Semi-outdoor breeding feeding optimization method and system

Abstract

The invention belongs to the technical field of feeding strategy optimization, and provides a semi-outdoor breeding feeding optimization method and system, and the method specifically comprises the steps: firstly, arranging a camera at a feeding point, collecting a single feeding video through the camera, obtaining the behavioral viscosity and the feces distribution degree from the single feeding video, and carrying out the calculation of the behavioral viscosity and the feces distribution degree; then evaluating a feeding point sewage level according to the behavior viscosity and the excrement distribution degree, and finally judging whether the feeding point layout has defects according to the feeding point sewage level. According to the method, short-term retention concentration caused by foraging behaviors of chicken flocks at different selected feeding point positions in a semi-outdoor breeding scene is effectively quantified, so that whether feeding points which are excessively concentrated or cannot provide feeding welfare of the chicken flocks appear in a feeding point layout state or not is evaluated, and the feed utilization rate and the feeding efficiency of the feeding points are improved; the natural foraging habit is closer to the chicken; and the stress behavior is reduced.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of feeding strategy optimization, and particularly relates to a semi-outdoor breeding feeding optimization method and system. BACKGROUND

[0002] Compared with fast large-scale broilers, Qingyuan chickens have a longer growth cycle. If the traditional method of cage breeding is used, the activity of the chickens will be limited, and the long growth cycle will make the meat loose and affect the taste. Therefore, the most suitable method for Qingyuan chickens is the semi-outdoor breeding mode. The semi-outdoor breeding mode guarantees the opportunity of Qingyuan chickens to forage and move freely. However, when it comes to the method of outdoor breeding, the feeding efficiency has always been based on experience. For example, the method of equally distributing feed from each feeding point. This method of equally distributing feed does not take into account the feeding efficiency of the feeding point, so there is a problem that the feed at some feeding points is quickly consumed, and the feeding amount and feeding welfare of the chicken group near the feeding point cannot be guaranteed. In addition, the feed at some feeding points is wasted or contaminated due to the lack of attraction or the chickens staying there. Therefore, this traditional method of equally distributing feed is currently only used in small-scale breeding farms, that is, the same amount of feed is used in each feeding point, and the outdoor factors such as terrain and light are ignored. However, in larger semi-outdoor breeding farms, such as those with an area of 200 square meters or more, the activity range or concentration of the chicken group is different, and the attractiveness of each feeding point to the chickens or the concentration of the chicken group foraging is also different, resulting in significant differences in manure distribution. This inconsistency is mainly due to the difference in the residence time of Qingyuan chickens at different locations. Therefore, the planning of the quantitative distribution of feeding points should be combined with the activity characteristics of the chicken group, and further adaptive optimization should be performed through sensor networks and big data collection. SUMMARY

[0003] The purpose of the present application is to provide a semi-outdoor breeding feeding optimization method and system to solve one or more technical problems in the prior art and at least provide a beneficial option or create conditions.

[0004] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a semi-outdoor breeding feeding optimization method is provided, which comprises the following steps: S100, arranging a camera at a feeding point and collecting a single feeding video through the camera; S200, obtaining behavior viscosity and manure distribution degree from the single feeding video; S300, evaluating the pollution level of the feeding point according to the behavior viscosity and the manure distribution degree; S400, judging whether the layout of the feeding point has defects according to the pollution level of the feeding point; Further, in step S100, the method of arranging a camera at the feeding point and collecting a single feeding video through the camera is: containing several feeding points in the farm, arranging a camera above each feeding point, and the camera is a wide-angle camera; collecting a video once each time feeding, which is recorded as a single feeding video.

[0005] Further, in step S200, the method of obtaining the behavior viscosity and the fecal pollution distribution from the single feeding video is: obtaining the number of chickens in each video frame from the single feeding video, which is recorded as the recognition quantity, and the attenuation rate of the recognition quantity in the single feeding process is recorded as the behavior viscosity. The linear fitting calculation process directly regards the recognition quantity sequence as the activity intensity change of the whole chicken group at the feeding point, which is easy to ignore the spatial distribution difference caused by terrain, light, air flow, shielding, social preference, etc., leading to the misjudgment of the overall viscosity result to a few hot areas, including underestimation or overestimation, but the calculation process is simple and consumes less. The subarea linear fitting calculation fully combines the flow structure formed by the center, the outer circle to the scattered group when the chicken group responds to the feeding stimulus, and the recognition quantity of each subarea is calculated separately to obtain the behavior viscosity, which is more in line with the natural order of the real chicken group flowing in and dispersing; avoiding the overall behavior viscosity being misled by the high activity of the edge area.

[0006] The fecal pollution area in each video frame is identified from the single feeding video, the fecal pollution diffusion rate is calculated according to the distribution of the pixels corresponding to the fecal pollution area, and the weighted average value of the fecal pollution diffusion rate is obtained by taking the feeding process as the weight to obtain the fecal pollution distribution; this process focuses on using the gaze point rendering method to perform semantic recognition on the video stream.

[0007] The fecal pollution diffusion rate is an index for quantifying the change trend of the fecal pollution distribution density relative to its distance from the center point, and its value is used to reflect the diffusion degree of the fecal pollution from the center to the outside. The smaller the index, the more concentrated the fecal pollution distribution, the better the pollution collection effect of the feeding point, and the higher the eating welfare efficiency of the feeding point to the chickens.

