel size, minor axis pixel size, target region area, and major-minor axis ratio in the image based on the bounding box. The major axis pixel size can be represented as the distance between the two largest pixels on the egg's contour, while the minor axis pixel size is determined based on the contour direction perpendicular to the major axis. These major and minor axis pixel sizes provide a preliminary description of the egg's size and shape.

[0032] To obtain the actual dimensions, the camera is calibrated after system installation. A calibration piece of known length can be placed in the visual inspection area, and the size conversion factor is determined by obtaining the actual length of the calibration piece and the pixel length in the image. The actual dimensions are determined according to the following formula: ,in, For the target actual size, The pixel size corresponding to the target. The size conversion factor is obtained by calibrating the camera. After converting the major and minor axes of the egg separately, the actual major and minor axis dimensions of the egg can be obtained, thus forming a size characteristic that reflects the size of the egg.

[0033] For example, after the camera is calibrated, the size conversion factor within the visual detection area is determined to be 0.05 mm / pi. When the major axis pixel size of a target egg is 100 pi and the minor axis pixel size is 72 pi, the corresponding actual major axis size is approximately 5 mm and the minor axis size is approximately 3.6 mm. In practical applications, the calibration factor can be adjusted according to the camera resolution, installation height, and visual detection area size. The above size data is only used to illustrate the size conversion process and does not limit the actual egg size range.

[0034] In the contour feature extraction process, edge detection is performed on the egg target region to obtain the egg contour. Based on the egg contour, the contour integrity, the ratio of the major and minor axes, and the degree of contour deviation are calculated. Normal eggs usually have a relatively continuous closed contour, while eggs with obvious damage, deformities, or severe defects show significant differences in contour features compared to normal eggs. Therefore, contour integrity and the ratio of the major and minor axes are used as feature parameters in the feature vector to be verified.

[0035] During color feature extraction, the egg target region undergoes color space transformation, and the color mean, color dispersion, and local color variation of the egg target region are calculated. These parameters reflect the overall color of the egg and its local color variations. When there are obvious stains, abnormal color areas, or other surface changes on the egg surface, the difference between the corresponding color features and those of normal egg samples increases.

[0036] During surface state feature extraction, texture analysis and local anomaly detection are performed on the egg target area to obtain texture features, surface spot features, crack features, and surface integrity features. By combining surface state features with size features, contour features, and color features, a feature vector to be verified can be formed for secondary verification using a strong verification distance algorithm.

[0037] In this embodiment, let the feature vector corresponding to the egg to be verified be: ,in, This indicates the egg to be verified. Feature parameters. The normal egg feature sample library pre-stores standard feature vectors corresponding to multiple normal eggs: ,in, This indicates the number of a normal egg sample. The normal egg characteristic sample library can be established by collecting a certain number of egg images under normal production conditions, and abnormal samples such as broken eggs, cracked eggs, severely stained eggs, and obviously deformed eggs are removed through manual verification.

[0038] After receiving the feature vector to be verified, the strong verification distance algorithm module calculates the distance between it and multiple standard feature vectors in the normal egg feature sample database. Considering that different features have varying degrees of importance for identifying abnormal eggs, corresponding feature weights are assigned to different feature parameters. The weighted feature distance is determined according to the following formula: ,in, This indicates that the egg to be verified is related to the first... Weighted feature distance between normal egg samples Indicates the first The feature weights corresponding to the feature parameters. This indicates the egg to be verified. Item feature parameters, Indicates the first The first normal egg sample Item feature parameters.

[0039] To avoid the excessive influence of feature parameters with different dimensions on distance calculation, each feature parameter can be normalized before the distance calculation. For example, egg size, color, texture, and contour features can be mapped to preset numerical ranges to give different feature parameters relatively consistent numerical ranges. For size features, standardization can be performed based on the average size and standard deviation of normal egg samples; for color, texture, and contour features, normalization can be performed based on the maximum and minimum values ​​in the sample library.

[0040] For the same egg to be verified, calculate the weighted feature distance between it and multiple standard feature vectors in the normal egg feature sample library, and obtain: ,in, This represents the minimum distance between the feature space of the egg to be verified and the normal egg sample. This indicates the number of normal egg feature samples. The strong validation distance algorithm module sets a preset distance threshold. and will and Comparison. When When the overall characteristics of the egg to be tested are considered to be similar to those of at least one normal egg sample, the egg is identified as a normal egg; when At that time, it was determined that the egg deviated significantly from the characteristic space of a normal egg, and the egg was identified as an abnormal egg.

