A chicken farming image monitoring method and system based on cloud computing

Abstract

The present invention relates to the field of image data processing technology, and in particular to a chicken farming image monitoring method and system based on cloud computing, the method comprising: obtaining a flock motion video representing the motion conditions of the chickens to be monitored in the chicken farm to be monitored, and extracting key frames from the flock motion video; determining the activity weakening coefficient of each chicken to be monitored between each adjacent key frame; performing key frame division according to the activity weakening coefficient of each chicken to be monitored between adjacent key frames; determining the static eigenvalue and motion eigenvalue of each chicken to be monitored in each corresponding key frame sequence; constructing a target eigenvector; clustering the target eigenvector, and performing abnormal monitoring on the chickens to be monitored based on the target clusters and discrete points. The present invention realizes the analysis of the activity behavior of the chickens by processing the key frames in the flock motion video, thereby realizing abnormal monitoring of the chickens, and improving the accuracy of abnormal monitoring of the chickens in the chicken farm.

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing, and in particular to a chicken breeding image monitoring method and system based on cloud computing. Background Art

[0002] In the chicken farming industry, if the dead chickens in the chicken farms cannot be identified and detected in time, the disease will often spread rapidly in the chicken farms, resulting in a large number of chickens getting sick in the chicken farms. Therefore, it is very important to monitor the abnormalities of chickens in the chicken farms.

[0003] At present, when monitoring the abnormality of chickens in chicken farms, the method usually adopted is: relying on human eyes and experience to judge the status of chickens in chicken farms to achieve abnormal monitoring of chickens. However, this operation often requires farmers to regularly inspect chicken farms, thereby increasing the labor intensity of farmers, and when relying on human eyes and experience to judge the status of chickens in chicken farms, the judgment results are affected by subjective factors, resulting in poor accuracy of the judgment results, which in turn leads to poor accuracy of abnormal monitoring of chickens in chicken farms.

[0004] At present, there is also a chicken abnormality monitoring method: installing cameras in chicken farms, shooting chicken videos in the chicken farms, and farmers monitor chicken abnormalities by observing and monitoring the chicken videos. However, although this operation does not require farmers to regularly inspect the chicken farms, it requires human eyes and experience to judge the status of the chickens in the chicken videos. The judgment results are also affected by subjective factors, resulting in poor accuracy in abnormal monitoring of chickens in chicken farms. Summary of the invention

[0005] In order to solve the technical problem of poor accuracy in abnormal monitoring of chickens in a chicken farm, the present invention proposes a chicken farming image monitoring method and system based on cloud computing.

[0006] In a first aspect, the present invention provides a chicken farming image monitoring method based on cloud computing, the method comprising:

[0007] Acquire a flock motion video representing the motion conditions of the chickens to be monitored in the chicken farm to be monitored, and extract key frames from the flock motion video;

[0008] According to the displacement change and area change of each monitored chicken between each adjacent key frame, the activity weakening coefficient of each monitored chicken between each adjacent key frame is determined;

[0009] Divide the key frames according to the activity weakening coefficient of each chicken to be monitored between all adjacent key frames, and obtain the key frame sequence corresponding to each chicken to be monitored;

[0010] Determine the static characteristic value of each chicken to be monitored in each key frame sequence corresponding to it according to the moving speed of each chicken to be monitored in each key frame sequence corresponding to it and the activity weakening coefficient between adjacent key frames in each key frame sequence corresponding to it;

[0011] Determine the motion feature value of each chicken to be monitored in each key frame sequence corresponding to it according to the movement of the chicken head and the change of the movement of the chicken body of each chicken to be monitored in each key frame sequence corresponding to it;

[0012] A target feature vector is formed according to the static feature value and motion feature value of each monitored chicken in each corresponding key frame sequence and the time corresponding to the corresponding key frame sequence;

[0013] All target feature vectors are clustered to obtain target clusters and discrete points, and based on the distribution of all target clusters and all discrete points, abnormal monitoring is performed on the monitored chickens.

[0014] In combination with the first aspect above, in a possible implementation, determining the activity weakening coefficient of each chicken to be monitored between each adjacent key frame according to the displacement change and area change of each chicken to be monitored between each adjacent key frame includes:

[0015] The area corresponding to each chicken to be monitored on each key frame is determined as the target area of ​​each chicken to be monitored on each key frame;

[0016] The distance between the centroids of the target areas of each monitored chicken on each adjacent key frame is determined as a displacement change index of each monitored chicken between each adjacent key frame;

[0017] The area difference between the target areas of each monitored chicken on each adjacent key frame is determined as the area change index of each monitored chicken between each adjacent key frame;

[0018] According to the displacement change index and area change index of each monitored chicken between each adjacent key frames, the activity weakening coefficient of each monitored chicken between each adjacent key frames is determined, wherein the displacement change index and the area change index are negatively correlated with the activity weakening coefficient.

[0019] In combination with the first aspect, in a possible implementation, the key frame division is performed according to the activity weakening coefficient of each chicken to be monitored between all adjacent key frames to obtain a key frame sequence corresponding to each chicken to be monitored, including:

[0020] Determine any chicken to be monitored as a marked chicken, and perform descending order on the activity weakening coefficients of the marked chicken between all adjacent key frames to obtain the activity weakening coefficient sequence corresponding to the marked chicken;

[0021] The difference between each adjacent activity weakening coefficient in the activity weakening coefficient sequence corresponding to the marked chicken is determined as the target difference;

[0022] The two activity weakening coefficients corresponding to the largest target difference are selected, and a cut is made in the middle of the two activity weakening coefficients, so that the activity weakening coefficient sequence is divided into two sub-weakening coefficient sequences;

[0023] The mean value of all active weakening coefficients in each sub-weakening coefficient sequence is determined as the sub-representative active weakening factor corresponding to each sub-weakening coefficient sequence;

[0024] Select the sub-weakening coefficient sequence with a larger corresponding sub-representative activity weakening factor from the two sub-weakening coefficient sequences as the target sub-coefficient sequence;

[0025] Taking the adjacent key frames corresponding to all active weakening coefficients in the target sub-coefficient sequence as segmentation points, all extracted key frames are segmented to obtain sub-key frame segments;

[0026] If the number of key frames in the sub-key frame segment is less than 2, the sub-key frame segment is merged with its adjacent sub-key frame segment, and the merged sub-key frame segment is determined as the key frame sequence corresponding to the marked chicken;

[0027] If the number of key frames in the sub-key frame segment is at least 2, the sub-key frame segment is determined as the key frame sequence corresponding to the marked chicken.

