Living body matching state determination method and device, equipment, medium and program product

Abstract

The invention discloses a living body matching state determination method and device, equipment, a medium and a program product. The method comprises the following steps: determining a first feature; wherein the first feature comprises a feature set of postures of a first living object in a first time period; acquiring a second feature; wherein the second feature comprises a feature set of the posture of a second living object in a second time period; determining a feature association degree between the first feature and the second feature; and based on the feature association degree, determining a matching degree between the first living object in the first time period and the second living object in the second time period. Through the method, the identity of any living object can be flexibly and accurately identified from the time dimension.

Description

Technical Field

[0001] This application relates to the field of livestock identification technology, and in particular to a method, apparatus, equipment, medium and program product for determining the matching status of live animals. Background Technology

[0002] In practical applications, static identification of live livestock such as pigs, cattle, and sheep is mainly achieved by analyzing data collected by sensors or image acquisition devices. However, while these methods can identify a single livestock animal at a specific moment, they cannot provide comprehensive tracking of its identity. Therefore, none of these solutions can meet the requirements for identifying live livestock animals. Summary of the Invention

[0003] Based on the above technical problems, embodiments of this application provide a method, apparatus, device, medium, and program product for determining the living body matching status.

[0004] The technical solution provided in this application is as follows:

[0005] This application first provides a method for determining the liveness matching status, the method comprising:

[0006] Determine a first feature; wherein the first feature includes a set of features of the body posture of the first living object in a first time period;

[0007] Obtain the second feature; wherein the second feature includes a set of features of the body posture of the second living object during the second time period;

[0008] Determine the feature correlation degree between the first feature and the second feature;

[0009] Based on the aforementioned feature correlation, the degree of matching between the first living object in the first time period and the second living object in the second time period is determined.

[0010] In some embodiments, the first feature includes at least the body shape, appearance, and gender characteristics of the first living object; the second feature includes at least the body shape, appearance, and gender characteristics of the second living object; determining the feature correlation degree between the first feature and the second feature includes:

[0011] Determine the growth parameters of the first living object;

[0012] Based on the growth parameters, the body shape feature, appearance feature and gender feature in the first feature are processed to obtain the processing result;

[0013] Based on the data associated with body shape features, appearance features, and gender features in the processing results, and the body shape features, appearance features, and gender features in the second feature, the feature correlation degree is determined.

[0014] In some embodiments, determining the feature correlation degree between the first feature and the second feature includes:

[0015] Determine the first data type of the feature data contained in the first feature;

[0016] Determine the second data type of the feature data contained in the second feature;

[0017] Based on the first data type and the second data type, the feature data contained in the first feature and the feature data contained in the second feature are processed to determine the feature correlation degree.

[0018] In some embodiments, processing the feature data contained in the first feature and the feature data contained in the second feature based on the first data type and the second data type to determine the feature correlation degree includes:

[0019] If both the first data type and the second data type are discrete types, determine the dot product result between the first vector corresponding to the first feature and the second vector corresponding to the second feature;

[0020] The feature correlation degree is determined based on the dot product result.

[0021] In some embodiments, processing the feature data contained in the first feature and the feature data contained in the second feature based on the first data type and the second data type to determine the feature correlation degree includes:

[0022] If both the first data type and the second data type are continuous types, determine the feature similarity between the feature data contained in the first feature and the feature data contained in the second feature;

[0023] The feature correlation degree is determined based on the feature similarity.

[0024] In some embodiments, obtaining the second feature includes:

[0025] Determine the object identifier of the second living object;

[0026] The second feature is obtained from the live animal feature data set based on the object identifier; wherein the live animal feature data set includes the association between the physical characteristics of live objects in the livestock live animal set during at least one time period and the identifier of the live object.

[0027] In some embodiments, the method further includes:

[0028] During the m-th lifespan of the p-th living object, an image set including the p-th living object is collected; wherein, the images in the image set include at least the type characteristics, body shape characteristics, appearance characteristics, and sex characteristics of the p-th living object; p and m are both integers greater than or equal to 1; the p-th living object is any living object in the livestock living object set;

[0029] Feature recognition is performed on the images in the image set to obtain the feature set of the p-th living object during the m-th life period;

[0030] The feature sets of the p-th living object from the first life period to the M-th life period are integrated to obtain the feature data set of the p-th living object; where M is an integer greater than 1; m is less than or equal to M;

[0031] The live feature data set is determined based on the feature data set of the first live object to the feature data set of the Pth live object; where P is an integer greater than 1, and p is less than or equal to P.

[0032] In some embodiments, the step of performing feature recognition on the images in the image set to obtain the feature set of the p-th living object during the m-th life period includes:

[0033] Confirm the prompt data;

[0034] Based on the aforementioned prompting data, a multimodal visual model is controlled to segment the images in the image set to obtain segmentation results;

[0035] The segmentation results are identified and classified to obtain classification results; wherein, the classification results include a set of features of at least two body parts of the p-th living object;

[0036] By integrating the features from the classification results, a feature set of the p-th living object during the m-th life period is obtained.

[0037] This application embodiment also provides a live matching state determination device, the live matching state determination device comprising:

[0038] A determining module is used to determine a first feature; wherein the first feature includes a feature set of the body posture of the first living object in a first time period;

[0039] The acquisition module is used to acquire the second feature; wherein the second feature includes a feature set of the body posture of the second living object during the second time period;

[0040] The determining module is further configured to determine the feature correlation degree between the first feature and the second feature; and based on the feature correlation degree, determine the matching degree between the first living object in the first time period and the second living object in the second time period.

[0041] This application embodiment also provides a liveness matching status determination device, which includes a processor and a memory; the memory stores a computer program; when the computer program is executed by the processor, it can implement the liveness matching status determination method as described above.

[0042] This application also provides a computer-readable storage medium storing a computer program; when the computer program is executed by a processor of an electronic device, it can implement the liveness matching state determination method as described above.

[0043] This application also provides a computer program product, which includes a computer program; when the computer program is executed by the processor of an electronic device, it can implement the liveness matching state determination method as described above.

[0044] In the liveness matching status determination method provided in this application embodiment, the first feature includes a set of body posture features of the first live object in a first time period, and the second feature includes a set of body posture features of the second live object in a second time period. Thus, through the first feature and the second feature, the body posture features of the first live object and the second live object in different time periods can be displayed. Furthermore, by determining the feature correlation degree between the first feature and the second feature, the degree of correlation between the first live object and the second live object in the body posture feature dimension can be comprehensively and accurately characterized. On this basis, based on the feature correlation degree, the matching degree between the first live object in the first time period and the second live object in the second time period is determined, realizing the correlation of the matching degree between the first live object and the second live object in the time dimension. On the other hand, when the matching degree characterizes whether the first live object and the second live object are the same object, and the first live object is any live object, the above process can flexibly and accurately identify the identity of any live object in the time dimension.