[0008] The process of taking the feeding process as the weight to calculate the weighted average value of the fecal pollution diffusion rate to obtain the fecal pollution distribution is to take exp(-TEnd) as the feeding process weight to calculate the weighted average value of the fecal pollution diffusion rate to obtain the fecal pollution distribution, where exp() is an exponential function with e as the base, and TEnd is the serial number of the video frame. The larger the serial number value, the closer to the end time of the feeding process, i.e., the end time of the video.

[0009] Because during the feeding process, chickens usually have a physiological delay defecation process after eating, especially within several minutes after concentrated eating, which is the peak of defecation, usually 5-15 minutes, so by weighting the process, the misleading caused by identification interference or the absence of fecal pollution at the initial stage of feeding is avoided, and the real defecation aggregation behavior of the chicken group at the feeding point is more accurately reflected; Further, in step S300, the method for evaluating the pollution level of the feeding point according to the behavioral viscosity and the fecal pollution distribution degree is: each feeding as a monitoring point, a preset time lag response window is a set of any monitoring point to the first n1 monitoring points, the size of the time lag response window ranges from n1∈[10, 30], and ti represents the serial number of each monitoring point in the time lag response window, and the smaller the value of ti, the closer to the current time; the binary group of the behavioral viscosity and the fecal pollution distribution degree obtained by the monitoring point at the same time is recorded as a monitoring group; that is, the behavioral viscosity and the fecal pollution distribution degree obtained by each monitoring point are constructed into a binary group for data storage; The covariance of the behavioral viscosity and the fecal pollution distribution degree in the time lag response window is the first joint degree C ti When it is negative, it is judged that the current is an asynchronous behavior mode; The covariance of the behavioral viscosity and the fecal pollution distribution degree in the time lag response window is the first joint degree C ti If the first joint degree is positive, it is defined as a synchronous behavior mode, otherwise as an asynchronous behavior mode; The behavioral viscosity and the fecal pollution distribution degree in the time lag response window are respectively sorted in descending order to obtain a sorting difference degree as a rank mismatch index, and the specific steps are: the serial number value of the behavioral viscosity in the ti th monitoring group after descending sorting of all data in the time lag response window is RV ti The serial number value of the fecal pollution distribution in the ti th monitoring group after descending sorting of all data in the time lag response window is RD ti , and the rank mismatch index is d ti =|RV ti -RD ti |; If the rank mismatch index is greater than 0.25×n1, the monitoring point is defined as a pollution mutation point, and the opportunity of a monitoring point being judged as a pollution mutation point is n1 times, that is, the n1 monitoring points after the occurrence have the opportunity to judge the pollution mutation point, and when it is marked as a pollution mutation point for the first time, it is permanently marked as a pollution mutation point; if the pollution mutation point occurs in an asynchronous behavior mode, it is defined as a pollution anchor point, and the diffusion weight of the pollution anchor point is ep_idx=C ti (1-exp(-d ti )) and the diffusion weight of the non-pollution anchor point is C ti The weighted average value of each fecal pollution distribution is calculated as the pollution level of the feeding point with the diffusion weight as the weight. Wherein each fecal pollution distribution only includes each monitoring point in the time lag response window corresponding to the current monitoring point.

[0010] Wherein exp() is the exponential function with e as the base.

[0011] The fecal pollution distribution degree is constructed based on a video image with the feeding point as the center, and the larger the value is, the more concentrated the fecal pollution is, the steeper the distribution is, and the more in line with the fecal pollution target. The behavior viscosity is based on the decay rate of the recognition quantity of the chicken flock activity, and is used to reflect the retention capacity of the chicken flock when the feeding point is active. The order mismatch index is the order mismatch value of the behavior viscosity and the fecal pollution distribution degree within the time lag response window, and is used to capture the nonlinear response between behavior deviation and pollution surge. When dti is large, it reflects environmental or behavioral abnormalities, such as weather abnormalities or artificial feeding incorporation, which causes the chicken flock to temporarily gather but quickly disperse, resulting in uneven distribution of fecal pollution. Therefore, the calculation principle of the feeding point fecal pollution level is to capture the more concentrated defecation behavior of the chicken flock near the feeding point when the chicken flock is densely active at the feeding point, resulting in an increased degree of fecal pollution distribution. In combination with the time lag response window, it more excludes the time correlation influence caused by environmental factors, and in combination with the pollution anchor point that marks the abnormal disconnection of the chicken flock behavior and the fecal pollution distribution, it fully considers the monitoring points that are highly related or significant to the abnormality of the chicken flock behavior, so as to quantify the concentration performance of the feeding point on the fecal pollution in a feeding cycle.

