Body surface parameter monitoring method and system for meat duck breeding

Abstract

The invention relates to the technical field of parameter monitoring, in particular to a body surface parameter monitoring method and system for meat duck breeding, and the method comprises the steps: collecting the original image data of a meat duck epidermis region, and synchronously obtaining an environment illumination intensity parameter; performing dynamic illumination compensation processing on the original image data to generate a standard illumination epidermis image; performing feather-epidermis region segmentation on the standard illumination epidermis image to generate a feather distribution mask and an epidermis exposure region coordinate set; extracting multi-dimensional physical sign parameters based on the epidermis exposure area coordinate set, wherein the multi-dimensional physical sign parameters comprise feather gap epidermis temperature, a feather coverage density index and an epidermis texture roughness coefficient; inputting the multi-dimensional physical sign parameters into a pre-trained growth state evaluation model, and outputting a physical sign anomaly probability value and a corresponding anomaly type code; according to the invention, feather dynamic shielding and ambient light interference are effectively overcome, a real-time linkage mechanism of physical sign parameters and feeding strategies is established, and the intelligent level and economic benefits of meat duck breeding are significantly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of parameter monitoring, and particularly to a method and system for monitoring body surface parameters for meat duck breeding. Background Art

[0002] During the breeding process of meat ducks, the real-time monitoring of body surface parameters is of great significance for evaluating growth status, preventing diseases, and optimizing feeding strategies. However, existing technologies mostly rely on manual visual inspection or single infrared temperature measurement, and there are significant technical bottlenecks: Firstly, the dynamic occlusion of feathers makes it difficult to extract epidermal features. Traditional image segmentation algorithms are difficult to accurately distinguish the epidermal regions at the feather gaps, resulting in measurement errors of physical sign parameters. Secondly, existing monitoring methods are mostly limited to temperature parameters, lacking comprehensive analysis of multi-dimensional physical signs such as feather density distribution and epidermal texture changes, and unable to fully reflect the health status of meat ducks. In addition, the epidermal specular reflection interference in a dynamic light environment is serious, and traditional image enhancement algorithms are prone to losing feather edge details, affecting the reliability of monitoring results.

[0003] In the prior art, some studies have tried to improve the monitoring accuracy through multi-spectral imaging or deep learning algorithms, but there are still the following deficiencies: One is that the multi-spectral data fusion method does not consider the dynamic changes of environmental light, resulting in unstable image quality; the second is that the feather-epidermis segmentation algorithm has insufficient processing ability for fine gaps and is prone to edge breaks; the third is that the extraction of physical sign parameters is disconnected from the growth model, and a real-time linkage mechanism between physical sign abnormalities and feeding strategies has not been established. Therefore, there is an urgent need for an innovative solution that can overcome the above technical bottlenecks and achieve accurate monitoring and intelligent control of multi-dimensional physical signs to meet the high-efficiency and intelligent requirements of modern meat duck breeding. Summary of the Invention

[0004] The present invention provides a method for monitoring body surface parameters for meat duck breeding.

[0005] A method for monitoring body surface parameters for meat duck breeding includes the following steps:

[0006] S1, collecting original image data of the epidermal region of meat ducks through a multi-spectral imaging device, and synchronously obtaining environmental light intensity parameters;

[0007] S2, performing dynamic light compensation processing on the original image data to generate a standard light epidermal image, and the dynamic light compensation processing includes a reflectance correction sub-module based on the environmental light intensity parameters;

[0008] S3, performing feather-epidermal region segmentation on the standard light epidermal image to generate a feather distribution mask and a coordinate set of the epidermal exposed region;

[0009] S4. Extract multi-dimensional physical sign parameters based on the epidermal exposure area coordinate set. The multi-dimensional physical signs include the epidermal temperature in the feather gap, the feather coverage density index, and the epidermal texture roughness coefficient.

[0010] S5. Input the multi-dimensional physical sign parameters into the pre-trained growth status evaluation model to output the physical sign anomaly probability value and the corresponding anomaly type code.

[0011] Optionally, S1 includes:

[0012] S11. Trigger the synchronous acquisition module of the multi-spectral imaging device. The multi-spectral imaging device includes a spectroprism group, a narrowband filter array, and an embedded four-channel photodiode array.

[0013] S12. Perform spectral splitting on the target meat duck to generate four-channel raw spectral data.

[0014] S13. Generate the raw image data of multi-spectral fusion.