[0045] When the technical solution provided in this application is applied to the insurance scenario for live livestock, and the first and second live objects are livestock such as pigs, cattle, and sheep, the above method can accurately identify the insured livestock represented by the first live object during the growth process of these animals. This reduces the probability that farmers will obtain unreasonable income by substituting uninsured, sick, or dead pigs, cattle, and sheep for insured ones, and reduces the probability of erroneous claims and economic losses for insurance organizations. This allows insurance organizations to conduct their insurance business stably and also provides a basis for insurance organizations to assess insurance claims. This provides data for contributing to and setting insurance rates; furthermore, when insured livestock are in an abnormal state such as illness or death, the above methods can also improve the efficiency of insurance companies' claims settlements, enabling farmers to quickly obtain insurance compensation, and also providing insurance companies with reference data on livestock breeding and insurance claims; at the same time, through the above operations, insurance companies can gain insight into the growth and development patterns of live livestock, thereby providing data support for the insurance of live livestock in different regions and / or breeds. For example, for regions or livestock types prone to disease, insurance companies can provide targeted insurance services and rates. Attached Figure Description

[0046] Figure 1 A flowchart illustrating the method for determining the liveness matching status provided in this application embodiment;

[0047] Figure 2 A schematic diagram illustrating the process of comparing live livestock in an embodiment of this application;

[0048] Figure 3 A schematic diagram illustrating the process of obtaining segmentation results provided in an embodiment of this application;

[0049] Figure 4 A schematic diagram illustrating the process of feature classification and archiving of living objects provided in the embodiments of this application;

[0050] Figure 5 Another schematic diagram of the process for comparing live livestock provided in the embodiments of this application;

[0051] Figure 6 This is a schematic diagram of the living body matching state determination device provided in the embodiments of this application;

[0052] Figure 7 This is a schematic diagram of the structure of the live matching status determination device provided in the embodiments of this application. Detailed Implementation

[0053] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0054] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0055] In related technologies, the identification and authentication of live livestock individuals are mainly achieved through two methods: processing the data collected by sensors to determine the identity of the live livestock, and extracting features from the image data collected by image acquisition devices to determine the identity of the live livestock.

[0056] However, sensor-based identification technology, which relies on sensors to acquire data, requires contact detection to identify live livestock. This can easily trigger stress responses in the animals, threatening their health and the safety of farm workers. On the other hand, image data acquired through image acquisition devices primarily relies on visual recognition technology to identify the livestock. However, this technology depends mainly on pixel features of the animal's head in the image data, such as only achieving cattle face recognition and pig face recognition. Therefore, this approach can only achieve static identification of a specific animal at a particular moment, and its accuracy is insufficient to meet the practical needs of identifying live livestock.

[0057] Based on the above technical problems, this application provides a method for determining the living body matching status. Figure 1 This is a flowchart illustrating the method for determining the liveness matching status provided in an embodiment of this application, as shown below. Figure 1 As shown, the process may include the following steps:

[0058] Step 101: Determine the first feature.

[0059] The first feature includes a set of features of the body posture of the first living object in the first time period.

[0060] In one implementation, the first time period may include the current time period or a historical time period.

[0061] In one implementation, the first live animal object may include any type of livestock; for example, the first live animal object may include any type of pig, cattle, or sheep.

[0062] In one implementation, the first living object can be any living object in the livestock living object set; for example, the livestock living object set can include at least one type of livestock living object.

[0063] In one implementation, the first living object can be at any stage of growth and development of the livestock of its type.

[0064] In one embodiment, the first feature may include at least one static feature selected from the body shape feature, appearance feature, gender feature, and age feature of the first living object; for example, the body shape feature may characterize at least one of the height, leg length, body length, waist circumference, abdominal circumference, and head circumference of the first living object; for example, the appearance feature may include the skin color, hair length and / or brightness, eye size, mouth shape, ear shape and / or size, and tail shape of the first living object; for example, the gender feature may include female and male; for example, the age feature may characterize the stage of the first living object in its life cycle, such as the age feature may include at least one of infancy, adolescence, adulthood, and old age; for example, the tail shape may include the thickness and length of the tail.

[0065] In one implementation, the first feature may include dynamic features such as the posture and / or behavior of the first living object; for example, the dynamic features may include the walking posture and / or foraging behavior of the first living object.

[0066] In one implementation, the first feature can be determined in the following way:

[0067] Images of a first living object are acquired from multiple angles using an image acquisition device to obtain an object image set. Feature extraction and recognition are performed on the images in the object image set to obtain a set of object features. The set of object features is then determined as the first feature. For example, the resolution of the image acquisition device can be 1920*1080. Furthermore, in a well-lit environment, the image acquisition device can be controlled to acquire images of the first living object in 360 degrees to improve the clarity of the images in the object image set.

[0068] Step 102: Obtain the second feature.

[0069] The second feature includes a set of features of the body posture of the second living object during the second time period.

[0070] In one implementation, the second time period can be a historical period of the first time period.

[0071] In one implementation, the second time period may include at least one time period of the life cycle of the second living object.

[0072] In one implementation, the type of the first living object and the type of the second living object can be the same.

[0073] In one implementation, the first living object and the second living object may be the same or different.

[0074] In one embodiment, the second feature may include at least one static feature selected from the body shape, appearance, sex, and age of the second living object; for example, the body shape feature may characterize at least one of the height, leg length, body length, waist circumference, abdominal circumference, and head circumference of the second living object; for example, the appearance feature may include the skin color, hair length and / or brightness, eye size, mouth shape, ear shape and / or size, and tail shape of the second living object; for example, the sex feature may include female and male; for example, the age feature may characterize the stage of the second living object in its life cycle, such as at least one of infancy, adolescence, adulthood, and old age; for example, the tail shape may include the thickness and length of the tail.

[0075] In one implementation, the second feature may include dynamic features such as the posture and / or behavior of the second living object; for example, the dynamic features may include the walking posture and / or foraging behavior of the second living object.

[0076] In one implementation, the second feature can be obtained in the following way:

[0077] The second feature is obtained from the feature set stored in the preset storage space; wherein, the feature set may include a set of features of the second living object in multiple age periods.

[0078] Step 103: Determine the feature correlation degree between the first feature and the second feature.

[0079] In one implementation, feature correlation can characterize whether there is a correlation between a first feature and a second feature.

[0080] In one implementation, feature correlation can characterize the correlation or similarity between a first feature and a second feature.

[0081] In one implementation, the feature correlation degree can be determined in the following way:

[0082] The Euclidean distance between the x-th feature data in the first feature and the x-th feature data in the second feature is determined as the x-th correlation degree. Then, the set of the first correlation degree to the x-th correlation degree is determined as the feature correlation degree. Here, x can be an integer greater than or equal to 1 and less than or equal to X. X is an integer greater than 1 and is used to characterize the number of feature data contained in the first feature and the second feature, respectively.

[0083] Step 104: Based on feature correlation, determine the degree of matching between the first live object in the first time period and the second live object in the second time period.

[0084] In one implementation, the degree of matching can characterize the similarity between the first living object and the second living object; for example, the degree of matching can characterize whether the first living object and the second living object are related by blood.