[0028] In combination with the first aspect, in a possible implementation, the method of determining the static feature value of each chicken to be monitored in each key frame sequence corresponding to the chicken to be monitored according to the moving speed of each chicken to be monitored in each key frame sequence corresponding to the chicken to be monitored and the activity weakening coefficient between adjacent key frames in each key frame sequence corresponding to the chicken to be monitored includes:

[0029] Determine any chicken to be monitored as a marked chicken, and determine any key frame sequence corresponding to the marked chicken as a marked key frame sequence;

[0030] Filter out a preset number of monitored chickens closest to the marked chicken from the first key frame in the marked key frame sequence as a preset number of reference chickens corresponding to the marked chicken;

[0031] The static characteristic value of each chicken to be monitored in each key frame sequence corresponding to it is determined based on the movement speed of each chicken to be monitored in each key frame sequence corresponding to it, the maximum value of the activity weakening coefficients between all adjacent key frames in each key frame sequence corresponding to it, and the preset number of reference chickens corresponding to each chicken to be monitored.

[0032] In combination with the first aspect above, in a possible implementation, the formula corresponding to the static feature value of the chicken to be monitored in its corresponding key frame sequence is:

[0033] ;

[0034] ;in, is the static feature value of the i-th chicken to be monitored in its corresponding k-th key frame sequence; i is the sequence number of the chicken to be monitored; k is the sequence number of the i-th chicken to be monitored in its corresponding key frame sequence; It is an exponential function with a natural constant as base; is the distance between the centroid of the target area on the first key frame and the centroid of the target area on the last key frame in the kth key frame sequence corresponding to the i-th monitored chicken; is the number of key frames of the i-th monitored chicken in its corresponding k-th key frame sequence; a is the sequence number of the key frame of the i-th monitored chicken in its corresponding k-th key frame sequence; represents the moving speed of the i-th monitored chicken in its corresponding k-th key frame sequence; is the maximum value of the activity weakening coefficients of the i-th monitored chicken among all adjacent key frames in its corresponding k-th key frame sequence; Characterizes the difference in activity changes between the i-th monitored chicken and its reference chicken in the k-th key frame sequence; N is a preset number; b is the serial number of the reference chicken corresponding to the i-th monitored chicken; It is the absolute value function; is the activity weakening coefficient between the ath key frame and the a+1th key frame of the i-th monitored chicken in its corresponding k-th key frame sequence; It is the activity weakening coefficient between the ath key frame and the a+1th key frame in the kth key frame sequence corresponding to the i-th monitored chicken, the b-th reference chicken corresponding to the i-th monitored chicken.

[0035] In combination with the first aspect, in a possible implementation, determining the motion feature value of each chicken to be monitored in each key frame sequence corresponding to it according to the movement of the chicken head and the change of the chicken body movement of each chicken to be monitored in each key frame sequence corresponding to it includes:

[0036] Identify a target eye point representing a chicken eye position, a target tail point representing the end point position of the chicken tail, and a target leg point representing the chicken leg position on the same side as the target eye point for each chicken to be monitored on each key frame;

[0037] Determine any chicken to be monitored as a marked chicken, and determine any key frame as a marked frame;

[0038] Connect the target chicken eye point and the target chicken tail point corresponding to the marked chicken on the marked frame to obtain the first line segment corresponding to the marked chicken on the marked frame;

[0039] Connect the target chicken eye point and the target chicken leg point corresponding to the marked chicken on the marked frame to obtain a second line segment corresponding to the marked chicken on the marked frame;

[0040] The angle between the first line segment and the second line segment corresponding to the marked chicken on the marked frame is determined as the state angle corresponding to the marked chicken on the marked frame;

[0041] The variance of the horizontal and vertical coordinate ratios of the target chicken eye points of each monitored chicken on all key frames in each key frame sequence is determined as the chicken head movement factor of each monitored chicken in each key frame sequence corresponding to it;

[0042] Determine the chicken body movement change factor of each chicken to be monitored in each key frame sequence corresponding to the chicken to be monitored according to the difference between the state angles of all adjacent key frames in each key frame sequence corresponding to the chicken to be monitored;

[0043] According to the chicken head movement factor and the chicken body movement change factor of each chicken to be monitored in each key frame sequence corresponding to it, the motion feature value of each chicken to be monitored in each key frame sequence corresponding to it is determined.

[0044] In combination with the first aspect above, in a possible implementation, the formula corresponding to the motion feature value of the chicken to be monitored in its corresponding key frame sequence is:

[0045] ;

[0046] ;in, is the motion feature value of the i-th chicken to be monitored in its corresponding k-th key frame sequence; i is the sequence number of the chicken to be monitored; k is the sequence number of the i-th chicken to be monitored in its corresponding key frame sequence; is the normalization function; is the chicken head movement factor of the i-th chicken to be monitored in its corresponding k-th key frame sequence; is the chicken body movement change factor of the i-th monitored chicken in its corresponding k-th key frame sequence; is the number of key frames of the i-th monitored chicken in its corresponding k-th key frame sequence; a is the sequence number of the key frame of the i-th monitored chicken in its corresponding k-th key frame sequence; It is the absolute value function; is the state angle of the i-th chicken to be monitored corresponding to the a-th key frame in its k-th key frame sequence; It is the state angle of the i-th monitored chicken corresponding to the a+1-th key frame in its k-th key frame sequence.

[0047] In combination with the first aspect above, in a possible implementation, the abnormal monitoring of the chicken to be monitored based on the distribution of all target clusters and all discrete points includes:

[0048] Determine the abnormal behavior index corresponding to each chicken to be monitored according to the number of discrete points in all target feature vectors corresponding to each chicken to be monitored and the total number of target clusters to which all target feature vectors corresponding to each chicken to be monitored belong;

[0049] If the abnormal behavior index corresponding to the chicken to be monitored is greater than the preset abnormal threshold, the chicken to be monitored is determined to be an abnormal behavior chicken.

[0050] In combination with the first aspect above, in a possible implementation, the formula corresponding to the abnormal behavior index of the chicken to be monitored is:

[0051] ;in, is the abnormal behavior index corresponding to the i-th chicken to be monitored; i is the serial number of the chicken to be monitored; is the normalization function; is the number of discrete points in all target feature vectors corresponding to the i-th chicken to be monitored; It is an exponential function with a natural constant as base; is the total number of target clusters to which all target feature vectors corresponding to the i-th chicken to be monitored belong; M is the total number of all target clusters; It is the cumulative value of the static eigenvalues ​​contained in all discrete points in all target eigenvectors corresponding to the i-th chicken to be monitored.