[0041] In this way, even if an egg has a low YOLO target confidence score, as long as its comprehensive features such as size, outline, color, and surface condition are close to those of a normal egg sample, it can still be identified as a normal egg using the strong validation distance algorithm. Correspondingly, for eggs with a high YOLO model confidence score but whose size, shape, or surface condition deviates significantly from those of normal samples, they can also be identified a second time using the strong validation distance algorithm, thus forming a dual identification mechanism based on target detection and feature distance verification.

[0042] Furthermore, in this embodiment, the monitoring and analysis module determines the anomaly score based on the target confidence level, minimum feature distance, and the degree of egg size anomaly. The following anomaly scoring model can be used: ,in, For abnormal scoring, The target confidence score output by the YOLO model. This is the minimum feature distance obtained by the strong validation distance algorithm. For the preset distance threshold, The degree of abnormality in egg size, , and These are the corresponding weighting coefficients.

[0043] Among them, the degree of abnormal egg size The degree of size anomaly can be determined based on the deviation between the current egg's actual size and the normal egg size range. For example, establishing normal egg size ranges for the long axis, short axis, and the ratio of the long axis to the short axis, when one or more size parameters of the current egg exceed the corresponding normal range, the degree of size anomaly is determined based on the extent of the deviation. By fusing target confidence, feature distance, and the degree of size anomaly, it is possible to avoid judging the egg's condition based on only a single parameter.

[0044] The monitoring and analysis module classifies anomaly scores based on preset anomaly scoring thresholds. For example, when the anomaly score is below the first anomaly scoring threshold, the egg is identified as a normal egg; when the anomaly score is between the first and second anomaly scoring thresholds, the egg is identified as a suspected anomaly egg; and when the anomaly score reaches the second anomaly scoring threshold, the egg is identified as a high-risk anomaly egg. For suspected anomaly eggs, the corresponding egg image and recognition result can be retained for subsequent manual verification.

[0045] In identifying abnormal egg types, the monitoring and analysis module determines the anomaly type based on size, contour, color, and surface condition characteristics. An egg is marked as having a size anomaly when its major or minor axis dimensions significantly deviate from the normal range; a shape anomaly when the ratio of the major and minor axes or the integrity of the contour significantly deviates from the normal range; a color anomaly when the color characteristics differ significantly from normal samples; a surface anomaly when there are obvious abnormalities in texture, spots, or local surface features; and a broken or partially chipped egg contour. An egg can correspond to one or more anomaly characteristics simultaneously.

[0046] To achieve dynamic monitoring of egg production performance, the system statistically analyzes the identification results at preset time intervals. For example, using 10 minutes, 30 minutes, 1 hour, or 24 hours as a statistical period, it counts the number of valid eggs, normal eggs, and abnormal eggs detected by the camera within the corresponding period. Each collection cycle records the number of valid eggs as follows: The normal number of eggs is The number of abnormal eggs is The proportion of abnormal eggs can be determined by the following formula: Simultaneously, the average egg size within the corresponding period is calculated based on the actual size of each egg. For example, the average major axis size and average minor axis size can be calculated separately, or the number of eggs in different size ranges can be further counted to obtain the egg size distribution.

[0047] The monitoring and analysis module obtains the time series of egg production based on multiple consecutive data collection cycles: ,in, Indicates the first The effective egg production count within each collection cycle. The rate of change in egg production performance is obtained by comparing the egg production counts of adjacent collection cycles. ,when A negative value indicates that the number of eggs laid in the current collection cycle has decreased compared to the previous collection cycle; when The rate of change remains consistently below the preset threshold, and the proportion of abnormal eggs... When the egg production rate remains above the preset abnormality threshold, the monitoring and analysis module determines that there is an abnormal change in the current egg production performance and outputs an early warning message for abnormal egg production performance.

[0048] To avoid false alarms caused by occasional fluctuations in a single data collection cycle, this embodiment can use multiple consecutive data collection cycles for joint judgment. For example, three consecutive data collection cycles can be set as the judgment window. An abnormal egg production performance warning is only generated when the change rate of egg production performance corresponding to the three data collection cycles is lower than a preset change rate threshold, and the proportion of abnormal eggs in at least two of the data collection cycles exceeds a preset abnormal proportion threshold. The warning information may include the warning time, current egg production, egg production change rate, proportion of abnormal eggs, number of abnormal eggs, and main types of abnormal eggs.

[0049] Furthermore, the monitoring and analysis module can also classify and store monitoring data according to chicken coop, egg-laying area, or flock number. When multiple cameras correspond to different egg-laying areas, the eggs collected by each camera are numbered and counted separately, and an independent egg-laying performance time series is established according to the corresponding area, thereby enabling comparison of egg production quantity, abnormal egg ratio, and egg size distribution between different egg-laying areas.