[0085] In one implementation, the degree of matching can characterize whether the first living object and the second living object are the same object at different times. In other words, the identity of the first living object can be confirmed by the degree of matching.

[0086] In one implementation, the degree of matching can be determined in the following way:

[0087] The degree of matching is determined based on the strength of the feature correlation; for example, the degree of matching can increase as the feature correlation increases.

[0088] In one implementation, if the degree of matching is greater than or equal to the degree threshold, it can be characterized that the first living object and the second living object are the same living object in different time periods; if the degree of matching is less than the degree threshold, it can be characterized that the first living object and the second living object are not the same living object.

[0089] As can be seen from the above, in the liveness matching status determination method provided in this application embodiment, the first feature includes a set of body posture features of the first live object in the first time period, and the second feature includes a set of body posture features of the second live object in the second time period. Thus, through the first feature and the second feature, the body posture features of the first live object and the second live object in different time periods can be displayed. Furthermore, by determining the feature correlation degree between the first feature and the second feature, the correlation degree between the first live object and the second live object in the body posture feature dimension can be comprehensively and accurately characterized. On this basis, based on the feature correlation degree, the matching degree between the first live object in the first time period and the second live object in the second time period is determined, realizing the correlation of the matching degree between the first live object and the second live object in the time dimension. On the other hand, when the matching degree characterizes whether the first live object and the second live object are the same object, and the first live object is any live object, the above process can flexibly and accurately identify the identity of any live object in the time dimension.

[0090] When the technical solution provided in this application is applied to the insurance scenario for live livestock, and the first and second live objects are livestock such as pigs, cattle, and sheep, the above method can accurately identify the insured livestock represented by the first live object during the growth process of these animals. This reduces the probability that farmers will obtain unreasonable income by substituting uninsured, sick, or dead pigs, cattle, and sheep for insured ones, and reduces the probability of erroneous claims and economic losses for insurance organizations. This allows insurance organizations to conduct their insurance business stably and also provides a basis for insurance organizations to assess insurance claims. This provides data for contributing to and setting insurance rates; furthermore, when insured livestock are in an abnormal state such as illness or death, the above methods can also improve the efficiency of insurance companies' claims settlements, enabling farmers to quickly obtain insurance compensation, and also providing insurance companies with reference data on livestock breeding and insurance claims; at the same time, through the above operations, insurance companies can gain insight into the growth and development patterns of live livestock, thereby providing data support for the insurance of live livestock in different regions and / or breeds. For example, for regions or livestock types prone to disease, insurance companies can provide targeted insurance services and rates.

[0091] Based on the foregoing embodiments, in the live matching status determination method provided in this application, the first feature includes at least the body shape feature, appearance feature, and gender feature of the first live object; the second feature includes at least the body shape feature, appearance feature, and gender feature of the second live object.

[0092] In one implementation, the first feature and the second feature may further include a type feature; for example, the type feature may include features such as mammal type, non-mammal type, and bird type.

[0093] In one implementation, the type feature can also characterize features such as pigs, cattle, and sheep in mammal types.

[0094] Accordingly, determining the feature correlation degree between the first feature and the second feature can be achieved through the following steps:

[0095] Step A1: Determine the growth parameters of the first living subject.

[0096] In one embodiment, the growth parameters may include the growth and development patterns of the livestock type represented by the first living object during its life cycle; for example, the growth and development patterns may include the changes in static data such as body shape and / or appearance of the livestock type represented by the first living object from birth to adulthood and old age, and may also include the changes in dynamic data such as walking and / or foraging during the above-mentioned stages.

[0097] In one implementation, the growth parameters can be determined in the following way:

[0098] Based on the type of livestock corresponding to the first living object, the parameters stored in the growth database are filtered, and the filtering results are determined as growth parameters; wherein, the growth database can store parameters that can characterize the growth and development patterns of various livestock.

[0099] Step A2: Process the body shape, appearance and gender features in the first feature based on the growth parameters to obtain the processing results.

[0100] In one implementation, the data in the processing result can be associated with the body shape characteristics, appearance characteristics, and gender characteristics of the first living object, respectively.

[0101] In one implementation, the processing result can be obtained in the following way:

[0102] The first time period corresponding to the first living object is determined within the life cycle corresponding to the growth parameters. The second time period is determined within the life cycle of the first living object. Then, based on the time difference between the first and second time periods and the growth and development pattern represented by the growth parameters, the above features in the first feature are scaled in size dimension and strengthened or weakened in color dimension to obtain the processing result.

[0103] If the type feature in the first feature matches the type feature in the second feature, the first feature is processed based on the growth parameters to obtain the processing result.

[0104] Step A3: Based on the data associated with body shape features, appearance features, and gender features in the processing results, and the body shape features, appearance features, and gender features in the second feature, determine the feature correlation degree.

[0105] In one implementation, the feature correlation degree can be determined in the following way:

[0106] The degree of feature correlation is determined by comparing the data associated with type features, appearance features, and gender features in the processing results with the differences between these data and the body shape features, appearance features, and gender features in the second feature.

[0107] As can be seen from the above, in the live matching state determination method provided in this application embodiment, the first feature includes at least the body shape features, appearance features, and gender features of the first live object, and the second feature includes at least the body shape features, appearance features, and gender features of the second live object. Thus, the first feature and the second feature can comprehensively and accurately characterize the feature attributes of the first live object and the second live object. Furthermore, after determining the growth parameters of the first live object, the body shape features, appearance features, and gender features in the first feature are processed based on the growth parameters to obtain the processing results, realizing targeted processing of various features in the first feature in the growth cycle dimension of the first live object. On this basis, based on the data associated with the body shape features, appearance features, and gender features in the processing results, and with the body shape features, appearance features, and gender features in the second feature, the feature correlation degree is determined, which can improve the completeness and comprehensiveness of the data included in the feature correlation degree, so that the feature correlation degree can reflect the degree of difference between the first live object and the second live object from the time dimension of the growth and development of the first live object.

[0108] Based on the foregoing embodiments, the method for determining the liveness matching state provided in this application can further be implemented by the following steps:

[0109] Step B1: Determine the first data type of the feature data contained in the first feature.

[0110] In one implementation, the first data type may include basic data types or extended data types of various feature data contained in the first feature; for example, the basic data type may include integer and floating-point types, and the extended data type may include data types other than the basic data type.

[0111] In one implementation, the first data type can be determined in the following way:

[0112] The data structure of the feature data contained in the first feature is analyzed to determine the first data type.

[0113] Step B2: Determine the second data type of the feature data contained in the second feature.

[0114] In one implementation, the second data type and its determination process can be similar to the first data type and its determination process, and will not be described in detail here.

[0115] Step B3: Based on the first data type and the second data type, process the feature data contained in the first feature and the feature data contained in the second feature to determine the feature correlation degree.

[0116] In one implementation, the feature correlation degree can be determined in the following way:

[0117] If the first data type matches the second data type, the feature data contained in the first feature and the feature data contained in the second feature are processed using the first strategy to determine the feature correlation degree; if the first data type does not match the second data type, the feature data contained in the first feature and the second feature are processed using the second strategy to determine the feature correlation degree.