[0012] Further, in step S300, the method for evaluating the feeding point fecal pollution level according to the behavior viscosity and the fecal pollution distribution degree is: each feeding as a monitoring point, a preset time lag response window is a set of any monitoring point to the first n1 monitoring point, the size of the time lag response window is in the range of n1∈[10, 30], tj represents the serial number of each monitoring point in the time lag response window, and the smaller the value of tj is, the closer to the current time it is; the binary group of the behavior viscosity and the fecal pollution distribution degree is synchronously obtained at the monitoring point, which is called a monitoring group; that is, the behavior viscosity and the fecal pollution distribution degree obtained by each monitoring point are constructed into a binary group for data storage; The average of all behavior viscosities in the time lag response window is calculated and recorded as the viscosity level SV tj , which is defined as the ratio of the viscosity level of any time lag response window to the viscosity level of the previous time lag response window as the behavior condensation potential PV tj ; The harmonic mean of all monitoring point fecal pollution distribution degrees in the time lag response window is calculated and recorded as SD tj ; the absolute value of the difference between the fecal pollution distribution degree and SD tj is defined as the fecal pollution deviation RD t ; if the fecal pollution distribution degree is less than SD tj , the fecal pollution deviation is multiplied by 1+cos(V tj ) to correct the fecal pollution distribution degree; the correction enhances the embodiment of the spatial correlation corresponding to the behavior viscosity, because the fecal pollution deviation is affected by the behavior mode; after the corrected fecal pollution deviation is multiplied by the reciprocal weight corresponding to the time difference degree and summed, the fecal pollution cumulative potential PD tjThe time difference degree is used to strengthen the influence of recent pollution changes. The fecal pollution deviation and the corresponding time difference degree refer to the number of monitoring points between the monitoring point where the variable is obtained and the current monitoring point, and the reciprocal of the obtained number is used as the weight to perform weighted summation on each fecal pollution deviation in the time lag response window. The result obtained by the weighted summation is the fecal accumulation potential.

[0013] The ratio of the geometric mean to the arithmetic mean of the behavior viscosity of all monitoring points in the time lag response window is the deviation stability order k tj . If the deviation stability order is greater than 1, the natural logarithm thereof is taken as the deviation measure, otherwise the natural exponential thereof is taken as the deviation measure. The deviation measure is denoted as eqk tj . The deviation measures of the natural logarithm and the natural exponential are denoted as ln(k tj ) and exp(k tj ) respectively. Here, the deviation stability order reflects the uniformity and dynamic instability of the behavior viscosity distribution.

[0014] The fecal accumulation potential is smoothed and the deviation measure is used as the weight factor thereof. The fecal accumulation potential and the behavior condensation potential are linearly weighted and summed to obtain the fecal accumulation level factor. The specific calculation process is as follows: HL=α·PV tj +β·sigmoid(PD tj )·eqk tj ; where α and β are preset threshold values, α+β=1, the sigmoid function is used to smooth the fecal accumulation potential data, and the instability of the result caused by the numerical value of the abnormal order of magnitude of fecal pollution is reduced. sigmoid() is an activation function. The weights α and β need to be adjusted according to the actual situation of the farm, the behavior characteristics of the chicken flock and the optimization goal. For example, when the farm pays more attention to whether the foraging behavior of the chicken flock needs to ensure the feed utilization rate or reduce stress as the goal, the value of the α weight is increased, and when the attention to the fecal accumulation effect and the cleaning cost is stronger, the value of the β weight is increased; by default, the same value can be taken. This method introduces the sigmoid function to complete the smoothing and normalization of the fecal pollution distribution, ensures the smooth change of the variable weight in the continuous interval, and avoids the interference of extreme numerical values on the stability and convergence of the model; the deviation measure dynamically adjusts the weight, adapts to the fluctuations of the chicken flock caused by weather changes, feeding disturbances, and the like, and is convenient for debugging.

[0015] The ratio of the behavior viscosity to the behavior coagulation potential reflects the relative level of the current flock activity intensity relative to the overall trend, embodies the dynamic aggregation characteristics and the relative changes of the behavior activity of the flock at a specific observation time; by introducing the hyperbolic tangent function to nonlinearly map the behavior uniformity index k and normalizing it by logarithmic transformation, the model effectively suppresses the excessive influence of individual behavior fluctuations on the overall evaluation, realizes the reasonable adjustment of the inherent randomness and spontaneous fluctuations in the flock, and conforms to the steady-state fluctuation law in the dynamic life system; introducing the Laplace operator to the absolute value operation of the spatial second-order derivative of the fecal pollution accumulation potential quantifies the spatial curvature and gradient of the pollutant distribution, accurately reveals the spatial heterogeneity and dynamic evolution of the fecal pollution aggregation and diffusion hotspots in the environment; by constructing a composite response function in the form of the product of multiple indexes to integrate the feedback mechanisms from the perspectives of biological behavior and ecology, the inherent coupling characteristics of the flock behavior and environmental pollution in the semi-outdoor breeding ecosystem are quantitatively predicted.

[0016] Beneficial effects: Since the fecal accumulation level of the feeding point is obtained by time series analysis based on the evaluation of behavior viscosity and fecal pollution distribution, the short-term retention concentration caused by the foraging behavior of the flock in different selected feeding point positions in the semi-outdoor breeding scene can be effectively quantified, so as to evaluate whether the state of the feeding point layout is excessively concentrated or cannot provide ideal fecal accumulation capacity, and to further extract the defect position of the feeding point layout, thereby improving the flock feeding welfare of the feeding point and the concentration degree of fecal pollution, reducing the risk of feed waste and improving the feeding efficiency of the feed.