[0050] To prevent the same egg from being counted repeatedly in consecutive video frames during continuous egg movement, the target detection module can establish target associations based on the positional changes of the egg's target bounding box in adjacent image frames. When the center distance between the target bounding boxes of two egg targets in adjacent frames is less than a preset distance and the overlap of the target bounding boxes reaches a preset condition, the two targets are associated with the same egg target, and a unique identification number is assigned to the egg target. When the egg target leaves the visual detection area, the corresponding egg is determined as a valid detection target and is counted only once in the statistics module. This reduces the statistical error in egg production caused by repeated counting of the same egg during continuous video acquisition.

[0051] In the actual operation, eggs enter the visual detection area corresponding to the camera via a conveyor belt, and the camera continuously acquires images of the eggs. The image processing module performs image cropping, resizing, and image enhancement. The target detection module loads the target detection model according to the YOLO model weight path and outputs the egg target bounding box and target confidence score. Egg targets that reach the confidence threshold are used as candidate egg targets, while those that do not reach the confidence threshold are used as targets to be verified. Subsequently, the feature extraction module extracts size, contour, color, and surface state features from the candidate egg targets and targets to be verified. The strong verification distance algorithm module calculates the distance between the extracted features and the normal egg feature sample library and obtains the minimum feature distance. The monitoring and analysis module combines the minimum feature distance, target confidence score, and degree of size abnormality to determine the egg status. Finally, the number of normal eggs, the number of abnormal eggs, the proportion of abnormal eggs, and the egg size distribution in each collection cycle are statistically analyzed to form dynamic changes in egg production performance.

[0052] For example, a camera is installed in a chicken coop to continuously capture data on the egg collection and conveying area. The system sets the confidence threshold to 0.80 and establishes a normal egg feature sample library containing the size, outline, color, and surface condition characteristics of normal eggs. When an egg with a YOLO confidence score of 0.76 is detected, it is not directly identified as an abnormal egg because it is below 0.80. Instead, it enters the strong verification distance algorithm module. If the minimum feature distance obtained after feature extraction is 0.32, and the preset distance threshold is 0.50, then because 0.32 is less than 0.50, the egg is still identified as a normal egg. Conversely, when another egg has a YOLO confidence score of 0.91, which is higher than 0.80, but the minimum feature distance obtained after feature distance calculation is 0.78, which is greater than the preset distance threshold of 0.50, then the egg is further identified as an abnormal egg, and its specific abnormality type is determined based on its size, outline, and surface condition characteristics. The above values ​​are only used to illustrate the system's judgment process and do not constitute a limitation on the specific values ​​of each threshold.

[0053] Through the aforementioned structure and processing method, the system in this embodiment does not merely count eggs based on camera images. Instead, it utilizes the YOLO model for rapid egg detection, performs initial screening using a confidence threshold, and then generates a multi-dimensional feature vector based on size, contour, color, and surface condition. A strong validation distance algorithm is then used for secondary validation against normal egg feature samples, ensuring that both egg size and condition are included in the identification process. Furthermore, the system performs statistical analysis and change analysis on egg production numbers, abnormal egg ratios, and egg size distribution across multiple consecutive acquisition cycles, thereby generating dynamic monitoring results for flock egg production performance.

[0054] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0055] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A system for dynamic monitoring of egg production performance and intelligent identification of abnormal eggs, characterized in that, It includes a camera, an image processing module, a target detection module, a feature extraction module, a strong verification distance algorithm module, and a monitoring and analysis module; The camera is used to capture images of eggs within the egg-laying area and transmit the egg images to the image processing module; The image processing module is used to preprocess the egg image and transmit the processed egg image to the target detection module; The target detection module is configured with a YOLO model weight path, which is used to load the target detection model according to the YOLO model weight path, and perform target detection on the processed egg image, outputting the egg target box, target confidence score and egg size parameters; The target detection module is equipped with a confidence threshold, which is used to identify targets whose confidence level reaches the confidence threshold as candidate egg targets, and targets whose confidence level is lower than the confidence threshold as targets to be verified. The feature extraction module is used to extract size features, contour features, color features, and surface state features based on the target boxes corresponding to the candidate egg target and the target to be verified, and to form a feature vector to be verified. The strong verification distance algorithm module is used to calculate the feature distance between the feature vector to be verified and the pre-established normal egg feature samples to obtain the corresponding minimum feature distance, and to perform secondary verification on the candidate egg target and the target to be verified based on the comparison result of the minimum feature distance and the preset distance threshold, so as to determine whether the corresponding egg is a normal egg or an abnormal egg. The monitoring and analysis module is used to generate dynamic monitoring data on egg production quantity, abnormal egg ratio, and egg size distribution based on the number of normal eggs, abnormal eggs, and egg size parameters identified within a continuous collection period, and to obtain the results of changes in egg production performance based on the dynamic monitoring data.