[0118] For example, the first strategy and the second strategy may be different, and the second strategy may include data transformation of the first feature and / or the second feature so that the type of feature data represented by the first feature can match the type of feature data represented by the second feature.

[0119] As can be seen from the above, the liveness matching state determination method provided in this application, after determining the first data type of the feature data contained in the first feature and the second data type of the feature data contained in the second feature, processes the feature data contained in the first feature and the feature data contained in the second feature based on the first data type and the second data type to determine the feature correlation degree. Thus, by using the first data type and the second data type, the data types of the feature data contained in the first feature and the second feature can be comprehensively and accurately characterized; moreover, by processing the first feature and the second feature based on the first data type and the second data type, targeted processing of the feature data contained in the first feature and the feature data contained in the second feature is achieved, thereby improving the accuracy of the feature correlation degree.

[0120] Based on the foregoing embodiments, the liveness matching state determination method provided in this application, which processes the feature data contained in the first feature and the feature data contained in the second feature based on the first data type and the second data type to determine the feature correlation degree, can be achieved through the following steps:

[0121] Step C1: If both the first data type and the second data type are discrete, determine the dot product between the first vector corresponding to the first feature and the second vector corresponding to the second feature.

[0122] Accordingly, if at least one of the first data type and the second data type is not discrete, then the dot product result between the first vector and the second vector can be uncertain.

[0123] In one implementation, both the first data type and the second data type are discrete types, which may include feature data contained in the first feature and feature data contained in the second feature, both of which are discrete data obtained after quantization.

[0124] In one implementation, the first vector may include multiple vectors, and the multiple vectors in the first vector are respectively used to characterize the type characteristics, body shape characteristics, appearance characteristics and gender characteristics of the first living object.

[0125] In one implementation, the second vector may include multiple vectors, and the multiple vectors in the second vector are respectively used to characterize the type characteristics, body shape characteristics, appearance characteristics and gender characteristics of the second living object.

[0126] In one implementation, the dot product result can characterize the degree of difference between the first vector and the second vector.

[0127] In one implementation, the dot product result can be obtained by performing a dot product between the vectors contained in the first vector and the vectors contained in the second vector.

[0128] Step C2: Determine the feature correlation degree based on the dot product result.

[0129] In one implementation, the feature correlation degree can be obtained in the following way:

[0130] The dot product result is weighted based on the number of feature data contained in the first feature and the number of feature data contained in the second feature, and the result of the weighted calculation is determined as the feature correlation degree.

[0131] As can be seen from the above, in the liveness matching state determination method provided in this application embodiment, if both the first data type and the second data type are discrete types, the dot product result between the first vector corresponding to the first feature and the second vector corresponding to the second feature is determined, and the feature correlation degree is determined based on the dot product result. Thus, the computational efficiency of the dot product result is improved, thereby improving the efficiency of determining the feature correlation degree; furthermore, by calculating the dot product result, the degree of feature difference between the first vector corresponding to the first feature and the second vector corresponding to the second feature can be more accurately reflected, thereby improving the accuracy of the feature correlation degree.

[0132] Based on the foregoing embodiments, the liveness matching state determination method provided in this application, which processes the feature data contained in the first feature and the feature data contained in the second feature based on the first data type and the second data type to determine the feature correlation degree, can also be achieved through the following steps:

[0133] Step D1: If both the first data type and the second data type are continuous types, determine the feature similarity between the feature data contained in the first feature and the feature data contained in the second feature.

[0134] Accordingly, if at least one of the first data type and the second data type is not a continuous type, then feature similarity can be determined without using the above method.

[0135] In one implementation, when both the first data type and the second data type are continuous types, feature similarity can characterize the correlation coefficient between the first feature and the second feature.

[0136] Accordingly, the feature similarity can be determined by calculating the correlation coefficient for the feature data in the first feature and the feature data in the second feature.

[0137] For example, feature similarity dim(t) A ,t k It can be calculated using equation (1):

[0138]

[0139] If both the first and second data types are discrete types, then through Calculate feature similarity, which can be the dot product result from the previous embodiment; if both the first and second data types are continuous types, then... Calculate feature similarity; and t A The first feature can be the characteristic of the first living object, such as object A, i.e., the first feature, t. k This can be a second feature of the second living object; k is used to characterize the time period identifier of the second time period; t Ai For the i-th feature data in the first feature, t ki For the i-th feature data in the second feature, dim(t) A ,t k ) is the feature comparison function.

[0140] Step D2: Determine the feature correlation degree based on feature similarity.

[0141] In one implementation, the feature correlation degree can be determined by weighting the feature similarity; for example, the feature correlation degree Sim can be calculated using equations (2) to (3). A .

[0142]

[0143] σ k =log(k+1) (3)

[0144] Equation (3) is the attenuation function, and its calculation result is used to represent the attenuation value corresponding to the time difference between the second time period represented by k and the first time period. Its value can increase as the time difference increases.

[0145] As can be seen from the above calculations, in the embodiments of this application, when determining the feature correlation degree, if both the first feature and the second feature are discrete types, the feature similarity can be determined by the Onehot encoded vector dot product method. However, when both the first feature and the second feature are continuous types, the pre-similarity can be used for calculation.

[0146] In one implementation, if the feature correlation degree is greater than 0.75, it can be determined that the first living object and the second living object are the same living object; if the feature correlation degree is greater than or equal to 0.5 but less than 0.75, it can be determined that the first living object and the second living object are similar, and the similarity between the first living object and the second living object can be confirmed by manual intervention; if the feature correlation degree is less than 0.5, it can be determined that the first living object and the second living object are not the same living object.

[0147] As can be seen from the above, in the liveness matching state determination method provided in this application embodiment, if both the first data type and the second data type are continuous types, the feature similarity between the feature data contained in the first feature and the feature data contained in the second feature is determined. In this way, the feature similarity between the first feature and the second feature, which are both continuous types, is determined in a targeted manner. Furthermore, determining the feature correlation degree based on feature similarity can improve the accuracy of feature correlation degree.

[0148] Based on the foregoing embodiments, the method for determining the liveness matching state provided in this application can obtain the second feature in the following ways:

[0149] Determine the object identifier of the second living object; obtain the second feature from the living feature data set based on the object identifier.

[0150] The live animal feature data set includes the association between the physical characteristics of live animals in the livestock live animal set and the identification of the live animals during at least one time period.

[0151] In one implementation, the object identifier of the second living object may include the name and / or number of the second living object.

[0152] In one implementation, the set of liveness features can be obtained in the following way:

[0153] During the growth and development of live animals in a livestock collection, a set of images of the live animals is collected in advance, and feature extraction and recognition are performed on the images in the set to obtain feature data. Then, the feature data is associated with the identifier of the live animals to obtain a set of live animal features.

[0154] In one implementation, at least one time period may include the life period encompassed by the life cycle of a living object.

[0155] In one implementation, the second feature can be obtained in the following way:

[0156] Based on the matching relationship between the identifiers of live objects in the association relationship between the object identifier and the live feature data set, the target relationship is determined from the live feature data set, and the feature data contained in the target relationship is determined as the second feature; wherein, the identifier of the live object contained in the target relationship can be matched with the object identifier.