[0015] S14. Synchronously acquire the ambient light intensity parameter.

[0016] Optionally, S2 includes:

[0017] S21. Obtain the raw image data of multi-spectral fusion and the ambient light intensity parameter.

[0018] S22. Construct a reflectivity correction sub-module to calculate the pixel-level dynamic compensation coefficient.

[0019] S23. Generate a standard illumination epidermal image based on the pixel-level dynamic compensation coefficient.

[0020] Optionally, S3 includes:

[0021] S31. Construct an improved U-Net++ segmentation network. Its input is the standard illumination epidermal image, and the output is the feather area probability map and the epidermal area probability map.

[0022] S32. Optimize the probability map through the edge repair sub-network to generate the corrected feather edge area.

[0023] S33. Generate a feather distribution mask based on the corrected feather edge area.

[0024] S34. Extract the epidermal exposure area coordinate set.

[0025] Optionally, S4 includes:

[0026] S41. Extract the epidermal temperature in the feather gap based on the epidermal exposure area coordinate set.

[0027] S42. Calculate the feather coverage density index based on the feather distribution mask;

[0028] S43. Extract the epidermal texture roughness coefficient based on the extracted epidermal exposure area coordinate set.

[0029] Optionally, the S5 includes:

[0030] S51. Discretize the input multi-dimensional physical sign parameters to construct a discrete feature vector;

[0031] S52. Input the discrete feature vector into the pre-trained Bayesian network and calculate the four types of abnormal probabilities through the conditional probability table;

[0032] S53. Determine the dominant abnormal type according to the four types of abnormal probabilities and their clinical weights, and generate an abnormal type code.

[0033] Optionally, the S51 includes:

[0034] S511. Discretize the input multi-dimensional physical sign parameters;

[0035] S512. Based on the discretized epidermal temperature in the feather gap, the feather coverage density index, and the epidermal texture roughness coefficient, construct a discretized feature vector.

[0036] Optionally, the S52 includes:

[0037] S521. Input the discretized feature vector into the pre-trained Bayesian network and calculate the four types of abnormal probabilities through the conditional probability table;

[0038] S522. Calculate the total probability of physical sign abnormalities and quantify the probability of at least one abnormality occurring.

[0039] Optionally, the S53 includes:

[0040] S531. Determine the dominant abnormal type according to the four types of abnormal probabilities and their clinical weights;

[0041] S532. Generate an abnormal type code based on the dominant abnormal type.

[0042] A body surface parameter monitoring system for meat duck breeding, used to implement the above-mentioned method for monitoring body surface parameters of meat ducks, includes the following modules:

[0043] Multi-spectral imaging device module: Collect the original image data of the epidermal area of the meat duck and synchronously obtain the environmental light intensity parameter;

[0044] Dynamic illumination compensation module: performs dynamic illumination compensation processing on the original image data to generate a standard illumination epidermal image, and the dynamic illumination compensation processing includes a reflectivity correction sub-module based on the ambient illumination intensity parameter;

[0045] Feather-epidermis region segmentation module: performs feather-epidermis region segmentation on the standard illumination epidermal image to generate a feather distribution mask and a set of coordinates of the epidermal exposure region;

[0046] Multi-dimensional sign extraction module: extracts multi-dimensional sign parameters based on the set of coordinates of the epidermal exposure region, including the epidermal temperature in the feather gap, the feather coverage density index, and the epidermal texture roughness coefficient;

[0047] Growth state evaluation model: inputs the multi-dimensional sign parameters into a pre-trained growth state evaluation model, and outputs a sign abnormality probability value and the corresponding abnormality type code.

[0048] Advantages of the present invention:

[0049] The present invention provides a method and system for monitoring body surface parameters for meat duck breeding, which has the following significant advantages: First, through multi-spectral imaging and dynamic illumination compensation technologies, it effectively overcomes the dynamic occlusion of feathers and environmental illumination interference, and improves the positioning accuracy of the epidermal exposure region; at the same time, the improved U-Net++ segmentation network and edge repair sub-network solve the problem of feather edge fracture, greatly improve the feather gap, and improve the detection accuracy of the epidermal exposure region, laying a solid foundation for the accurate extraction of sign parameters.