[0017] Further, in step S400, the method for judging whether the feeding point layout has defects according to the fecal accumulation level of the feeding point is: the number set composed of the fecal accumulation level of each feeding point is recorded as the fecal accumulation level set, the upper quartile and the lower quartile of the fecal accumulation level set are recorded as the aggregation overflow and the aggregation loss, respectively, and the monitoring period is set, with a default value of 5-10 natural days; when the aggregation overflow has a monotonic increasing trend within the monitoring period, it is determined that there is a risk of fecal accumulation overload, and when the aggregation loss has a monotonic decreasing trend within the monitoring period, it is determined that there is a risk of fecal accumulation imbalance; when the fecal accumulation overload risk or the fecal accumulation imbalance risk occurs, it is determined that the feeding point has defects.

[0018] The flock's response to the environment has daily stability and self-organization, and the change trend of the manure distribution itself is a feedback to the good or bad of the feeding layout. If the aggregation overflow amount monotonously increases in the monitoring period, it means that a few feeding points continuously obtain an excessive manure aggregation level, reflecting the trend of the flock excessively concentrating on individual feeding points, and the derived scenarios include manure accumulation overload, excessive flushing burden, insufficient feed distribution, and insufficient flock activity. When the aggregation loss amount monotonously decreases in the monitoring period, it means that the manure aggregation capacity of individual feeding points has significantly declined, reflecting the avoidance trend of the flock to the feeding points. Due to seasonal changes in light and temperature, the feeding points lose the ability to induce the flock's feeding behavior, resulting in immediate resource waste.

[0019] Further, S500, optimizing the defective feeding point layout: calculating the average value of the feeding point manure aggregation level of each feeding point in the monitoring period, denoted as the manure aggregation period level; when it is determined that the feeding point has defects, and the defect belongs to the manure aggregation overload risk, the feeding point with the maximum manure aggregation period level is sent to the client, and the feeding point is warned to present a trend of flock over-reliance, and a new feeding point needs to be added around; when it is determined that the feeding point has defects, and the defect belongs to the manure aggregation imbalance risk, the feeding point with the minimum manure aggregation period level is sent to the client, and the feeding point is warned to present a trend of flock avoidance, and the manure aggregation capacity of the feeding point is insufficient.

[0020] This method realizes the load balancing of the flock's spatial activity by identifying the traffic hotspots and coldspots through the period level, so as to alleviate the pressure on the flock welfare consumption of individual feeding points and the problem of regional manure backlog.

[0021] Preferably, all undefined variables in the present application can be threshold values set by humans if not specifically defined.

[0022] The present application also provides a semi-outdoor breeding feeding optimization system, which comprises a processor, a memory, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to realize the steps in the semi-outdoor breeding feeding optimization method. The semi-outdoor breeding feeding optimization system can run on desktop computers, notebook computers, palm computers, and cloud data centers. The executable system can include, but is not limited to, a processor, a memory, a server cluster, and the like. The processor executes the computer program to run in the following system units: A video data acquisition unit is arranged at the feeding point to acquire a single feeding video through a camera. A distribution state identification unit is arranged to obtain behavior viscosity and manure distribution from the single feeding video. The waste collection assessment unit is used to assess the waste collection level at the feeding point based on behavioral viscosity and waste distribution. The abnormal feeding point layout judgment unit is used to determine whether there are defects in the feeding point layout based on the sewage collection level of the feeding points.

[0023] The beneficial effects of this invention are as follows: This invention provides a method and system for optimizing feeding in semi-outdoor farming. Since the level of waste collection at feeding points is obtained through time series analysis based on behavioral viscosity and fecal distribution, it can effectively quantify the short-term concentration of chickens' foraging behavior under different selected feeding point locations in semi-outdoor farming scenarios. This allows for the assessment of whether there are excessively concentrated feeding points or feeding points that cannot provide ideal feeding welfare for the chickens in the state of the feeding point layout. This provides a mathematical reference basis for further extracting the location of defects in the feeding point layout, thereby improving the feed supply efficiency of feeding points and reducing feed pollution or waste.

[0024] In addition, this process of selecting feeding points based on the distribution of feces and the characteristics of chickens' stay time can make the feeding points closer to the chickens' natural foraging habits and reduce stress behaviors, thereby indirectly improving the welfare and health of the flock. Furthermore, because the feeding points are more adapted to the chickens' habits, it can also prevent feed from being quickly dispersed or wasted by a few individuals and reduce feed waste rate. Attached Figure Description

[0025] The above and other features of the present invention will become more apparent from the detailed description of the embodiments shown in conjunction with the accompanying drawings. In the accompanying drawings, the same reference numerals denote the same or similar elements. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without any creative effort. In the drawings: Figure 1 The diagram shows a flowchart of an optimized feeding method for semi-outdoor aquaculture. Figure 2 The diagram shows the structure of a semi-outdoor aquaculture feeding optimization system. Detailed Implementation

[0026] The following will provide a clear and complete description of the concept, specific structure, and technical effects of the present invention in conjunction with the embodiments and accompanying drawings, so as to fully understand the purpose, solution, and effects of the present invention. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0027] like Figure 1 The diagram shows a flowchart of an optimized feeding method for semi-outdoor aquaculture. The following section will combine... Figure 1 This invention describes a method for optimizing feeding in semi-outdoor aquaculture according to an embodiment of the present invention, the method comprising the following steps: S100, arranging a camera at the feeding point and collecting a single feeding video through the camera; S200, obtaining behavior viscosity and fecal pollution distribution from the single feeding video; S300, evaluating the fecal pollution level of the feeding point according to the behavior viscosity and the fecal pollution distribution; S400, judging whether the feeding point layout has defects according to the fecal pollution level of the feeding point; Further, in step S100, the method of arranging a camera at the feeding point and collecting a single feeding video through the camera is: a plurality of feeding points are included in the farm, a camera is arranged above each feeding point, and the camera is a wide-angle camera; the camera is used to collect single feeding video data of the whole process of the chicken group at the feeding point, and the video is collected once each time, which is recorded as a single feeding video.