[0157] As can be seen from the above, the method for determining the matching status of a living object provided in this application, after determining the object identifier of the second living object, obtains a second feature from the living object feature data set based on the object identifier. The living object feature data set includes the association between the physical characteristics of living objects in the livestock live object set during at least one time period and the identifier of the living object. Thus, through the living object feature data set, the physical characteristics of living objects in the livestock live object set during at least one time period can be comprehensively and thoroughly characterized, and the efficiency of obtaining the second feature can be improved.

[0158] Based on the foregoing embodiments, the liveness matching status determination method provided in this application can also perform the following steps:

[0159] Step E1: During the m-th lifetime of the p-th living object, acquire a set of images including the p-th living object.

[0160] The images in the image set include at least the type features, body shape features, appearance features, and gender features of the p-th living object; p and m are both integers greater than or equal to 1; the p-th living object is any living object in the livestock living object set.

[0161] In one embodiment, the image set may include multiple images, and the features contained in the images may be the same or different.

[0162] In one implementation, the image set can be obtained in the following way:

[0163] During the m-th lifetime period, the features of the p-th living object are collected from all directions using an image acquisition device to obtain the p-th video data. Then, the p-th video data is decoded and processed by frame extraction to obtain an image set.

[0164] Step E2: Perform feature recognition on the images in the image set to obtain the feature set of the p-th living object during the m-th life period.

[0165] In one implementation, the feature set of the p-th living object at the m-th lifetime can be obtained in the following way:

[0166] By performing feature recognition on the images in the image set using a neural network model or an artificial intelligence model, multiple recognition results are obtained. These multiple recognition results are then integrated to obtain the feature set of the p-th living object at the m-th life period.

[0167] Step E3: Integrate the feature sets of the p-th living object from the first life period to the M-th life period to obtain the feature data set of the p-th living object.

[0168] Where M is an integer greater than 1; m is less than or equal to M.

[0169] In one implementation, the Mth lifetime period may include the lifetime period in which the pth living object is located when the feature set of the pth living object was last determined.

[0170] In one implementation, the feature data set of the p-th living object can be determined in the following way:

[0171] Based on the sequential relationship between the first life period and the Mth life period in the life cycle of the p-th living object, the feature sets of the p-th living object from the first life period to the Mth life period are integrated along the time dimension to obtain the feature data set of the p-th living object.

[0172] Step E4: Determine the live feature data set based on the feature data set of the first live object to the feature data set of the Pth live object.

[0173] Where P is an integer greater than 1, and p is less than or equal to P.

[0174] In one implementation, the liveness feature dataset can be determined in the following way:

[0175] The p-th relationship is obtained by associating the identifier of the p-th living object with the feature data set of the p-th living object, and the set of the p-th relationships is determined as the feature data corresponding to the p-th living object in the living feature data set.

[0176] Figure 2 This is a flowchart illustrating the process of comparing live livestock in an embodiment of this application, as shown below. Figure 2 As shown, the process may include the following steps:

[0177] Step 201: Collect live animal data.

[0178] For example, livestock live data may include the image set described in the foregoing embodiments.

[0179] Step 202: Feature extraction.

[0180] For example, features can be extracted from images in an image set using neural network models, artificial intelligence models, or big data models, thereby obtaining the body shape features, appearance features, gender features, and type features of a living object.

[0181] Step 203: Construct a time feature map.

[0182] For example, the temporal feature map may include the live feature data set in the foregoing embodiments. The feature set of the p-th live object in each life period can be integrated according to the time sequence by the method provided in the foregoing embodiments to obtain the temporal feature map of the p-th live object.

[0183] Step 204: Live animal comparison.

[0184] For example, a first feature of any living object in the current time period can be obtained, and a second feature can be obtained from the time feature map based on the object identifier. Based on the feature correlation between the first feature and the second feature, the degree of matching between any living object and the living object corresponding to the second feature can be determined, thereby realizing the live comparison between any living object and the living object corresponding to the second feature.

[0185] Step 205: Output the results.

[0186] For example, the method provided in the foregoing embodiments can be used to determine whether any living object and the living object corresponding to the second feature are the same object based on the degree of matching, and the result can be output.

[0187] Through the above process, the collection of livestock live data, feature extraction, and construction of time feature maps are carried out in sequence, thus providing comprehensive and detailed data support for livestock live comparison and improving the accuracy of livestock live comparison results.

[0188] As can be seen from the above, in the live animal matching status determination method provided in this application embodiment, during the m-th life period of the p-th live animal, an image set including the p-th live animal is collected, and the images in the image set include at least the type features, body shape features, appearance features, and gender features of the p-th live animal. Thus, through the images in the image set, the overall features of the p-th live animal during the m-th life period can be comprehensively and accurately displayed. Furthermore, feature recognition is performed on the images in the image set to obtain the feature set of the p-th live animal during the m-th life period, realizing comprehensive recognition of the features carried by the images in the image set. On this basis, by integrating the feature sets of the p-th live animal from the first life period to the M-th life period, the feature data set of the p-th live animal is obtained. Thus, through the feature data set of the p-th live animal, the feature change pattern of the p-th live animal during each life period can be comprehensively displayed from the time dimension of the p-th live animal's life period. On the other hand, by determining the live animal feature data set based on the feature data set of the first live animal to the feature data set of the p-th live animal, the continuous tracking of the features of any live animal in the livestock live animal set in the time dimension is realized.

[0189] Based on the foregoing embodiments, the liveness matching status determination method provided in this application involves performing feature recognition on images in an image set to obtain the feature set of the p-th live object during its m-th lifetime. This can be achieved through the following steps:

[0190] Step F1: Confirm the prompt data.

[0191] In one implementation, the prompt data includes predetermined data for instructing the segmentation of images in the image set; for example, the prompt data may include the name or number of a body part of the p-th living object, such as the face, torso, limbs, and tail; for example, the prompt data may also include the overall name of the p-th living object, such as the overall pixel region of a pig, cow, or sheep.

[0192] Step F2: Based on the prompt data, control the multimodal vision model to segment the images in the image set to obtain the segmentation results.

[0193] In one implementation, the segmentation result may include sub-images obtained by segmenting a portion of pixel regions in an image set; for example, the segmentation result may include a set of pixel data corresponding to the body parts represented by the cue data.

[0194] In one implementation, the segmentation result can be obtained in the following way:

[0195] A masking layer is constructed based on the prompt data, and the multimodal vision model is controlled through the masking layer to perform masking processing on the images in the image set to obtain the masking result. The masking result is then segmented to obtain the segmentation result.

[0196] Figure 3 This is a schematic diagram of the process for obtaining segmentation results provided in an embodiment of this application, such as... Figure 3 As shown, the process may include the following steps:

[0197] Step 301: Obtain the image set.

[0198] For example, the images in the image collection can be arrays of 1024*1024 in three channels.

[0199] Step 302, Image Encoding.