2. The system for dynamic monitoring of egg production performance and intelligent identification of abnormal eggs according to claim 1, characterized in that, The camera is fixedly installed above the egg-laying and conveying area. The camera's shooting area covers the visual detection area through which the eggs pass. The camera is connected to a supplementary lighting component, which is used to provide stable illumination to the visual detection area to improve the image contrast between the target area of ​​the egg and the background area.

3. The system for dynamic monitoring of egg production performance and intelligent identification of abnormal eggs according to claim 1, characterized in that, The image processing module includes an image cropping unit, an image resizing unit, and an image enhancement unit. The image cropping unit is used to extract a visual detection region from the egg image. The image resizing unit is used to adjust the visual detection region to an image size that matches the input size of the target detection model. The image enhancement unit is used to process the brightness, contrast, and noise of the resized image.

4. The system for dynamic monitoring of egg production performance and intelligent identification of abnormal eggs according to claim 1, characterized in that, The target detection module includes a model loading unit, a target recognition unit, and a confidence filtering unit; The model loading unit is used to read the model weight file according to the YOLO model weight path and load the corresponding target detection model; The target recognition unit is used to perform target recognition on the processed egg image using the target detection model, and outputs the egg target box and the target confidence score corresponding to the egg target box; The confidence screening unit is used to compare the target confidence with the confidence threshold, and to determine the target box that reaches the confidence threshold as a candidate egg target.

5. The system for dynamic monitoring of egg production performance and intelligent identification of abnormal eggs according to claim 4, characterized in that, The confidence screening unit is also used to mark target boxes below the confidence threshold as targets to be verified, and transmit the targets to be verified to the feature extraction module. The strong verification distance algorithm module performs secondary verification based on the size features, contour features, color features and surface state features corresponding to the targets to be verified, so as to reduce the probability that low confidence targets are directly misjudged as abnormal eggs.

6. The system for dynamic monitoring of egg production performance and intelligent identification of abnormal eggs according to claim 1, characterized in that, The feature extraction module is used to obtain the major axis pixel size, minor axis pixel size, target area area and major-minor axis ratio of the egg based on the egg target box, and convert the pixel size into the actual size according to the preset camera calibration parameters to form the egg size feature; The feature extraction module is also used to extract the mean color, color dispersion, texture features, contour integrity, and surface anomaly features based on the pixel information of the egg target area, so as to form the feature vector to be verified.

7. The system for dynamic monitoring of egg production performance and intelligent identification of abnormal eggs according to claim 1, characterized in that, The strong verification distance algorithm module establishes a normal egg feature sample library, which includes standard feature vectors corresponding to multiple normal eggs. The strong verification distance algorithm module calculates the weighted feature distance between the feature vector to be verified and each of the standard feature vectors according to preset feature weights, and determines the minimum value among the weighted feature distances as the minimum feature distance corresponding to the feature vector to be verified. The weighted feature distance is determined by the square root of the sum of the squares of the differences between each feature parameter and the corresponding feature weight.

8. The system for dynamic monitoring of egg production performance and intelligent identification of abnormal eggs according to claim 7, characterized in that, The strong verification distance algorithm module compares the minimum feature distance with a preset distance threshold; when the minimum feature distance is less than or equal to the preset distance threshold, the corresponding egg target is identified as a normal egg; when the minimum feature distance is greater than the preset distance threshold, the corresponding egg target is identified as an abnormal egg, and the abnormality type of the abnormal egg is determined according to the size feature and the surface state feature.

9. The system for dynamic monitoring of egg production performance and intelligent identification of abnormal eggs according to claim 8, characterized in that, The monitoring and analysis module is used to determine an anomaly score based on the target confidence level, the minimum feature distance, and the degree of egg size abnormality, and to classify abnormal eggs according to the comparison result of the anomaly score and the preset anomaly score threshold; the anomaly types include size abnormality, shape abnormality, color abnormality, surface abnormality, and breakage abnormality.

10. The system for dynamic monitoring of egg production performance and intelligent identification of abnormal eggs according to claim 9, characterized in that, The monitoring and analysis module counts the number of eggs laid, the number of normal eggs, the number of abnormal eggs, the proportion of abnormal eggs, the average egg size, and the number of eggs in different size ranges according to a preset collection cycle, and determines the rate of change in egg production performance based on the number of eggs laid in adjacent collection cycles. The monitoring and analysis module outputs an early warning message for abnormal egg production performance when the rate of change in egg production performance is lower than a preset rate of change threshold and the proportion of abnormal eggs reaches a preset abnormal proportion threshold.

Images