[0050] The present invention constructs for the first time a multi-dimensional sign parameter system including the epidermal temperature in the feather gap, the feather coverage density index, and the epidermal texture roughness coefficient, realizes the collaborative analysis of sign parameters and the comprehensive evaluation of abnormal states; in addition, through the Bayesian network and the clinical weight decision-making mechanism, a real-time linkage mechanism between sign parameters and feeding strategies is established, significantly improving the intelligent level and economic benefits of meat duck breeding. Description of the drawings

[0051] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings in the following description are only those of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0052] Figure 1 It is a flowchart of the method according to the embodiment of the present invention;

[0053] Figure 2 It is a system module diagram according to the embodiment of the present invention. Detailed implementation manners

[0054] The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it should be noted here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments. For some well-known technologies, those skilled in the art can also adopt other alternative methods for implementation; moreover, the accompanying drawings are only for more specifically describing the embodiments, and are not intended to specifically limit the present invention.

[0055] It should be noted that in the specification, references to "one embodiment", "an embodiment", "exemplary embodiments", "some embodiments", etc. indicate that the described embodiments may include specific features, structures, or characteristics, but not necessarily every embodiment includes such specific features, structures, or characteristics. Additionally, when describing specific features, structures, or characteristics in connection with an embodiment, implementing such features, structures, or characteristics in connection with other embodiments (whether explicitly described or not) should be within the knowledge of those skilled in the relevant art.

[0056] Generally, terms can be understood, at least in part, from their use in context. For example, at least in part depending on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in a singular sense, or can be used to describe a combination of features, structures, or characteristics in a plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey a set of exclusive factors, but rather, at least in part depending on the context, can allow for the existence of other factors that may not be explicitly described.

[0057] As Figure 1 shown, a method for monitoring body surface parameters for meat duck breeding includes the following steps:

[0058] S1, acquiring original image data of the epidermis region of meat ducks through a multispectral imaging device, and synchronously obtaining environmental light intensity parameters;

[0059] S2, performing dynamic light compensation processing on the original image data to generate a standard light epidermis image, and the dynamic light compensation processing includes a reflectivity correction sub-module based on the environmental light intensity parameters;

[0060] S3, performing feather-epidermis region segmentation on the standard light epidermis image to generate a feather distribution mask and a set of coordinates of the epidermis exposure region;

[0061] S4, extracting multi-dimensional physical sign parameters based on the set of coordinates of the epidermis exposure region, and the multi-dimensional physical signs include the epidermis temperature between feathers, the feather coverage density index, and the epidermis texture roughness coefficient;

[0062] S5, inputting the multi-dimensional physical sign parameters into a pre-trained growth state evaluation model, and outputting a physical sign abnormality probability value and the corresponding abnormality type code.

[0063] S1 includes:

[0064] S11, trigger the synchronous acquisition module of the multispectral imaging device, where the multispectral imaging device includes a beam splitting prism group, a narrowband filter array, and an embedded four-channel photodiode array;

[0065] S12, perform beam splitting on the target meat duck to generate four-channel original spectral data:

[0066] Channel 1 (visible light band): 500 - 600 nm, for capturing epidermal texture;

[0067] Channel 2 (near-infrared band): 850 - 900 nm, for penetrating through the feather gaps;

[0068] Channel 3 (short-wave infrared): 940 - 960 nm, for reconstructing the epidermal temperature field;

[0069] Channel 4 (ultraviolet band): 320 - 380 nm, for detecting epidermal lesions;

[0070] S13, generate the original image data of multispectral fusion, expressed as:

[0071]

[0072] where, I k (x,y) represents the pixel intensity value of the k-th channel, ω k is the channel weight coefficient (ω1 = 0.3, ω2 = 0.4, ω3 = 0.2, ω4 = 0.1), and δ k is the filter transmittance compensation factor (δ1 = 1.2, δ2 = 1.5, δ3 = 1.0, δ4 = 0.8);

[0073] S14, synchronously acquire the environmental light intensity parameter, expressed as:

[0074] E env = w1·E vis + w2·E nir + w3·E uv + w4·E wide ;

[0075] where, E vis , E nir , E uv respectively represent the illuminance measurement values in the visible light (400 - 700 nm), near-infrared (700 - 1100 nm), and ultraviolet (300 - 400 nm) bands, and E wide is the full-spectrum (300 - 1100 nm) illuminance value, and the weight coefficients satisfy w1 + w2 + w3 + w4 = 1, preferably w1 = 0.5, w2 = 0.3, w3 = 0.1, w4 = 0.1.