[0200] For example, images in an image set can be encoded using a 4-layer, 16-kernel convolutional network to obtain a 32-channel, 16384-dimensional vector.

[0201] Step 303: Obtain the vectorized representation of the image.

[0202] For example, a 16384-dimensional vector with 32 channels corresponding to each image in the image set can be obtained, thereby obtaining a vectorized representation of the images in the image set.

[0203] Step 304: Extract features.

[0204] For example, feature extraction can be performed on the vectorized image using a multi-layer Transformer Encoder to obtain the feature extraction result.

[0205] For example, by using a bidirectional multi-head attention structure, a 16-layer Transformer Encoder module stack can be formed through attention units, feedforward units, and residual linking units. The first dimension of the output of the last hidden state is used as the output of the encoder structure and passed to the next structure.

[0206] Step 305: Encoding the prompt word.

[0207] For example, prompt data can be predetermined and encoded to obtain prompt word encoding results; for example, the prompt word encoding results can include the embedding results of the prompt data.

[0208] Step 306: Mask layer encoding.

[0209] For example, the prompt word encoding result and the feature extraction result can be input into the masking layer to control the display and hiding of specific areas in the feature extraction result.

[0210] Step 307: Output the segmentation results.

[0211] For example, the segmentation result output by the mask layer can be a 128*128 soft mask, used to mark the contour boundaries of the body parts indicated by the cue data in the image; for example, the cue word encoding result is also output to the segmentation result simultaneously to associate the segmentation result with the cue word encoding result.

[0212] Through the above process, accurate and flexible segmentation of any image in the image set is achieved.

[0213] Step F3: Identify and classify the segmentation results to obtain the classification results.

[0214] The classification result includes a set of features of at least two body parts of the p-th living object.

[0215] For example, the segmentation results can be extracted and identified using a deep learning (DL) model with multi-task learning, thereby obtaining feature data to represent dimensions such as type, gender, appearance, and body shape. Then, the feature data can be classified and archived using a deep learning model with multi-task learning to obtain classification results. For example, the feature data of the appearance dimension can include features of parts such as skin color, pattern, nose, lips, mouth, and ears.

[0216] For example, a multi-task learning DL model may include a multi-output convolutional neural network (CNN); for example, the CNN may include an input layer, a feature extraction layer, and multiple output layers.

[0217] For example, the input layer of a CNN can take in a sub-image corresponding to a 128*128 segmentation result with 3 channels.

[0218] For example, the feature extraction layer of a CNN may include three convolutional layers, three pooling layers, a flattening layer, and a fully connected layer; wherein, the first convolutional layer may include 32 3x3 convolutional kernels and use ReLU as the activation function, and the first pooling layer can achieve 2x2 max pooling; the second convolutional layer includes 64 3x3 convolutional kernels and uses ReLU as the activation function, and the second pooling layer can achieve 2x2 max pooling; the third convolutional layer may include 128 3x3 convolutional kernels and use ReLU as the activation function, and the third pooling layer can achieve 2x2 max pooling; the flattening layer is used to flatten the three-dimensional feature map into a one-dimensional vector; the fully connected layer may include 512 neurons and use ReLU as the activation function.

[0219] For example, the feature extraction layer may also include a Dropout layer with a parameter of 0.6 to reduce the probability of overfitting.

[0220] For example, the output layer may include eight output units. The first output unit may include two neurons for extracting gender features and classifying them using a softmax activation function. The second output unit may include three neurons for extracting category features, including type features such as pig, cow, and sheep, and classifying them using a softmax activation function. The third output unit may include five neurons for extracting skin color features, including black, white, yellow, brown, and gray, and classifying them using a softmax activation function. The fourth output unit may include three neurons for extracting pattern features and classifying them using a softmax activation function. The fifth output unit may include 32 neurons for extracting lip features, and its output may be an embedding vector. The sixth output unit may include 32 neurons for extracting nose features, and its output may be an embedding vector. The seventh output unit may include 32 neurons for extracting mouth features, and its output may be an embedding vector. The eighth output unit may include 32 neurons for extracting ear features, and its output may be an embedding vector.

[0221] The CNN described above effectively identifies and classifies the features of various body parts of different living objects using a single backbone neural network and a multi-branch network of neurons.

[0222] Before using a CNN, sample data needs to be collected to train the initial CNN. For example, the sample data can broadly cover color images of various livestock, including pigs, cattle, and sheep, to improve the diversity and representativeness of livestock types and individuals. For example, after obtaining the color image data, it can be standardized by cropping and / or scaling to adjust it to a 128*128*3 vector, and then the initial CNN can be trained based on this vector. During this training process, the CNN will capture the correlations between features of various types of live objects, thereby improving the accuracy of CNN feature recognition. The number of images in the color image data can be 300,000, and the ratio of images for pigs, cattle, and sheep can be 1:1:1.

[0223] For example, color image data can carry annotation information to indicate the feature type and feature value corresponding to the pixel data contained in the color image data; and, during training, the training process can be optimized by adjusting hyperparameters, wherein the hyperparameters can include the following parameters:

[0224] The learning rate can be set to an initial learning rate of 0.001, with a learning rate decay parameter of 0.1. The optimizer can be set to Stochastic Gradient Descent (SGD), with a momentum of 0.9 and a weight decay of 5 * 10^6. -4 The batch size is 64; the L2 weight decay of regularization is 1*10. -4 Or 5*10 -4 The dropout ratio is 0.45; the number of iterations (epochs) is set to 300; early stopping is set to 10; and training is stopped if the validation set performance does not improve within 10 epochs.

[0225] For example, the loss function for the above training can be as shown in (4):

[0226]

[0227] Where Lt represents the loss of the t-th task, wt represents the loss weight of the t-th task, N represents the total number of tasks, and the total loss L can be expressed as the weighted sum of the losses of all tasks; n can be an integer greater than 1.

[0228] For the t-th task, if it is a multi-class classification problem, its cross-entropy loss L t It can be calculated using equation (5):

[0229]

[0230] Where: m is the batch size, K is... t It is the number of categories for the t-th task, y t,i,k This is the true label of the i-th sample belonging to class k on task t. Its value can be 0 or 1, where a value of 1 indicates that the i-th sample belongs to class k on task t, and a value of 0 indicates that the i-th sample does not belong to class k on task t. It is the predicted probability that the i-th sample belongs to category k on the t-th task.

[0231] Step F4: Integrate the features from the classification results to obtain the feature set of the p-th living object during the m-th life period.

[0232] In one implementation, the feature set of the p-th living object during its m-th lifetime can be obtained in the following way:

[0233] The features in the classification results are integrated according to the identifier of the p-th living object and the m-th life period of the p-th living object to obtain the feature set of the p-th living object in the m-th life period.

[0234] In one implementation, the feature set of the p-th living object at any time stage can be stored in the form of a structure or in the form of JSON (JavaScript Object Notation); wherein, some feature data in the feature set stored in JSON format can be as shown in Table 1.