[0028] The single feeding whole process includes a preparation stage and an eating stage, wherein the preparation stage is a time period before the feeding is executed, with a duration of 1-5 minutes, used to identify the ground fecal pollution state before feeding as a state baseline, to ensure that the fecal pollution distribution after feeding is caused by the behavior of the current chicken group; the preparation stage is not necessary, and the purpose is to improve the initial accuracy of the fecal pollution distribution. The process of using the single feeding video to obtain the fecal pollution distribution under the default state includes the preparation stage and the eating stage, and the process of obtaining the behavior viscosity only includes the eating stage.

[0029] The eating stage is a time period after the feeding is executed, with a duration of 15-60 minutes, used to extract the behavior viscosity and the fecal pollution distribution. The duration is related to the type of the chicken to be fed, for example, the feeding behavior of Qingyuan chicken is most concentrated within 10 minutes after the feeding is started, and then enters the stage of scattered feeding and staying in place, and 30 minutes is enough to cover the main activity behavior, so 30 minutes is set.

[0030] Further, in step S200, the method of obtaining the behavior viscosity and the fecal pollution distribution from the single feeding video is: obtaining the number of chickens in each video frame from the single feeding video, which is recorded as the recognition quantity, and recording the decay rate of the recognition quantity in the single feeding process as the behavior viscosity. The recognition algorithm of the number of chickens in the video frame is YOLOv5 or Faster R-CNN target detection algorithm, and the calculation method of the decay rate includes two kinds, the first kind is linear fitting calculation, and the specific process is: arranging the recognition quantity in time sequence to form a recognition quantity sequence, using make_interp_spline() function to perform cubic spline fitting on the recognition quantity sequence to obtain a fitting curve, and calculating the average value of the slope in the interval with the maximum value in the fitting curve and the end of the curve to obtain the behavior viscosity.

[0031] The second method is partitioned linear fitting calculation. The video frame is divided into three equal parts according to the distance from the feeding point, which are used as recognition areas. The recognition amount of chickens in each recognition area is recorded as the partition recognition amount. The partition recognition amount is used as the input of the first linear fitting method to calculate the sub-behavior viscosity. The behavior viscosity is the weighted average of the viscosity of each sub-behavior, where the weighting value is the difference between the maximum and minimum values ​​of the partition recognition amount during the feeding process. The fecal and soiled areas in each video frame are identified from a single feeding video. The fecal and soiled diffusion rate is calculated based on the distribution of pixels corresponding to the fecal and soiled areas. The fecal and soiled distribution degree is obtained by calculating the weighted average of the fecal and soiled diffusion rate with the feeding process as the weight. The process of identifying feces-stained areas in each video frame from a single feeding video involves using YOLOv5 or Faster R-CNN object detection algorithms to identify and exclude chicken objects. For each video frame, KCir concentric circles are constructed with equal radii centered at the feeding point. The area between two concentric circles with consecutive radii constitutes a distance-degree region. The centroid distance of the distance-degree region is the average of the radii of the two concentric circles. The percentage of pixels marked as feces-stained areas within the distance-degree region is the feces-stain ratio. The pair formed by the feces-stain ratio and the centroid distance is denoted as a distribution pair. The slope of the sequence of distribution pairs is calculated as the feces diffusion rate; the slope represents the decrease in the feces-stain ratio per unit increase in centroid distance. The process of calculating the weighted average of the manure diffusion rate using the feeding process as the weight to obtain the manure distribution degree is as follows: the weighted average of the manure diffusion rate is calculated using exp(-TEnd) as the weight of the feeding process to obtain the manure distribution degree, where exp() is an exponential function with the natural constant e as the base, and TEnd is the sequence number of the video frame. The larger the sequence number value, the closer it is to the end time of the feeding process, that is, the end time of the video.