[0076] S2 includes:

[0077] S21, obtaining the original image data of multispectral fusion and the environmental light intensity parameters;

[0078] S22, constructing a reflectance correction sub-module and calculating the pixel-level dynamic compensation coefficient, expressed as:

[0079]

[0080] where E ref is the standard reference light intensity, set to 500 lux (corresponding to the overcast sky diffuse light condition), γ is the band reflectance gain coefficient, with a value range of 0.8 - 1.2 (taking 1.0 in the visible light band and 1.15 in the near-infrared band), = 0.1 is the light intensity smoothing factor, β = 0.2 is the image intensity distribution adjustment weight, and μ and σ are the mean and standard deviation of the original image data within a 10×10 neighborhood respectively;

[0081] S23, generating a standard illumination epidermis image based on the pixel-level dynamic compensation coefficient, expressed as:

[0082] I std (x,y) = C comp (x,y) · I raw (x,y) · δ k + T cal ;

[0083] where δ k is the filter transmittance compensation factor, and T cal = 0.05 is the temperature drift calibration term, used to correct the thermal radiation deviation of the short-wave infrared channel.

[0084] S3 includes:

[0085] S31, constructing an improved U-Net++ segmentation network, whose input is the standard illumination epidermis image and the output is the feather area probability map and the epidermis area probability map. The specific steps are as follows:

[0086] (1) Feather area probability map generation: generating the feather area probability map P f (x,y) through a 3×3 convolution and a Softmax activation function, expressed as:

[0087]

[0088] where W f is the convolution kernel weight, b f is the bias term, is the feature map of the last layer of the decoder;

[0089] (2) Epidermal region probability map generation: Parallelly calculate the epidermal region probability map P e (x, y), and impose the constraint condition, expressed as:

[0090]

[0091] P f (x, y) + P e (x, y) ≤ 1;

[0092] S32. Optimize the probability map through the edge repair sub-network to generate the corrected feather edge region, expressed as:

[0093]

[0094] Among them, is the Laplacian operator, used to detect feather edge mutations, G σ (x, y) is the Gaussian kernel function (σ = 1.5), which realizes edge smoothing, and λ = 0.3 is the edge enhancement coefficient;

[0095] S33. Based on the corrected feather edge region, generate the feather distribution mask, expressed as:

[0096]

[0097] Among them, the threshold τ f = 0.65 is determined through ROC curve analysis to balance the recall rate and precision rate of feather detection;

[0098] S34. Extract the coordinate set Ω e of the epidermal exposure area, expressed as:

[0099] Ω e = {(x, y)|M f (x, y) = 0 and P e (x, y) > τ e};

[0100] Among them, τ e = 0.8 is the confidence threshold of the epidermal region.

[0101] S4 includes:

[0102] S41. Based on the coordinate set of the epidermal exposure area, extract the epidermal temperature of the feather gap. The specific steps are as follows:

[0103] (1) In the standard illumination epidermal image of channel 3 (short-wave infrared 940 - 960 nm), extract the radiation intensity value I e (x, y) of the pixels corresponding to the coordinate set Ω IR ;

[0104] (2) Calculate the skin temperature through the radiation-temperature calibration formula, expressed as:

[0105]

[0106] where a = 28.6, b = 0.45, c = 0.12 are calibration coefficients (determined through blackbody radiation experiments), δ3 = 1.0 is the short-wave infrared channel transmittance compensation factor defined in claim 2, is the near-infrared ambient illuminance component;

[0107] S42. Based on the feather distribution mask, calculate the feather coverage density index. The specific steps include:

[0108] (1) On the feather distribution mask M f (x, y), divide the grid area Γ with a unit of 50×50 pixels i ;

[0109] (2) Calculate the feather coverage density of each unit, expressed as:

[0110]

[0111] (3) Global density index calculation:

[0112]

[0113] where ω max = 0.6, ω avg = 0.4 are weight coefficients (satisfying ω max + ω avg = 1), and N is the total number of grids;

[0114] S43. Based on the extracted skin exposure area coordinate set, extract the skin texture roughness coefficient. The specific steps are as follows:

[0115] (1) Perform local binary pattern (LBP) encoding on the image within the Ω e area, expressed as:

[0116]

[0117] where P = 8 is the number of neighborhood sampling points, R = 2 is the sampling radius, g c is the gray value of the central pixel, g p is the neighborhood pixel value, and s(x) = 1 (if x ≥ 0) otherwise 0;

[0118] (2) Calculate the LBP histogram entropy value as the roughness coefficient:

[0119] where p(k) is the occurrence probability of the LBP pattern k.