[0052] In a second aspect, the present invention provides a chicken farming image monitoring system based on cloud computing, the system comprising:

[0053] A video acquisition key frame extraction module is used to acquire a flock motion video representing the motion conditions of the chickens to be monitored in the chicken farm to be monitored, and to extract key frames from the flock motion video;

[0054] An activity weakening coefficient determination module is used to determine the activity weakening coefficient of each chicken to be monitored between each adjacent key frame according to the displacement change and area change of each chicken to be monitored between each adjacent key frame;

[0055] A key frame division module is used to divide the key frames according to the activity weakening coefficient of each chicken to be monitored between all adjacent key frames to obtain a key frame sequence corresponding to each chicken to be monitored;

[0056] A stationary feature value determination module is used to determine the stationary feature value of each chicken to be monitored in each key frame sequence corresponding to it according to the moving speed of each chicken to be monitored in each key frame sequence corresponding to it and the activity weakening coefficient between adjacent key frames in each key frame sequence corresponding to it;

[0057] A motion feature value determination module is used to determine the motion feature value of each chicken to be monitored in each key frame sequence corresponding to it according to the movement of the chicken head and the change of the movement of the chicken body of each chicken to be monitored in each key frame sequence corresponding to it;

[0058] A feature vector construction module is used to construct a target feature vector according to the static feature value and motion feature value of each monitored chicken in each corresponding key frame sequence and the time corresponding to the corresponding key frame sequence;

[0059] The cluster monitoring module is used to cluster all target feature vectors to obtain target clusters and discrete points, and based on the distribution of all target clusters and all discrete points, perform abnormal monitoring on the monitored chickens.

[0060] In a third aspect, a server is provided, comprising a memory and a processor. The memory is used to store executable program code, and the processor is used to call and run the executable program code from the memory, so that the device executes the method in the first aspect or any possible implementation of the first aspect.

[0061] In a fourth aspect, a computer program product is provided, comprising: a computer program code, which, when executed on a computer, enables the computer to execute the method in the first aspect or any possible implementation of the first aspect.

[0062] In a fifth aspect, a computer-readable storage medium is provided, which stores a computer program code. When the computer program code runs on a computer, the computer executes the method in the above-mentioned first aspect or any possible implementation manner of the first aspect.

[0063] The present invention has the following beneficial effects:

[0064] The present invention discloses a chicken farming image monitoring method based on cloud computing. By processing key frames in a chicken flock motion video, the chicken activity behavior analysis is realized, thereby realizing chicken abnormal monitoring, solving the technical problem of poor accuracy in abnormal monitoring of chickens in a chicken farm, and improving the accuracy of abnormal monitoring of chickens in a chicken farm. Compared with judging the state of chickens in a chicken farm by relying on human eyes and experience, the present invention is based on cloud computing. By analyzing key frames in a chicken flock motion video, multiple factors related to the chicken activity behavior state are quantified, such as activity weakening coefficient, static eigenvalue and motion eigenvalue, thereby quantifying the target feature vector representing the activity behavior state of the monitored chicken in the time period corresponding to different key frame sequences, and clustering all target feature vectors, which can facilitate the subsequent analysis of the aggregation of target feature vectors representing different activity behavior states, thereby realizing relatively objective abnormal monitoring of the monitored chicken, reducing the influence of human subjective factors to a certain extent, thereby improving the accuracy of abnormal monitoring of the monitored chicken. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0066] Figure 1 It is a flow chart of a chicken farming image monitoring method based on cloud computing of the present invention;

[0067] Figure 2 A schematic diagram of the composition structure of a chicken farming image monitoring system based on cloud computing of the present invention;

[0068] Figure 3 The figure is a schematic diagram of the structure of a computer device of the present invention. DETAILED DESCRIPTION

[0069] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the technical solutions proposed by the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0070] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0071] refer to Figure 1 , showing the process of some embodiments of a chicken farming image monitoring method based on cloud computing of the present invention. The chicken farming image monitoring method based on cloud computing includes the following steps:

[0072] Step S1, obtaining a flock motion video representing the motion conditions of the chickens to be monitored in the chicken farm to be monitored, and extracting key frames from the flock motion video.

[0073] The chicken farm to be monitored may be a chicken farm to be monitored for abnormality. The chickens to be monitored may be chickens in the chicken farm to be monitored. The chicken movement video may be a video representing the movement of chickens in the chicken farm to be monitored.

[0074] It should be noted that the chickens in the chicken farm to be monitored can be located on the same horizontal plane. Generally speaking, chicken farms are often composed of a single-layer structure or a multi-layer structure. If a chicken farm is composed of a single-layer structure, the chicken farm can be directly recorded as the chicken farm to be monitored. If a chicken farm is composed of a multi-layer structure, each layer in the chicken farm can be recorded as the chicken farm to be monitored.

[0075] As an example, this step may include the following steps:

[0076] The first step is to obtain a flock movement video representing the movement of the chickens to be monitored in the chicken farm to be monitored.

[0077] For example, a camera installed in a chicken farm to be monitored can be used to collect a video of a flock of chickens in the chicken farm to be monitored during the current time period as a flock of chickens motion video. The end time of the current time period can be the current time. The duration corresponding to the current time period can be 1 hour. Each frame of the flock of chickens motion video can contain all the chickens in the chicken farm to be monitored.

[0078] It should be noted that if the space of the chicken farm to be monitored is large and one camera cannot capture the entire space, multiple cameras can be used to shoot and then spliced ​​together, so that each frame of the final video contains the entire space of the chicken farm to be monitored, and the final video is recorded as a chicken flock motion video. Each chicken to be monitored in each frame of the chicken flock motion video can be marked by the YOLOV5 (convolutional neural network) model. The same chicken to be monitored in the chicken flock motion video can be identified by image tracking technology.

[0079] In the second step, the 3D-CNN (3D Convolutional Neural Network) key frame extraction algorithm can be used to extract key frames from the chicken movement video to obtain multiple key frames.

[0080] The key frames in the embodiment of the present invention may be sequential, and the sequence may be the sequence of acquisition time.

[0081] It should be noted that, compared with analyzing all frame images, the embodiment of the present invention analyzes the extracted key frames, which can reduce the subsequent calculation amount to a certain extent.

[0082] Step S2, determining the activity weakening coefficient of each chicken to be monitored between each adjacent key frame according to the displacement change and area change of each chicken to be monitored between each adjacent key frame.

[0083] It should be noted that the adjacent key frames in the embodiment of the present invention may be key frames with adjacent acquisition times.

[0084] As an example, this step may include the following steps:

[0085] In the first step, the area corresponding to each chicken to be monitored on each key frame is determined as the target area of ​​each chicken to be monitored on each key frame.

[0086] The target area may be the area where the chicken to be monitored is located on the key frame.

[0087] In the second step, the distance between the centroids of the target areas of each monitored chicken on each adjacent key frame is determined as the displacement change index of each monitored chicken between each adjacent key frame.

[0088] The distance between the centroids of the target areas on adjacent key frames may be equal to the distance between the pixel coordinates corresponding to the centroids of the target areas on adjacent key frames. The pixel coordinates corresponding to the centroids may be the coordinates of the centroids in the target coordinate system corresponding to the centroids. The target coordinate system corresponding to the centroids may be a coordinate system having the lower left end point of the key frame to which the centroids belong as the origin, the horizontal direction of the key frame to which the centroids belong as the horizontal axis, and the vertical direction of the key frame to which the centroids belong as the vertical axis.