[0032] Further, in step S300, the method for assessing the sewage collection level at the feeding point based on behavioral viscosity and fecal distribution degree is as follows: each feeding is treated as a monitoring point, and a preset time-delay response window is the set of any monitoring point up to the previous n1 monitoring points. The size of the time-delay response window is within the range of n1∈[10,30], and ti represents the sequence number of each monitoring point in the time-delay response window. The smaller the value of ti, the closer it is to the current time. The binary data of behavioral viscosity and fecal distribution degree obtained at the monitoring point are recorded as a monitoring group. That is, the behavioral viscosity and fecal distribution degree obtained at each monitoring point are constructed into a binary data set for data storage. The covariance of the behavioral viscosity and fecal distribution within the time-delay response window is used as the first degree of unity C. ti When it is negative, it is determined that the current behavior mode is asynchronous. The covariance of the behavioral viscosity and fecal distribution degree within the time-delay response window is the first degree of unity C. ti, if the first copula degree is positive, it is defined as a synchronous behavior pattern, otherwise it is an asynchronous behavior pattern; The rank difference degree is obtained by descending order sorting of the behavior viscosity and the fecal pollution distribution degree in the time lag response window, respectively. The specific steps are as follows: the sequence number value of the behavior viscosity in the ti th monitoring group descendingly sorted in all data in the time lag response window is RV ti , the sequence number value of the fecal pollution distribution in the ti th monitoring group descendingly sorted in all data in the time lag response window is RD ti , and the rank order mismatch index is d ti =|RV ti -RD ti |; If the rank order mismatch index is greater than 0.25*n1, the monitoring point is defined as a pollution mutation point. The opportunity of a monitoring point being judged as a pollution mutation point is n1 times, that is, the n1 monitoring points after it have the opportunity to judge it as a pollution mutation point. When it is marked as a pollution mutation point for the first time, it is permanently marked as a pollution mutation point. If the pollution mutation point occurs in an asynchronous behavior pattern, it is defined as a pollution anchor point. The diffusion weight of the pollution anchor point is ep_idx=C ti (1-exp(-d ti )), and the diffusion weight of the non-pollution anchor point is C ti . The weighted average value of each fecal pollution distribution is calculated as the feeding point pollution level, wherein each fecal pollution distribution only includes each monitoring point in the time lag response window corresponding to the current monitoring point.

[0033] Where exp() is the exponential function with base e.

[0034] Further, in step S300, the method for evaluating the feeding point pollution level according to the behavior viscosity and the fecal pollution distribution is: each feeding is taken as a monitoring point, a preset time lag response window is a set of any monitoring point to the previous n1 monitoring points, the size of the time lag response window ranges from n1∈[10, 30], and tj represents the sequence number of each monitoring point in the time lag response window, and the smaller the tj value, the closer to the current time; the behavior viscosity and the fecal pollution distribution are synchronously acquired at the monitoring point, and a binary tuple is recorded as a monitoring group; that is, the behavior viscosity and the fecal pollution distribution obtained by each monitoring point are constructed into a binary tuple for data storage; The average of all behavior viscosities in the time lag response window is calculated and recorded as viscosity level SV tj , and the ratio of any time lag response window to the viscosity level of the previous time lag response window is defined as behavior condensation potential PV tj; to ensure stability, if the current behavior condensation potential exceeds the previous monitoring point behavior condensation potential three times, it is marked as an abnormal point, and the value is replaced by 1.5-2.5 times the previous monitoring point behavior condensation potential, to avoid abnormal fluctuations affecting subsequent analysis. The previous time lag response window of any time lag response window is the time lag response window of the first monitoring point without intersection in the reverse time direction of the time lag response window; The harmonic mean of the fecal pollution distribution degree of all monitoring points in the time lag response window is calculated as SD tj ; the absolute value of the difference between the fecal pollution distribution degree and SD tj is defined as the fecal pollution deviation RD t ; if the fecal pollution distribution degree is less than SD tj , the fecal pollution deviation is multiplied by 1+cos(V tj ) to correct the fecal pollution distribution degree; the correction enhances the spatial correlation corresponding to the behavior viscosity, because the fecal pollution deviation is affected by the behavior pattern; the corrected fecal pollution deviation is multiplied by the reciprocal weight of the corresponding time difference degree, and the sum is obtained as the fecal cumulative potential PD tj ; the time difference degree strengthens the influence of recent pollution changes. The fecal pollution deviation and the corresponding time difference degree refer to the number of monitoring points between the monitoring point where the variable is obtained and the current monitoring point, and the reciprocal of the obtained number is used as the weight to weight and sum the fecal pollution deviation in the time lag response window. The result is the fecal cumulative potential.

[0035] The ratio of the geometric mean to the arithmetic mean of the behavior viscosity of all monitoring points in the time lag response window is the deviation stability order k tj , if the deviation stability order is greater than 1, take its natural logarithm as the deviation measure, otherwise take the natural exponential as the deviation measure; the deviation measure is denoted as eqk tj , the deviation measure of the natural logarithm and the natural exponential are denoted as ln(k tj ) and exp(k tj ) respectively; here the deviation stability order reflects the uniformity and dynamic instability of the behavior viscosity distribution.

[0036] The fecal cumulative potential is smoothed with the deviation measure as its weight factor, and the behavior condensation potential is linearly weighted and summed to obtain the pollution level factor. The specific calculation process is: HL=α·PV tj +β·sigmoid(PD tj )·eqk tj; wherein alpha and beta are preset threshold values, alpha+beta=1, the sigmoid function is used to smooth the fecal accumulation potential data, and reduce the instability of result misjudgment caused by the abnormal number of fecal pollution; sigmoid() is an activation function. The weights alpha and beta need to be adjusted according to the actual situation of the farm, the behavior characteristics of the chicken group and the optimization goal. For example, when the farm pays more attention to whether the foraging behavior of the chicken group needs to ensure the feed utilization rate or reduce stress as the goal, the value of the alpha weight is increased, and when the attention to the effect of the pollution collection and the cleaning cost is stronger, the value of the beta weight is increased; by default, the same value can be taken.