[0235] Table 1 may include two columns of data: field name and field type. The field name is used to represent the name of the feature represented by the feature data. The field type may include the type of value data of the feature represented by the field name. For example, the timestamp may include the life period of the p-th living object, and the data type of the life period is int64. The living object number may be used to represent the number of the p-th living object, and its data type can be string. The feature name may include gender features, body shape features, appearance features, and type features in the aforementioned embodiments, and its data type can be string. The feature value may represent the value of the feature indicated by the feature name, and its data type can be string. For example, the feature value may represent the length of the limbs.

[0236] Field Name Field type Timestamp int64 Live animal number string Feature Name string Eigenvalues string

[0237] Table 1

[0238] Figure 4 This is a schematic diagram illustrating the process of feature classification and archiving of living objects provided in an embodiment of this application, as shown below. Figure 4 As shown, the process may include the following steps:

[0239] Step 401, Begin.

[0240] Step 402: Video stream input.

[0241] For example, the p-th video data of the p-th living object can be acquired by an image acquisition device and input as the p-th video data.

[0242] Step 403: Frame extraction from the video stream.

[0243] For example, the p-th video data after decoding can be sampled and extracted using a uniform sampling method to obtain an image set.

[0244] Step 404: Target feature extraction.

[0245] For example, images in an image set can be segmented and features extracted using a multimodal visual model to obtain feature extraction results; for example, target features may include the gender, appearance, and body shape-related features of the p-th living object in the image.

[0246] Step 405: Feature classification and archiving.

[0247] For example, the features in the feature extraction results can be classified using CNN to obtain the classification results, and the feature data in the classification results can be archived to obtain the feature set of the p-th live object, and then the live feature data set of all live objects in the livestock live set can be obtained.

[0248] Step 406, End.

[0249] The above method can efficiently and flexibly obtain the feature set of any living object in the livestock live animal collection.

[0250] As can be seen from the above, the liveness matching state determination method provided in this application, after determining the prompt data, controls a multimodal visual model to segment the images in the image set based on the prompt data to obtain the segmentation result. This not only improves the targeting of the image segmentation in the image set, but also improves the accuracy of the segmentation operation. Furthermore, after identifying and classifying the segmentation result to obtain the classification result, the features in the classification result are integrated to obtain the feature set of the p-th live object. The classification result includes the set of features of at least two body parts of the p-th live object. In this way, the comprehensiveness and completeness of the p-th live object can be improved.

[0251] Figure 5 Another schematic diagram of the process for comparing live livestock provided in the embodiments of this application, as shown below. Figure 5 As shown, this process can include three stages: the main stage, the construction of a feature map, and the extraction of target features. Specifically, it can include the following steps:

[0252] Step 501: Collect live animal data.

[0253] For example, livestock live data may include video streams or p-th video data.

[0254] Step 502: Video stream input.

[0255] Step 503: Frame extraction from the video stream.

[0256] Step 504: Image encoding.

[0257] For example, the images obtained by extracting frames from a video stream can be encoded in the form of image data encoding as described in the foregoing embodiments.

[0258] Step 505: Obtain the vectorized representation.

[0259] Step 506: Feature extraction.

[0260] Step 507: Mask layer encoding.

[0261] Step 508: Encoding the prompt word.

[0262] Step 509, Mask Output.

[0263] For example, the result of the masking output can be the segmentation result in the foregoing embodiments.

[0264] Step 510: Target feature classification.

[0265] For example, the target feature can be the feature extraction result in the foregoing embodiments.

[0266] For example, target features can be processed by multi-classification using CNN.

[0267] Step 511, Feature archiving.

[0268] For example, a temporal feature map can be constructed through feature archiving.

[0269] Step 512: Live animal comparison.

[0270] For example, the comparison between the first living object and the second living object can be achieved by the method provided in the foregoing embodiments.

[0271] Step 513, Output the results.

[0272] For example, the result of whether the first living object and the second living object are the same object can be output.

[0273] Through the above process, based on the feature data of the living objects stored in the time-map features, the feature data of the living objects to be identified are compared in all aspects, so as to achieve comprehensive tracking of the features between different living objects from the dimension of the growth cycle of the living objects; when the above technical solution is applied to the livestock insurance scenario, it can achieve accurate identification of the identity of livestock before and after insurance.

[0274] Based on the foregoing embodiments, this application also provides a liveness matching state determination device. Figure 6 This is a schematic diagram of the structure of the liveness matching state determination device provided in the embodiments of this application, as shown below. Figure 6 As shown, the liveness matching status determination device 6 includes:

[0275] The determining module 601 is used to determine the first feature; wherein, the first feature includes a set of features of the body posture of the first living object in the first time period;

[0276] The acquisition module 602 is used to acquire the second feature; wherein, the second feature includes a feature set of the body posture of the second living object during the second time period;

[0277] The determination module 601 is also used to determine the feature correlation degree between the first feature and the second feature; based on the feature correlation degree, to determine the matching degree between the first living object in the first time period and the second living object in the second time period.

[0278] In some embodiments, the first feature includes at least the body shape, appearance, and sex characteristics of the first living object; the second feature includes at least the body shape, appearance, and sex characteristics of the second living object;

[0279] The determination module 601 is used to determine the growth parameters of the first living object; process the body shape feature, appearance feature and gender feature in the first feature based on the growth parameters to obtain the processing result; and determine the feature correlation degree based on the data associated with the body shape feature, appearance feature and gender feature in the processing result and the body shape feature, appearance feature and gender feature in the second feature.

[0280] In some embodiments, the determining module 601 is configured to determine a first data type of the feature data contained in the first feature; determine a second data type of the feature data contained in the second feature; and process the feature data contained in the first feature and the feature data contained in the second feature based on the first data type and the second data type to determine the feature correlation degree.

[0281] In some embodiments, the determining module 601 is used to determine the dot product result between the first vector corresponding to the first feature and the second vector corresponding to the second feature if both the first data type and the second data type are discrete types; and to determine the feature correlation degree based on the dot product result.

[0282] In some embodiments, the determining module 601 is configured to determine the feature similarity between the feature data contained in the first feature and the feature data contained in the second feature if both the first data type and the second data type are continuous types; and determine the feature correlation degree based on the feature similarity.

[0283] In some embodiments, the determining module 601 is used to determine the object identifier of the second living object;

[0284] The acquisition module 602 is used to acquire a second feature from the live feature data set based on the object identifier; wherein, the live feature data set includes the association between the physical characteristics of live objects in the livestock live set and the identifiers of the live objects in at least one time period.

[0285] In some embodiments, the acquisition module 602 is used to acquire an image set including the p-th live object during the m-th life period of the p-th live object; wherein the images in the image set include at least the type characteristics, body shape characteristics, appearance characteristics and sex characteristics of the p-th live object; p and m are both integers greater than or equal to 1; the p-th live object is any live object in the livestock live object set;

[0286] The determination module 601 is used to perform feature recognition on the images in the image set to obtain the feature set of the p-th living object in the m-th life period; and to integrate the feature sets of the p-th living object from the first life period to the M-th life period to obtain the feature data set of the p-th living object; where M is an integer greater than 1; and m is less than or equal to M.