[0089] In the third step, the area difference between the target areas of each monitored chicken on each adjacent key frame is determined as the area change index of each monitored chicken between each adjacent key frame.

[0090] The fourth step is to determine the activity weakening coefficient of each monitored chicken between each adjacent key frame according to the displacement change index and area change index of each monitored chicken between each adjacent key frame.

[0091] Among them, both the displacement change index and the area change index can be negatively correlated with the activity weakening coefficient.

[0092] For example, the formula for determining the activity weakening coefficient of the chicken to be monitored between adjacent key frames may be:

[0093] ;in, is the activity weakening coefficient of the i-th monitored chicken between the j-th key frame and the j+1-th key frame. i is the sequence number of the monitored chicken. j is the sequence number of the key frame, which is obtained by sorting all the key frames in the order of the acquisition time. It is an exponential function with a natural constant as its base. It is the displacement change index of the i-th monitored chicken between the j-th key frame and the j+1-th key frame. It is the absolute value function. is the area of ​​the target region of the i-th monitored chicken in the j-th key frame, which can be represented by the number of pixels in the target region of the i-th monitored chicken in the j-th key frame. is the area of ​​the target region of the i-th monitored chicken in the j-th key frame, which can be represented by the number of pixels in the target region of the i-th monitored chicken in the j+1-th key frame. It is the area change index of the i-th monitored chicken between the j-th key frame and the j+1-th key frame.

[0094] It should be noted that in reality, sick chickens tend to remain motionless for a long time and have poor activity. For example, chickens with Marek's disease tend to remain motionless for a long time. When the value is larger, it often indicates that the displacement change of the i-th monitored chicken between the j-th key frame and the j+1-th key frame is larger, which often indicates that the i-th monitored chicken is more likely to walk during the time period between the j-th key frame and the j+1-th key frame. The smaller it is, the greater the change in area of ​​the i-th monitored chicken between the j-th key frame and the j+1-th key frame, which means that the i-th monitored chicken is more likely to change its shape during the period between the j-th key frame and the j+1-th key frame, for example, it may open or close its wings. When is larger, it often indicates that the i-th monitored chicken is more likely to maintain the same state during the time period between the j-th key frame and the j+1-th key frame, and it often indicates that the activity level of the i-th monitored chicken is worse during the time period between the j-th key frame and the j+1-th key frame.

[0095] Step S3, dividing the key frames according to the activity weakening coefficient of each chicken to be monitored between all adjacent key frames, and obtaining a key frame sequence corresponding to each chicken to be monitored.

[0096] Among them, a key frame sequence corresponding to the chicken to be monitored can represent a behavioral stage of the chicken to be monitored.

[0097] As an example, this step may include the following steps:

[0098] In the first step, any chicken to be monitored is determined as a marked chicken, and the activity weakening coefficients of the marked chicken between all adjacent key frames are sorted in descending order to obtain the activity weakening coefficient sequence corresponding to the marked chicken.

[0099] In the second step, the difference between each adjacent activity weakening coefficient in the activity weakening coefficient sequence corresponding to the above-mentioned marked chicken is determined as the target difference.

[0100] Among them, the target difference is a non-negative number. Since one target difference is calculated from two activity weakening coefficients, one target difference corresponds to two activity weakening coefficients.

[0101] The third step is to select two activity weakening coefficients corresponding to the largest target difference, and cut in the middle of the two activity weakening coefficients to divide the activity weakening coefficient sequence into two sub-weakening coefficient sequences.

[0102] For example, if the activity weakening coefficient sequence is {0.7, 0.6, 0.5, 0.3, 0.2, 0.1}, then the two activity weakening coefficients corresponding to the maximum target difference are 0.5 and 0.3 respectively, and the two sub-weakening coefficient sequences obtained by segmentation can be {0.7, 0.6, 0.5} and {0.3, 0.2, 0.1} respectively.

[0103] It should be noted that, if there are multiple maximum target differences listed in parallel, one may be randomly selected from the multiple target differences listed in parallel and recorded as the maximum target difference required by the embodiment of the present invention.

[0104] In the fourth step, the mean value of all active weakening coefficients in each sub-weakening coefficient sequence is determined as the sub-representative active weakening factor corresponding to each sub-weakening coefficient sequence.

[0105] The fifth step is to select the sub-weakening coefficient sequence with a larger corresponding sub-representative activity weakening factor from the two sub-weakening coefficient sequences as the target sub-coefficient sequence.

[0106] In the sixth step, the adjacent key frames corresponding to all active weakening coefficients in the target sub-coefficient sequence are used as segmentation points to segment all the extracted key frames to obtain sub-key frame segments.

[0107] Among them, there are at least two key frames in the key frame sequence corresponding to the marked chicken.

[0108] In the seventh step, if the number of key frames in the sub-key frame segment is less than 2, the sub-key frame segment is merged with its adjacent sub-key frame segment, and the merged sub-key frame segment is determined as the key frame sequence corresponding to the above-mentioned marked chicken.

[0109] In the eighth step, if the number of key frames in the sub-key frame segment is at least 2, the sub-key frame segment is determined as the key frame sequence corresponding to the above-mentioned marked chicken.

[0110] For example, if the target sub-coefficient sequence is {0.7, 0.6, 0.5}, there are 10 key frames, which are {key frame 1, key frame 2, key frame 3, key frame 4, key frame 5, key frame 6, key frame 7, key frame 8, key frame 9, key frame 10}, and the two key frames corresponding to the activity weakening coefficient of 0.7 are key frame 1 and key frame 2, the two key frames corresponding to the activity weakening coefficient of 0.6 are key frame 4 and key frame 5, and the two key frames corresponding to the activity weakening coefficient of 0.5 are key frame 7 and key frame 8, then the sub-key frame sequence is {0.7, 0.6, 0.5}, and the two key frames corresponding to the activity weakening coefficient are key frame 7 and key frame 8. There can be 4 frame segments, namely {key frame 1}, {key frame 2, key frame 3, key frame 4}, {key frame 5, key frame 6, key frame 7} and {key frame 8, key frame 9, key frame 10}; since {key frame 1} contains only one key frame, it needs to be merged with its adjacent sub-key frame segments, so we can finally get 3 key frame sequences corresponding to the marked chicken, which can be {key frame 1, key frame 2, key frame 3, key frame 4}, {key frame 5, key frame 6, key frame 7} and {key frame 8, key frame 9, key frame 10} respectively.

[0111] Step S4, determining the static feature value of each monitored chicken in each corresponding key frame sequence according to the moving speed of each monitored chicken in each corresponding key frame sequence and the activity weakening coefficient between adjacent key frames in each corresponding key frame sequence.

[0112] As an example, this step may include the following steps:

[0113] In the first step, any chicken to be monitored is determined as a marked chicken, and any key frame sequence corresponding to the marked chicken is determined as a marked key frame sequence.