[0037] Or build a pollution collection level model as follows: according to the behavior condensation potential PV tj , the fecal accumulation potential PD tj and the deviation measure eqk tj , build a pollution collection level model to obtain the pollution collection level factor HL: ; wherein D tj is the fecal distribution degree, tanh() is the inverse tangent function, mean{} is the average value function, and the average value of the data corresponding to each monitoring point with serial number tj in the time delay response window is calculated when calling; Further, in step S400, the method for judging whether the feeding point layout has defects according to the feeding point pollution collection level is: the number set composed of the pollution collection levels of each feeding point is recorded as a pollution collection level set, the upper quartile and the lower quartile of the pollution collection level set are recorded as the aggregation overflow and the aggregation loss respectively, and the monitoring period is set, and the default value is 5-10 natural days; when the aggregation overflow has a monotonic increasing trend within the monitoring period, it is determined that there is a risk of pollution collection overload, and when the aggregation loss has a monotonic decreasing trend within the monitoring period, it is determined that there is a risk of pollution collection imbalance; when the risk of pollution collection overload or the risk of pollution collection imbalance occurs, it is determined that the feeding point has defects.

[0038] Further, it further includes S500, optimizing the feeding point layout with defects: calculating the average value of the feeding point pollution collection level of each feeding point in the monitoring period, which is recorded as the pollution collection period level; when it is determined that the feeding point has defects, and the defect belongs to the risk of pollution collection overload, the feeding point with the maximum pollution collection period level is sent to the client, and it is warned that the feeding point presents a trend of excessive dependence of the chicken group, and new feeding points need to be added around; when it is determined that the feeding point has defects, and the defect belongs to the risk of pollution collection imbalance, the feeding point with the minimum pollution collection period level is sent to the client, and it is warned that the feeding point presents a trend of avoidance of the chicken group, and it is warned that the pollution collection capacity of the feeding point is insufficient.

[0039] The embodiment of the application provides a semi-outdoor breeding feeding optimization system, which is used for Figure 2As shown is a structure diagram of a semi-outdoor breeding feeding optimization system of the present application. The semi-outdoor breeding feeding optimization system of the embodiment comprises a processor, a memory, and a computer program stored in the memory and executable on the processor. The processor implements the steps in the semi-outdoor breeding feeding optimization method embodiment when executing the computer program.

[0040] The system comprises a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to run in the units of the following system: A video data acquisition unit is configured to arrange a camera at a feeding point and acquire a single feeding video through the camera. A distribution state identification unit is configured to acquire behavior viscosity and fecal pollution distribution from the single feeding video. A pollution collection evaluation unit is configured to evaluate the pollution collection level of the feeding point according to the behavior viscosity and the fecal pollution distribution. A feeding point layout anomaly judgment unit is configured to judge whether the feeding point layout has defects according to the pollution collection level of the feeding point.

[0041] The semi-outdoor breeding feeding optimization system can run in a desktop computer, a notebook computer, a palm computer, a cloud server, and other computing devices. The semi-outdoor breeding feeding optimization system can run in a system that can include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the example is only an example of the semi-outdoor breeding feeding optimization system and does not constitute a limitation on the semi-outdoor breeding feeding optimization system, which can include more or fewer components, or combine certain components, or different components, for example, the semi-outdoor breeding feeding optimization system can also include an input / output device, a network access device, a bus, and the like.

[0042] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic components, discrete hardware components, etc. The general-purpose processor can be a microprocessor or can also be any conventional processor. The processor is a control center of the one semi-outdoor breeding feeding optimization system running system, and connects each part of the one semi-outdoor breeding feeding optimization system running system through various interfaces and lines.

[0043] The memory can be used to store the computer program and / or modules, and the processor realizes various functions of the one semi-outdoor breeding feeding optimization system by running or executing the computer program and / or modules stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area. The program storage area can store an operating system, at least one application program required by a function (such as a sound playing function, an image playing function, etc.), etc. The data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, for example, a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state memory devices.

[0044] Although the description of the present application has been quite detailed and particularly described with respect to several embodiments, it is not intended to be limited to any of these details or embodiments or any special embodiment, so as to effectively cover the intended scope of the present application. In addition, the present application is described above in the embodiments that the inventors can foresee, and the purpose is to provide a useful description, and non-essential modifications to the present application that have not yet been foreseen can still represent equivalent modifications of the present application.

Claims

1. A method for optimizing feeding in semi-outdoor farming, characterized by, The method comprises the following steps: S100, arranging a camera at a feeding point and collecting a single feeding video through the camera; S200, obtaining behavior viscosity and fecal pollution distribution from the single feeding video; S300, evaluating the fecal pollution level of the feeding point according to the behavior viscosity and the fecal pollution distribution; S400, judging whether the layout of the feeding point is defective according to the fecal pollution level of the feeding point; The method for evaluating the fecal pollution level of the feeding point according to the behavior viscosity and the fecal pollution distribution is to extract monitoring data by using a time lag response window, calculate the covariance and order difference of the two, identify pollution mutation points and pollution anchor points, and weight the fecal pollution distribution according to the diffusion weight, so as to quantify the fecal pollution capacity of the feeding point and obtain the fecal pollution level of the feeding point.

2. The method according to claim 1, wherein, In step S100, the method for arranging a camera at a feeding point and collecting a single feeding video is as follows: a plurality of feeding points are arranged in a breeding farm, and a camera is arranged above each feeding point, which is a wide-angle camera; a video is collected once each time of feeding, which is referred to as a single feeding video.