[0287] The determination module 601 is further configured to determine the live feature data set based on the feature data set of the first live object to the feature data set of the Pth live object; wherein, P is an integer greater than 1, and p is less than or equal to P.

[0288] In some embodiments, the determining module 601 is used to determine cue data; based on the cue data, control a multimodal visual model to segment the images in the image set to obtain a segmentation result; identify and classify the segmentation result to obtain a classification result; wherein, the classification result includes a set of features of at least two body parts of the p-th living object;

[0289] The determination module 601 is also used to integrate the features in the classification results to obtain the feature set of the p-th living object in the m-th life period.

[0290] Based on the foregoing embodiments, this application also provides a liveness matching status determination device. Figure 7 This is a schematic diagram of the structure of the liveness matching status determination device provided in the embodiments of this application, as shown below. Figure 7 As shown, the liveness matching status determination device 7 includes a processor 701 and a memory 702; the memory 702 stores a computer program; when the computer program is executed by the processor 701, it can implement any liveness matching status determination method.

[0291] Based on the foregoing embodiments, this application also provides a computer-readable storage medium storing a computer program; when the computer program is executed by the processor of an electronic device, it can implement any of the aforementioned methods for determining the liveness matching state.

[0292] Based on the foregoing embodiments, this application also provides a computer program product, which includes a computer program; when the computer program is executed by the processor of an electronic device, it can implement any of the aforementioned methods for determining the liveness matching state.

[0293] The description of the various embodiments above tends to emphasize the differences between the various embodiments. The similarities or similarities between them can be referred to, and for the sake of brevity, they will not be repeated here.

[0294] The methods disclosed in the various method embodiments provided in this application can be arbitrarily combined to obtain new method embodiments without conflict.

[0295] The features disclosed in the various product embodiments provided in this application can be arbitrarily combined without conflict to obtain new product embodiments.

[0296] The features disclosed in the various method or device embodiments provided in this application can be arbitrarily combined without conflict to obtain new method or device embodiments.

[0297] It should be noted that the aforementioned computer-readable storage media can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), magnetic random access memory (FRAM), flash memory, magnetic surface memory, optical disc, or compact disc read-only memory (CD-ROM), etc.; or it can be various electronic devices that include one or any combination of the above-mentioned memories, such as mobile phones, computers, tablet devices, personal digital assistants, etc.

[0298] It should be noted that, in this document, 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 a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0299] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0300] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware nodes. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0301] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0302] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0303] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0304] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A method for determining the liveness matching state, characterized in that, The method includes: Determine a first feature; wherein the first feature includes a set of features of the body posture of the first living object in the first time period; Obtain the second feature; wherein the second feature includes a set of features of the body posture of the second living object during the second time period; Determine the feature correlation degree between the first feature and the second feature; Based on the aforementioned feature correlation, the degree of matching between the first living object in the first time period and the second living object in the second time period is determined.

2. The method according to claim 1, characterized in that, The first feature includes at least the body shape, appearance, and gender characteristics of the first living object; the second feature includes at least the body shape, appearance, and gender characteristics of the second living object; determining the feature correlation degree between the first feature and the second feature includes: Determine the growth parameters of the first living object; Based on the growth parameters, the body shape feature, appearance feature and gender feature in the first feature are processed to obtain the processing result; Based on the data associated with body shape features, appearance features, and gender features in the processing results, and the body shape features, appearance features, and gender features in the second feature, the feature correlation degree is determined.

3. The method according to claim 1, characterized in that, Determining the feature correlation degree between the first feature and the second feature includes: Determine the first data type of the feature data contained in the first feature; Determine the second data type of the feature data contained in the second feature; Based on the first data type and the second data type, the feature data contained in the first feature and the feature data contained in the second feature are processed to determine the feature correlation degree.

4. The method according to claim 3, characterized in that, The step of processing the feature data contained in the first feature and the feature data contained in the second feature based on the first data type and the second data type to determine the feature correlation degree includes: If both the first data type and the second data type are discrete types, determine the dot product result between the first vector corresponding to the first feature and the second vector corresponding to the second feature; The feature correlation degree is determined based on the dot product result.

5. The method according to claim 3, characterized in that, The step of processing the feature data contained in the first feature and the feature data contained in the second feature based on the first data type and the second data type to determine the feature correlation degree includes: If both the first data type and the second data type are continuous types, determine the feature similarity between the feature data contained in the first feature and the feature data contained in the second feature; The feature correlation degree is determined based on the feature similarity.

6. The method according to claim 1, characterized in that, The acquisition of the second feature includes: Determine the object identifier of the second living object; The second feature is obtained from the live animal feature data set based on the object identifier; wherein the live animal feature data set includes the association between the body shape features of live objects in the livestock live animal set during at least one time period and the identifier of the live object.

7. The method according to claim 6, characterized in that, The method further includes: During the m-th lifespan of the p-th living object, an image set including the p-th living object is collected; wherein, the images in the image set include at least the type characteristics, body shape characteristics, appearance characteristics, and sex characteristics of the p-th living object; p and m are both integers greater than or equal to 1; the p-th living object is any living object in the livestock living object set; Feature recognition is performed on the images in the image set to obtain the feature set of the p-th living object during the m-th life period; The feature sets of the p-th living object from the first life period to the M-th life period are integrated to obtain the feature data set of the p-th living object; where M is an integer greater than 1; m is less than or equal to M; The live feature data set is determined based on the feature data set of the first live object to the feature data set of the Pth live object; where P is an integer greater than 1, and p is less than or equal to P.

8. The method according to claim 7, characterized in that, The step of performing feature recognition on the images in the image set to obtain the feature set of the p-th living object during the m-th life period includes: Confirm the prompt data; Based on the aforementioned prompting data, a multimodal visual model is controlled to segment the images in the image set to obtain segmentation results; The segmentation results are identified and classified to obtain classification results; wherein, the classification results include a set of features of at least two body parts of the p-th living object; By integrating the features from the classification results, a feature set of the p-th living object during the m-th life period is obtained.

9. A device for determining the state of a living body matching, characterized in that, The live matching status determination device includes: A determining module is used to determine a first feature; wherein the first feature includes a feature set of the body posture of the first living object in a first time period; The acquisition module is used to acquire the second feature; wherein the second feature includes a feature set of the body posture of the second living object during the second time period; The determining module is further configured to determine the feature correlation degree between the first feature and the second feature; and based on the feature correlation degree, determine the matching degree between the first living object in the first time period and the second living object in the second time period.

10. A device for determining the state of a living body matching, characterized in that, The liveness matching status determination device includes a processor and a memory; the memory stores a computer program; when the computer program is executed by the processor, it can implement the liveness matching status determination method as described in any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that, The storage medium stores a computer program; when the computer program is executed by the processor of the electronic device, it can implement the liveness matching state determination method as described in any one of claims 1 to 8.

12. A computer program product, characterized in that, The program product includes a computer program; when the computer program is executed by the processor of an electronic device, it is capable of implementing the liveness matching state determination method as described in any one of claims 1 to 8.

Images