[0114] In the second step, a preset number of monitored chickens closest to the marked chicken are selected from the first key frame in the marked key frame sequence as the preset number of reference chickens corresponding to the marked chicken.

[0115] The preset number may be a preset number of adjacent chickens, and the preset number may be 6.

[0116] The third step is to determine the static characteristic value of each monitored chicken in each corresponding key frame sequence based on the movement speed of each monitored chicken in each corresponding key frame sequence, the maximum value of the activity weakening coefficients between all adjacent key frames in each corresponding key frame sequence, and the preset number of reference chickens corresponding to each monitored chicken.

[0117] For example, the formula for determining the static feature value of the chicken to be monitored in its corresponding key frame sequence can be:

[0118] ;

[0119] ;in, is the static feature value of the i-th monitored chicken in its corresponding k-th key frame sequence. i is the sequence number of the monitored chicken. k is the sequence number of the i-th monitored chicken in its corresponding key frame sequence. It is an exponential function with a natural constant as its base. It is the distance between the centroid of the target area on the first key frame and the centroid of the target area on the last key frame in the kth key frame sequence corresponding to the i-th chicken to be monitored; it can represent the movement displacement of the i-th chicken to be monitored in the time period corresponding to the kth key frame sequence corresponding to the i-th chicken to be monitored. is the number of key frames of the i-th chicken to be monitored in its corresponding k-th key frame sequence. a is the sequence number of the key frame of the i-th chicken to be monitored in its corresponding k-th key frame sequence. It represents the moving speed of the i-th chicken to be monitored in its corresponding k-th key frame sequence. It is the maximum value of the activity weakening coefficients of the i-th monitored chicken among all adjacent key frames in its corresponding k-th key frame sequence. Characterizes the difference in activity changes between the i-th monitored chicken and its reference chicken in the k-th key frame sequence. N is a preset number. b is the serial number of the reference chicken corresponding to the i-th monitored chicken. It is the absolute value function. is the activity weakening coefficient between the ath key frame and the a+1th key frame in the kth key frame sequence corresponding to the i-th monitored chicken. It is the activity weakening coefficient between the ath key frame and the a+1th key frame in the kth key frame sequence corresponding to the i-th monitored chicken, the b-th reference chicken corresponding to the i-th monitored chicken.

[0120] It should be noted that when When the value is larger, it often indicates that the i-th chicken to be monitored has a faster moving speed in the time period corresponding to the k-th key frame sequence, and the i-th chicken to be monitored has a higher activity level in the time period corresponding to the k-th key frame sequence. The larger is the value, the more likely it is that the i-th chicken to be monitored has maintained the same state during the time period corresponding to its corresponding k-th key frame sequence. When it is larger, it often means that the activity level of the i-th monitored chicken is smaller during the time period corresponding to its corresponding k-th key frame sequence, which often means that the i-th monitored chicken is more likely to be sick during the time period corresponding to its corresponding k-th key frame sequence.

[0121] In reality, because chickens are gregarious animals, according to the distance dependence of the flock law, when chickens that are closer to each other perform certain behaviors, they tend to have certain interactive effects on the surrounding chickens, while chickens that are farther away from each other tend to break away from the flock. In other words, under normal circumstances, the behaviors of adjacent chickens are often consistent, and if the behavior of a chicken is inconsistent with that of its neighbors, it often indicates that the chicken may be abnormal, that is, it may be sick. When the value is larger, it often indicates that the behavior change of the i-th monitored chicken in the time period corresponding to its corresponding k-th key frame sequence is more inconsistent with the behavior change of its neighboring chickens, which often indicates that the i-th monitored chicken is more likely to be sick in the time period corresponding to its corresponding k-th key frame sequence. The larger the value is, the more likely it is that the i-th chicken to be monitored is sick during the time period corresponding to its corresponding k-th key frame sequence.

[0122] Step S5, determining the motion feature value of each monitored chicken in each corresponding key frame sequence according to the movement of the chicken head and the change of the chicken body movement of each monitored chicken in each corresponding key frame sequence.

[0123] As an example, this step may include the following steps:

[0124] The first step is to identify a target eye point representing the position of an eye, a target tail point representing the end point of the tail, and a target leg point representing the position of the leg on the same side as the target eye point for each chicken to be monitored in each key frame.

[0125] The position of the eye can be represented by the position of the center of the eye, the position of the end point of the tail can be represented by the position of the highest point of the tail, and the position of the drumstick can be represented by the lowest point of the drumstick.

[0126] For example, the key frame can be input into a pre-trained target recognition network, and the target recognition network can identify a position point representing any eye position of the chicken to be monitored from the key frame as the target eye point; the target recognition network can identify a position point representing the end point of the tail of the chicken to be monitored from the key frame as the target tail point, and the target recognition network can identify a position point representing the chicken leg position on the same side as the target eye point from the key frame as the target leg point. The target recognition network can be a CNN (Convolutional Neural Networks) for identifying eyes, tails, and legs.

[0127] Optionally, the training process of the target recognition network may include the following sub-steps:

[0128] The first sub-step is to construct a CNN as a pre-training object recognition network.

[0129] In the second sub-step, a plurality of historical images of chicken flocks with the tails, ipsilateral eyes and drumsticks labeled are obtained.

[0130] The historical chicken flock images may be images of chicken flocks in a chicken farm taken historically.

[0131] It should be noted that, in general, the eyes, drumsticks and tails of chickens in chicken farms can often be shown in the images taken. In this case, the eyes, drumsticks and tails can be directly marked on the images. For chickens that are blocked by other chickens and cannot be fully shown in the images, the angle of shooting can be changed during monitoring to show the eyes, drumsticks and tails. It is also possible to evaluate the positions of the eyes, drumsticks and tails through manual experience and mark them.

[0132] In the third sub-step, a target recognition network is trained using multiple historical chicken images as a training set and the positions of the annotated chicken tails, ipsilateral chicken eyes and chicken legs as training labels to obtain a trained target recognition network.

[0133] Optionally, since the corn eye is approximately circular, the target corn eye point can also be obtained by the Hough circle transform method.

[0134] In the second step, any chicken to be monitored is determined as a marked chicken, and any key frame is determined as a marked frame.

[0135] The third step is to connect the target chicken eye point and the target chicken tail point corresponding to the marked chicken on the marked frame to obtain the first line segment corresponding to the marked chicken on the marked frame.

[0136] The fourth step is to connect the target chicken eye point and the target chicken leg point corresponding to the marked chicken on the marked frame to obtain the second line segment corresponding to the marked chicken on the marked frame.

[0137] The fifth step is to determine the angle between the first line segment and the second line segment corresponding to the marked chicken on the marked frame as the state angle corresponding to the marked chicken on the marked frame.