[0120] S5 includes:

[0121] S51, discretize the input multi-dimensional physical sign parameters to construct a discrete feature vector;

[0122] S52, input the discrete feature vector into the pre-trained Bayesian network, and calculate four types of abnormal probabilities through the conditional probability table;

[0123] S53, determine the dominant abnormal type according to the four types of abnormal probabilities and their clinical weights, and generate an abnormal type code.

[0124] S51 includes:

[0125] S511, discretize the input multi-dimensional physical sign parameters, specifically including:

[0126] (1) The epidermal temperature T of the feather gap is expressed as:

[0127]

[0128] (2) The feather coverage density index F density is expressed as:

[0129]

[0130] (3) The epidermal texture roughness coefficient H rough is expressed as:

[0131]

[0132] S512, based on the discretized epidermal temperature of the feather gap, the feather coverage density index, and the epidermal texture roughness coefficient, construct a discretized feature vector, expressed as:

[0133] X disc =[T disc ,F disc ,H disc ,E temp ,E hum ,A day ,F ver ;

[0134] where E temp, E hum is the environmental temperature and humidity, A day is the day-age stage code (0 for the brooding period, 1 for the growth period, 2 for the fattening period), and F ver is the version number of the feeding plan.

[0135] S52 includes:

[0136] S521. Input the discretized feature vector into the pre-trained Bayesian network, and calculate the four types of anomaly probabilities through the conditional probability table, expressed as:

[0137]

[0138] where x i is the i-th component of the feature vector X disc , and Pa(x i ) is the set of parent nodes of x i (defined by the network structure), and P(x i | Pa(x i )) is the value in the conditional probability table (CPT);

[0139] S522. Calculate the total probability of physical sign anomalies, quantifying the probability of at least one anomaly occurring, expressed as:

[0140]

[0141] S53 includes:

[0142] S531. Determine the dominant anomaly type according to the four types of anomaly probabilities and their clinical weights, expressed as:

[0143] where the weight w k reflects the anomaly severity (w 01 = 1.2, w 02 = 1.5, w 03 = 1.8, w 04 = 1.0);

[0144] S532. Generate an anomaly type code based on the dominant anomaly type, expressed as:

[0145]

[0146] As Figure 2 shown, a body surface parameter monitoring system for meat duck breeding, which is used to implement the above-mentioned method for monitoring body surface parameters of meat ducks, includes the following modules:

[0147] Multi-spectral imaging device module: Collect the original image data of the epidermal area of the meat duck and synchronously obtain the ambient light intensity parameters;

[0148] Dynamic light compensation module: Perform dynamic light compensation processing on the original image data to generate a standard light epidermal image. The dynamic light compensation processing includes a reflectivity correction sub-module based on the ambient light intensity parameters;

[0149] Feather-epidermis region segmentation module: Segment the standard illumination epidermis image into feather-epidermis regions, generating a feather distribution mask and a coordinate set of the epidermis exposure area;

[0150] Multi-dimensional sign extraction module: Extract multi-dimensional sign parameters based on the coordinate set of the epidermis exposure area, including the epidermis temperature in the feather gap, the feather coverage density index, and the epidermis texture roughness coefficient;

[0151] Growth status assessment model: Input the multi-dimensional sign parameters into a pre-trained growth status assessment model to output the sign abnormality probability value and the corresponding abnormality type code.

[0152] The present invention covers any alternatives, modifications, equivalent methods, and solutions made within the spirit and scope of the present invention. To enable the public to have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention. However, those skilled in the art can fully understand the present invention without the description of these details. Additionally, well-known methods, processes, procedures, components, and circuits are not described in detail to avoid unnecessary confusion to the essence of the present invention.

[0153] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A method for monitoring body surface parameters for meat duck farming, characterized in that: The following steps are involved: S1, collecting raw image data of the duck skin area through a multispectral imaging device, and synchronously obtaining the ambient light intensity parameters; S2, performing dynamic illumination compensation processing on the original image data to generate a standard illumination epidermal image, wherein the dynamic illumination compensation processing includes a reflectance correction submodule based on an ambient illumination intensity parameter; S3, performing feather-epidermis region segmentation on the standard illumination epidermis image to generate a feather distribution mask and an epidermis exposure region coordinate set; S4, extracting multidimensional physical sign parameters based on the epidermal exposure area coordinate set, the multidimensional physical signs include feather gap epidermal temperature, feather coverage density index, and epidermal texture roughness coefficient; S5, input the multi-dimensional physical sign parameters into the pre-trained growth status assessment model, and output the physical sign abnormality probability value and the corresponding abnormality type code.