3. The method according to claim 1, wherein, In step S200, the method for obtaining behavior viscosity and fecal pollution distribution from a single feeding video is as follows: the number of chickens in each video frame is obtained from the single feeding video, which is referred to as an identification quantity, and the decay rate of the identification quantity in the single feeding process is referred to as behavior viscosity; the fecal pollution area in each video frame is identified from the single feeding video, the fecal pollution diffusion rate is calculated according to the distribution of the corresponding pixels of the fecal pollution area, and a weighted average value of the fecal pollution diffusion rate is obtained by taking the feeding process as a weight, so as to obtain the fecal pollution distribution.

4. The method according to claim 1, wherein, In step S300, the method for evaluating the fecal pollution level of the feeding point according to the behavior viscosity and the fecal pollution distribution is as follows: each feeding is taken as a monitoring point, a preset time lag response window is a set of any monitoring point to the first n1 monitoring points, and a binary group of behavior viscosity and fecal pollution distribution obtained synchronously at the monitoring point is referred to as a monitoring group. The covariance of the behavior viscosity and the fecal pollution distribution degree in the time delay response window is a first correlation degree C ti When the value is negative, it is determined that the current is an asynchronous behavior mode. The order mismatch index is obtained by respectively ranking the behavioral viscosity and the fecal pollution distribution in descending order within the time delay response window. If the order mismatch index is greater than 0.25*n1, the monitoring point is defined as a pollution mutation point. If the pollution mutation point occurs in an asynchronous behavior mode, it is defined as a pollution anchor point. The diffusion weight of the pollution anchor point is ep_idx=C ti (1-exp(-d ti )) and the diffusion weight of the non-pollution anchor point is C ti , The fecal pollution accumulation potential is smoothed, and the deviation degree is taken as a weight factor to linearly weight and sum the behavior condensation potential, so as to obtain a fecal pollution level factor. The weighted average value of the diffusion weight for each fecal pollution distribution is the feeding point set pollution level.

5. The method of claim 1, wherein the method is a method of optimizing feeding in a semi-outdoor farming. In step S300, the method for evaluating the pollution level of the feeding point according to the behavioral viscosity and the fecal pollution distribution degree is: each feeding as a monitoring point, a preset time lag response window is a set of any monitoring point to the first n1 monitoring points, and a binary tuple of the behavioral viscosity and the fecal pollution distribution degree synchronously obtained at the monitoring points is recorded as a monitoring group; the average of all the behavioral viscosities in the time lag response window is calculated and recorded as the viscosity level SV tj , and the ratio of the viscosity level of any time lag response window to the viscosity level of the previous time lag response window is defined as the behavioral condensation potential PV tj . Calculate the harmonic mean SD of fecal waste distribution at all monitoring points within the time-delay response window. tj Defined as fecal distribution degree and SD tj The absolute value of the difference is the fecal waste deviation RD. t If the distribution of fecal waste is less than SD tj Then multiply the deviation of fecal matter by 1 + cos(V) tj Perform fecal distribution correction; multiply the corrected fecal deviation by the inverse weight of the corresponding time difference and sum them to obtain the fecal accumulation potential PD. tj ; The ratio of the geometric mean to the arithmetic mean of the viscosity of all monitoring points in the time delay response window is the deviation stability order k tj If the deviation stability order is greater than 1, the natural logarithm of the deviation stability order is taken as the deviation measure, otherwise the natural exponential is taken as the deviation measure; In step S400, the method for judging whether the layout of the feeding point is defective according to the fecal pollution level of the feeding point is as follows: a set of fecal pollution levels of each feeding point is referred to as a fecal pollution level set, the upper quartile and the lower quartile of the fecal pollution level set are referred to as an aggregation overflow and an aggregation loss, a preset monitoring period is set, when the aggregation overflow has a monotonic increasing trend within the monitoring period, it is determined that there is a risk of fecal pollution overload, and when the aggregation loss has a monotonic decreasing trend within the monitoring period, it is determined that there is a risk of fecal pollution imbalance; 6. The method of claim 1, wherein, When the risk of fecal pollution overload or the risk of fecal pollution imbalance occurs, it is determined that the feeding point has defects. ​ 7. The method of claim 1, wherein the method is a method of optimizing feeding in a semi-outdoor farming. Also comprising S500, optimizing the feeding point layout with defects: calculating the average value of the feeding point set pollution level of each feeding point in the monitoring period, denoted as the set pollution cycle level; when it is determined that the feeding point has defects, and the defects belong to the set pollution overload risk, the feeding point with the maximum set pollution cycle level is sent to the client, warning that the feeding point presents a trend of excessive dependence of chicken flock, and new feeding points need to be added around; when it is determined that the feeding point has defects, and the defects belong to the set pollution imbalance risk, the feeding point with the minimum set pollution cycle level is sent to the client, warning that the feeding point presents a trend of chicken flock avoidance, and warning that the set pollution capacity of the feeding point is insufficient.

8. A semi-outdoor farming feeding optimization system, characterized in that, The semi-outdoor breeding feeding optimization system comprises a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the semi-outdoor breeding feeding optimization method in any one of claims 1-7, and the semi-outdoor breeding feeding optimization system runs in a computing device of a desktop computer, a notebook computer, a palm computer, and a cloud data center.

Images