[0138] In the sixth step, the variance of the horizontal and vertical coordinate ratios of the target chicken eye points of each monitored chicken on all key frames in each key frame sequence is determined as the chicken head movement factor of each monitored chicken in each key frame sequence.

[0139] Among them, the method for obtaining the ratio of the horizontal and vertical coordinates corresponding to the target corn point can be: the ratio of the horizontal coordinate to the vertical coordinate included in the pixel coordinates corresponding to the target corn point can be determined as the ratio of the horizontal and vertical coordinates corresponding to the target corn point. The pixel coordinates corresponding to the target corn point can be the coordinates of the target corn point in its corresponding target coordinate system. The target coordinate system corresponding to the target corn point can be a coordinate system with the lower left end point of the key frame to which the target corn point belongs as the origin, the horizontal direction of the key frame to which the target corn point belongs as the horizontal axis, and the vertical direction of the key frame to which the target corn point belongs as the vertical axis.

[0140] The seventh step is to determine the chicken body movement change factor of each monitored chicken in each key frame sequence corresponding to it according to the difference between the state angles of each monitored chicken in all adjacent key frames in each key frame sequence.

[0141] In the eighth step, the motion feature value of each monitored chicken in each corresponding key frame sequence is determined according to the chicken head movement factor and the chicken body movement change factor of each monitored chicken in each corresponding key frame sequence.

[0142] For example, the formula for determining the motion feature value of the chicken to be monitored in its corresponding key frame sequence may be:

[0143] ;

[0144] ;in, is the motion feature value of the i-th chicken to be monitored in its corresponding k-th key frame sequence. i is the sequence number of the chicken to be monitored. k is the sequence number of the i-th chicken to be monitored in its corresponding key frame sequence. is a normalization function. is the chicken head movement factor of the i-th chicken to be monitored in its corresponding k-th key frame sequence. It is the chicken body movement change factor of the i-th monitored chicken in its corresponding k-th key frame sequence. is the number of key frames of the i-th chicken to be monitored in its corresponding k-th key frame sequence. a is the sequence number of the key frame of the i-th chicken to be monitored in its corresponding k-th key frame sequence. It is the absolute value function. It is the state angle of the i-th chicken to be monitored corresponding to the a-th key frame in its k-th key frame sequence. It is the state angle of the i-th monitored chicken corresponding to the a+1-th key frame in its k-th key frame sequence.

[0145] It should be noted that when The larger the value is, the greater the movement of the head of the i-th chicken to be monitored is during the time period corresponding to the k-th key frame sequence. The larger the value is, the greater the degree of movement of the i-th chicken to be monitored is during the time period corresponding to the k-th key frame sequence. When it is larger, it often means that the i-th chicken to be monitored has a greater degree of movement during the time period corresponding to its corresponding k-th key frame sequence, which often means that the i-th chicken to be monitored is more likely to be not sick during the time period corresponding to its corresponding k-th key frame sequence.

[0146] Step S6, constructing a target feature vector according to the static feature value and motion feature value of each monitored chicken in each corresponding key frame sequence and the time corresponding to the corresponding key frame sequence.

[0147] The time corresponding to the key frame sequence can be represented by the middle moment of the time period corresponding to the key frame sequence. For example, if the start time of the time period corresponding to the key frame sequence is 15:54:00 on October 28, 2024, and the end time is 15:54:10 on October 28, 2024, then the time corresponding to the key frame sequence can be 15:54:05 on October 28, 2024.

[0148] As an example, first, any chicken to be monitored can be determined as a marked chicken, and any key frame sequence corresponding to the marked chicken can be determined as a marked key frame sequence. Then, the static feature value and motion feature value of the marked chicken in the marked key frame sequence, as well as the time corresponding to the marked key frame sequence, can be used to form a target feature vector, which is a target feature vector corresponding to the marked chicken.

[0149] Step S7, clustering all target feature vectors to obtain target clusters and discrete points, and performing abnormal monitoring on the monitored chickens based on the distribution of all target clusters and all discrete points.

[0150] The discrete points represent discrete target feature vectors. The target feature vectors in the same target cluster are relatively similar.

[0151] As an example, this step may include the following steps:

[0152] In the first step, all target feature vectors are clustered to obtain target clusters and discrete points.

[0153] For example, all target feature vectors can be clustered using the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm, and the clusters obtained by clustering are used as target clusters, and discrete target feature vectors that are not clustered are recorded as discrete points, where the clustering radius can be 0.1, and the number of target feature vectors within the radius can be 10% of the number of chickens to be monitored.

[0154] In the second step, the abnormal behavior index corresponding to each chicken to be monitored is determined according to the number of discrete points in all target feature vectors corresponding to each chicken to be monitored and the total number of target clusters to which all target feature vectors corresponding to each chicken to be monitored belong.

[0155] For example, the formula for determining the abnormal behavior index corresponding to the chicken to be monitored can be:

[0156] ;in, is the abnormal behavior index corresponding to the i-th chicken to be monitored. i is the serial number of the chicken to be monitored. is a normalization function. is the number of discrete points in all target feature vectors corresponding to the i-th chicken to be monitored. It is an exponential function with a natural constant as its base. is the total number of target clusters to which all target feature vectors corresponding to the i-th chicken to be monitored belong. M is the total number of all target clusters. It is the cumulative value of the static eigenvalues ​​contained in all discrete points in all target eigenvectors corresponding to the i-th chicken to be monitored.

[0157] It should be noted that based on the flock law, discrete points can represent abnormal activities, and target clusters can represent relatively clustered normal activities. When is larger, it often means that the i-th chicken to be monitored has more discrete points, which often means that the i-th chicken to be monitored has more abnormal activities, which often means that the i-th chicken to be monitored is more likely to be sick. When the value is larger, it often means that the i-th chicken to be monitored covers more different normal behaviors, which often means that the i-th chicken to be monitored is more likely to be free of disease. The larger is the value, the more likely the i-th chicken to be monitored is to be sick. The larger it is, the more likely it is that the i-th chicken to be monitored is sick.

[0158] In the third step, if the abnormal behavior index corresponding to the chicken to be monitored is greater than the preset abnormal threshold, the chicken to be monitored is determined as an abnormal behavior chicken.

[0159] The preset abnormal threshold may be a preset threshold for determining whether the chicken's behavior is abnormal. For example, the preset abnormal threshold may be 0.7.

[0160] It should be noted that chickens with abnormal behavior may represent suspected sick chickens. Through marking and early warning, management personnel can be notified in time to re-examine these suspected sick chickens to avoid the widespread spread of poultry diseases.