2. A method for monitoring body surface parameters for meat duck farming according to claim 1, characterized in that: The S1 includes: S11, triggering a synchronous acquisition module of a multi-spectral imaging device, wherein the multi-spectral imaging device includes a beam splitter prism group, a narrow-band filter array, and an embedded four-channel photodiode array; S12, performing spectroscopic processing on the target duck to generate four-channel original spectral data; S13, generating multi-spectral fused raw image data; S14, synchronously collecting ambient light intensity parameters.

3. A method for monitoring body surface parameters for meat duck farming according to claim 2, characterized in that: The S2 includes: S21, obtaining multi-spectral fused original image data and ambient light intensity parameters; S22, constructing a reflectivity correction submodule to calculate pixel-level dynamic compensation coefficients; S23, generating a standard illumination epidermis image based on the pixel-level dynamic compensation coefficient.

4. A method for monitoring body surface parameters for meat duck farming according to claim 3, characterized in that: The S3 includes: S31, construct an improved U-Net++ segmentation network, whose input is a standard illumination epidermis image, and outputs a feather area probability map and an epidermis area probability map; S32, optimizing the probability map through the edge repair sub-network to generate a corrected feather edge region; S33, generating a feather distribution mask based on the corrected feather edge region; S34, extracting the epidermal exposed area coordinate set.

5. A method for monitoring body surface parameters for meat duck farming according to claim 4, characterized in that: The S4 includes: S41, extracting the cuticle temperature in the feather gap based on the cuticle exposed area coordinate set; S42, calculating a feather cover density index based on the feather distribution mask; S43, extracting the epidermis texture roughness coefficient based on the extracted epidermis exposure area coordinate set.

6. A method for monitoring body surface parameters for meat duck farming according to claim 5, characterized in that: The S5 includes: S51, discretizing the input multi-dimensional physical sign parameters to construct a discrete feature vector; S52, inputting the discrete feature vector into the pre-trained Bayesian network, and calculating the probabilities of four types of anomalies through the conditional probability table; S53, determining the dominant abnormality type based on the four types of abnormality probabilities and their clinical weights, and generating an abnormality type code.

7. A method for monitoring body surface parameters for meat duck farming according to claim 6, characterized in that: The S51 includes: S511, discretizing the input multi-dimensional physical sign parameters; S512, constructing a discretized feature vector based on the discretized feather gap epidermal temperature, feather coverage density index, and epidermal texture roughness coefficient.

8. The method for monitoring body surface parameters for meat duck farming according to claim 7, characterized in that: The S52 includes: S521, inputting the discretized feature vector into the pre-trained Bayesian network, and calculating the four types of abnormal probabilities through the conditional probability table; S522, calculating the total probability of abnormal physical signs, and quantifying the probability of at least one abnormality occurring.

9. A method for monitoring body surface parameters for meat duck farming according to claim 8, characterized in that: The S53 includes: S531, determine the dominant abnormality type based on the four types of abnormality probabilities and their clinical weights.

10. A body surface parameter monitoring system for meat duck farming, used to implement a body surface parameter monitoring method for meat duck farming as claimed in any one of claims 1 to 9, characterized in that: Includes the following modules: Multispectral imaging device module: collects raw image data of the duck skin area and simultaneously obtains ambient light intensity parameters; Dynamic illumination compensation module: performing dynamic illumination compensation processing on the original image data to generate a standard illumination epidermal image, wherein the dynamic illumination compensation processing includes a reflectivity correction submodule based on an ambient illumination intensity parameter; Feather-epidermis region segmentation module: performs feather-epidermis region segmentation on the standard illumination epidermis image to generate feather distribution mask and epidermis exposure region coordinate set; Multi-dimensional physical sign extraction module: extracts multi-dimensional physical sign parameters based on the epidermal exposure area coordinate set, including feather gap epidermal temperature, feather coverage density index and epidermal texture roughness coefficient; Growth status assessment model: multi-dimensional physical sign parameters are input into the pre-trained growth status assessment model, and the probability value of physical sign abnormality and the corresponding abnormality type code are output.

Images