[0161] refer to Figure 2 Based on the same inventive concept as the above method embodiment, the present invention provides a chicken farming image monitoring system based on cloud computing, the system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the above computer program is executed by the processor, the steps of implementing a chicken farming image monitoring method based on cloud computing may specifically include:

[0162] The video acquisition key frame extraction module 201 is used to acquire a flock motion video representing the motion conditions of the chickens to be monitored in the chicken farm to be monitored, and to extract key frames from the flock motion video;

[0163] An activity weakening coefficient determination module 202 is used to determine the activity weakening coefficient of each chicken to be monitored between each adjacent key frame according to the displacement change and area change of each chicken to be monitored between each adjacent key frame;

[0164] The key frame division module 203 is used to divide the key frames according to the activity weakening coefficient of each chicken to be monitored between all adjacent key frames to obtain a key frame sequence corresponding to each chicken to be monitored;

[0165] The static feature value determination module 204 is used to determine the static feature value of each chicken to be monitored in each key frame sequence corresponding to it according to the moving speed of each chicken to be monitored in each key frame sequence corresponding to it and the activity weakening coefficient between adjacent key frames in each key frame sequence corresponding to it;

[0166] The motion feature value determination module 205 is used to determine the motion feature value of each chicken to be monitored in each key frame sequence corresponding to it according to the movement of the chicken head and the change of the chicken body movement of each chicken to be monitored in each key frame sequence corresponding to it;

[0167] The feature vector construction module 206 is used to construct a target feature vector according to the static feature value and motion feature value of each monitored chicken in each corresponding key frame sequence and the time corresponding to the corresponding key frame sequence;

[0168] The cluster monitoring module 207 is used to cluster all target feature vectors to obtain target clusters and discrete points, and perform abnormal monitoring on the monitored chickens based on the distribution of all target clusters and all discrete points.

[0169] Figure 3 is a schematic diagram of the structure of a computer device provided by an embodiment of the present invention. Figure 3 As shown, the computer device 300 includes: a memory 301, a processor 302, and a computer program 303 stored in the memory 301 and running on the processor 302, wherein when the processor 302 executes the computer program 303, the computer device can execute any one of the cloud computing-based chicken farming image monitoring methods introduced above.

[0170] Based on the same inventive concept as the above method embodiment, the present invention provides a server, including a memory and a processor. The memory is used to store executable program code, and the processor is used to call and run the executable program code from the memory, so that the device executes any of the above-mentioned chicken farming image monitoring methods based on cloud computing.

[0171] Based on the same inventive concept as the above-mentioned method embodiment, the present invention provides a computer program product, which includes: computer program code, when the computer program code runs on a computer, enables the computer to execute any one of the above-mentioned cloud computing-based chicken farming image monitoring methods.

[0172] Based on the same inventive concept as the above-mentioned method embodiment, the present invention provides a computer-readable storage medium, which stores a computer program code. When the computer program code runs on a computer, the computer executes any one of the above-mentioned cloud computing-based chicken farming image monitoring methods.

[0173] In summary, compared to relying on human eyes and experience to judge the status of chickens in a chicken farm, the present invention is based on cloud computing, and by analyzing key frames in the chicken movement video, multiple factors related to the activity behavior status of chickens are quantified, such as activity attenuation coefficient, static eigenvalue and motion eigenvalue, thereby quantifying the target feature vector that characterizes the activity behavior status of the monitored chickens in the time period corresponding to different key frame sequences. Clustering of all target feature vectors can facilitate subsequent analysis of the clustering of target feature vectors characterizing different activity behavior states, thereby relatively objectively realizing abnormal monitoring of the monitored chickens, reducing the influence of human subjective factors to a certain extent, and thus improving the accuracy of abnormal monitoring of the monitored chickens.

[0174] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features can be replaced by equivalents. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A chicken farming image monitoring method based on cloud computing, characterized in that: The following steps are involved: Acquire a flock motion video representing the motion conditions of the chickens to be monitored in the chicken farm to be monitored, and extract key frames from the flock motion video; According to the displacement change and area change of each monitored chicken between each adjacent key frame, the activity weakening coefficient of each monitored chicken between each adjacent key frame is determined; Divide the key frames according to the activity weakening coefficient of each chicken to be monitored between all adjacent key frames, and obtain the key frame sequence corresponding to each chicken to be monitored; Determine the static characteristic value of each chicken to be monitored in each key frame sequence corresponding to it according to the moving speed of each chicken to be monitored in each key frame sequence corresponding to it and the activity weakening coefficient between adjacent key frames in each key frame sequence corresponding to it; Determine the motion feature value of each chicken to be monitored in each key frame sequence corresponding to it according to the movement of the chicken head and the change of the movement of the chicken body of each chicken to be monitored in each key frame sequence corresponding to it; A target feature vector is formed according to the static feature value and motion feature value of each monitored chicken in each corresponding key frame sequence and the time corresponding to the corresponding key frame sequence; Cluster all target feature vectors to obtain target clusters and discrete points, and perform abnormal monitoring on the monitored chickens based on the distribution of all target clusters and all discrete points; Determining the activity weakening coefficient of each chicken to be monitored between each adjacent key frame according to the displacement change and area change of each chicken to be monitored between each adjacent key frame includes: The area corresponding to each chicken to be monitored on each key frame is determined as the target area of ​​each chicken to be monitored on each key frame; The distance between the centroids of the target areas of each monitored chicken on each adjacent key frame is determined as the displacement change index of each monitored chicken between each adjacent key frame; The area difference between the target areas of each monitored chicken on each adjacent key frame is determined as the area change index of each monitored chicken between each adjacent key frame; According to the displacement change index and area change index of each monitored chicken between each adjacent key frame, the activity weakening coefficient of each monitored chicken between each adjacent key frame is determined, wherein the displacement change index and the area change index are both negatively correlated with the activity weakening coefficient; The method of determining the static feature value of each chicken to be monitored in each key frame sequence corresponding to the chicken to be monitored according to the moving speed of each chicken to be monitored in each key frame sequence corresponding to the chicken to be monitored and the activity weakening coefficient between adjacent key frames in each key frame sequence corresponding to the chicken to be monitored comprises: Determine any chicken to be monitored as a marked chicken, and determine any key frame sequence corresponding to the marked chicken as a marked key frame sequence; Filter out a preset number of monitored chickens closest to the marked chicken from the first key frame in the marked key frame sequence as a preset number of reference chickens corresponding to the marked chicken; Determine the static feature value of each chicken to be monitored in each key frame sequence corresponding to it according to the moving speed of each chicken to be monitored in each key frame sequence corresponding to it, the maximum value of the activity weakening coefficients between all adjacent key frames in each key frame sequence corresponding to it, and the preset number of reference chickens corresponding to each chicken to be monitored; The formula corresponding to the static characteristic value of the monitored chicken in its corresponding key frame sequence is: ； ;in, It is i Just wait for the monitoring chicken to be in its corresponding k The stationary feature values ​​in a sequence of key frames; i is the serial number of the chicken to be monitored; k It is i Only the sequence number of the chicken in its corresponding key frame sequence is to be monitored; It is an exponential function with a natural constant as base; It is i Just wait for the monitoring chicken to be in its corresponding k The distance between the centroid of the target area on the first key frame and the centroid of the target area on the last key frame in a sequence of key frames; It is i Just wait for the monitoring chicken to be in its corresponding k The number of keyframes in a keyframe sequence; a It is i Just wait for the monitoring chicken to be in its corresponding k The sequence number of the key frame in the key frame sequence; Indicates i Just wait for the monitoring chicken to be in its corresponding k The speed of movement in a sequence of key frames; It is i Just wait for the monitoring chicken to be in its corresponding k The maximum value of the activity weakening coefficients between all adjacent key frames in a key frame sequence; Characterization i The monitored chickens and their reference chickens are k The difference in activity changes in a sequence of key frames; N is the preset quantity; b It is i The serial number of the reference chicken corresponding to the chicken to be monitored; It is the absolute value function; It is i Just wait for the monitoring chicken to be in its corresponding k In the key frame sequence, the a keyframes and a +1 activity weakening factor between keyframes; It is i Only the corresponding b Only refer to the chicken, in i Only the corresponding k In the key frame sequence, the a keyframes and a +1 activity weakening factor between keyframes; The method of determining the motion feature value of each chicken to be monitored in each key frame sequence corresponding to the chicken to be monitored according to the movement of the chicken head and the change of the movement of the chicken body in each key frame sequence corresponding to the chicken to be monitored includes: Identify a target eye point representing a chicken eye position, a target tail point representing the end point position of the chicken tail, and a target leg point representing the chicken leg position on the same side as the target eye point for each chicken to be monitored on each key frame; Determine any chicken to be monitored as a marked chicken, and determine any key frame as a marked frame; Connect the target chicken eye point and the target chicken tail point corresponding to the marked chicken on the marked frame to obtain the first line segment corresponding to the marked chicken on the marked frame; Connect the target chicken eye point and the target chicken leg point corresponding to the marked chicken on the marked frame to obtain a second line segment corresponding to the marked chicken on the marked frame; The angle between the first line segment and the second line segment corresponding to the marked chicken on the marked frame is determined as the state angle corresponding to the marked chicken on the marked frame; The variance of the horizontal and vertical coordinate ratios of the target chicken eye points of each monitored chicken on all key frames in each key frame sequence is determined as the chicken head movement factor of each monitored chicken in each key frame sequence corresponding to it; Determine the chicken body movement change factor of each chicken to be monitored in each key frame sequence corresponding to the chicken to be monitored according to the difference between the state angles of all adjacent key frames in each key frame sequence corresponding to the chicken to be monitored; Determine the motion feature value of each chicken to be monitored in each key frame sequence corresponding to it according to the chicken head movement factor and the chicken body movement change factor of each chicken to be monitored in each key frame sequence corresponding to it; The formula corresponding to the motion feature value of the chicken to be monitored in its corresponding key frame sequence is: ； ;in, It is i Just wait for the monitoring chicken to be in its corresponding k Motion feature values ​​in a key frame sequence; i is the serial number of the chicken to be monitored; k It is i Only the sequence number of the chicken in its corresponding key frame sequence is to be monitored; is the normalization function; It is i Just wait for the monitoring chicken to be in its corresponding k The chicken head movement factor in the key frame sequence; It is i Just wait for the monitoring chicken to be in its corresponding k The chicken body movement variation factor in a keyframe sequence; It is i Just wait for the monitoring chicken to be in its corresponding k The number of keyframes in a keyframe sequence; a It is i Just wait for the monitoring chicken to be in its corresponding k The sequence number of the key frame in the key frame sequence; It is the absolute value function; It is i Only the chicken to be monitored corresponds to its k The first key frame in the sequence a The state angle on the key frame; It is i Only the chicken to be monitored corresponds to its k The first key frame in the sequence a +1 state angle at keyframe.

2. The method for monitoring chicken farming images based on cloud computing according to claim 1, characterized in that: The key frame division is performed according to the activity weakening coefficient of each chicken to be monitored between all adjacent key frames to obtain a key frame sequence corresponding to each chicken to be monitored, including: Determine any chicken to be monitored as a marked chicken, and perform descending order on the activity weakening coefficients of the marked chicken between all adjacent key frames to obtain the activity weakening coefficient sequence corresponding to the marked chicken; The difference between each adjacent activity weakening coefficient in the activity weakening coefficient sequence corresponding to the marked chicken is determined as the target difference; The two activity weakening coefficients corresponding to the largest target difference are selected, and a cut is made in the middle of the two activity weakening coefficients, so that the activity weakening coefficient sequence is divided into two sub-weakening coefficient sequences; The mean value of all active weakening coefficients in each sub-weakening coefficient sequence is determined as the sub-representative active weakening factor corresponding to each sub-weakening coefficient sequence; Select the sub-weakening coefficient sequence with a larger corresponding sub-representative activity weakening factor from the two sub-weakening coefficient sequences as the target sub-coefficient sequence; Taking the adjacent key frames corresponding to all active weakening coefficients in the target sub-coefficient sequence as segmentation points, all extracted key frames are segmented to obtain sub-key frame segments; If the number of key frames in the sub-key frame segment is less than 2, the sub-key frame segment is merged with its adjacent sub-key frame segment, and the merged sub-key frame segment is determined as the key frame sequence corresponding to the marked chicken; If the number of key frames in the sub-key frame segment is at least 2, the sub-key frame segment is determined as the key frame sequence corresponding to the marked chicken.

3. The chicken farming image monitoring method based on cloud computing according to claim 1 is characterized in that: Based on the distribution of all target clusters and all discrete points, abnormal monitoring of the monitored chickens includes: Determine the abnormal behavior index corresponding to each chicken to be monitored according to the number of discrete points in all target feature vectors corresponding to each chicken to be monitored and the total number of target clusters to which all target feature vectors corresponding to each chicken to be monitored belong; If the abnormal behavior index corresponding to the chicken to be monitored is greater than the preset abnormal threshold, the chicken to be monitored is determined to be an abnormal behavior chicken.

4. The chicken farming image monitoring method based on cloud computing according to claim 3 is characterized in that: The formula corresponding to the abnormal behavior index of the chicken to be monitored is: ;in, It is i Only the corresponding abnormal behavior indicators of the chickens are to be monitored; i is the serial number of the chicken to be monitored; is the normalization function; It is i The number of discrete points in all target feature vectors corresponding to the chickens to be monitored; It is an exponential function with a natural constant as base; It is i The total number of target clusters to which all target feature vectors corresponding to the chickens to be monitored belong; M is the total number of all target clusters; It is i The accumulated value of the static eigenvalues ​​contained in all discrete points in all target eigenvectors corresponding to the chicken to be monitored.

5. A chicken farming image monitoring system based on cloud computing, characterized in that: It comprises a processor and a memory, wherein the processor is used to process instructions stored in the memory to implement a chicken farming image monitoring method based on cloud computing as described in any one of claims 1 to 4.

Images