A method for measuring the weight distribution of chicken groups based on poultry grading weighing

Abstract

The present invention discloses a method for measuring the weight distribution of a chicken group based on poultry grading. The method comprises: weighing the chicken group by cage and grading the quality of the chicken group into several quality grades, inputting the weight of each cage of broilers of each quality grade into a handheld distribution weighing device for processing, obtaining the weight distribution before grading and the weight difference before and after the weight dividing line after grading, and then establishing a chicken group weight distribution measurement model after linear superposition of distribution; sampling and iteratively processing the weight data set constructed by the weight of each cage of broilers to obtain the posterior probability of the weight data set, and then obtaining the estimated distribution parameters, and inputting the estimated distribution parameters into the model for processing and outputting the weight distribution of the broilers after grading to achieve distribution weight measurement. The method of the present invention can reduce the number of times broilers are weighed in actual production, and realize high-precision automated analysis of the weight distribution of broiler groups in a quality grading manner, so as to further guide poultry production and breeding.

Description

Technical Field

[0001] The invention relates to a method for measuring the weight of a chicken group by weight distribution, and relates to the field of poultry breeding and production, and in particular to a method for measuring the weight of a chicken group by weight distribution based on poultry classification weighing. Background Art

[0002] The weight distribution of poultry is a key indicator of poultry breeding conditions. Deviations between the measured values ​​of weight distribution, such as average weight, standard deviation, and uniformity, and the expected values ​​usually indicate problems such as malnutrition or disease infection in poultry. In the poultry industry, large poultry processing companies generally have strict requirements on the weight of finished poultry, so the weight distribution of poultry for sale will directly affect the sales profits of poultry breeding companies.

[0003] In order to obtain the weight distribution of poultry, in addition to the most widely used manual weighing method and weight prediction model, the automatic weighing system has also been put into use in some modern poultry houses. However, in terms of live poultry weighing, due to the large number of poultry, the cost of manual sampling and weighing is high. Most automatic weighing systems that rely on image analysis also face the problems of low accuracy and high equipment prices. Summary of the invention

[0004] In order to solve the problems existing in the background technology, the present invention provides a method for measuring the weight distribution of a chicken group based on poultry grading weighing.

[0005] The technical solution adopted by the present invention is:

[0006] The chicken group redistribution weight measurement method based on poultry graded weighing of the present invention comprises:

[0007] Step S1: weigh each broiler in a chicken farm by cage, and classify them into several quality grades after quality classification.

[0008] Step S2: The weight of each cage of broilers in each quality grade is input into a handheld distribution weighing device for processing, and the weight distribution of each cage of broilers before grading and the weight difference before and after the weight dividing line after grading are obtained according to the weight of each cage of broilers in each quality grade. Then, a redistribution weighing model of the chicken population is established after linear superposition of the distribution.

[0009] Step S3: construct the weight of the broilers in each cage into a weight data set, obtain the posterior probability of the weight data set after sampling and iterative processing, and then obtain the estimated distribution parameters according to the posterior probability.

[0010] Step S4: The estimated distribution parameters are input into the chicken group redistribution weighing model for processing. After processing, the chicken group redistribution weighing model outputs the graded broiler weight distribution and displays it after preprocessing on a handheld distribution weighing device, thereby finally realizing the distribution weighing of the chicken group.

[0011] In the step S1, the broilers in the chicken farm are evenly divided according to the number and placed in N chicken cages, and then the broilers in each cage are weighed, and the broilers in each cage are divided into two or three quality levels according to the live quality grade from high to low according to the weight of the broilers in each cage, thereby completing the quality grade. The two-level quality grade division is a two-level scheme, which means that the chicken farm initially divides the chickens into two levels, namely, Grade I & Grade II and Grade III, and the three-level quality grade division is a three-level scheme, which means that the chicken farm initially divides the chickens into three levels, namely, Grade I, Grade II and Grade III; the first level in the two-level scheme includes Grade I & Grade II in the three-level scheme.

[0012] In step S2, when each cage of broiler chickens is divided into two quality levels, it is as follows:

[0013] Firstly, according to the weight of each cage of broiler chickens in each quality grade, Gaussian distribution is used to obtain the weight distribution of each cage of broiler chickens before grading and the weight distribution before and after the weight dividing line of the two quality grades after grading. According to the weight distribution before and after the weight dividing line of the two quality grades after grading, the weight difference before and after the weight dividing line after grading is obtained. Then, the weight distribution of each cage of broiler chickens before grading and the weight difference before and after the weight dividing line after grading are linearly superimposed to establish a redistribution weight measurement model for the chicken population under two quality grades.

[0014] The chicken group redistribution weight measurement model under the two-level quality level is as follows:

[0015] f 1( m , a , s )=(2 a -1) N ( m │ m , s 2 )+(1- a ) N ( m │0.6 m , s 2 )

[0016] in, f 1( ) is the weight distribution of broiler chickens after classification under two quality grades; m is the body weight of broilers in each cage; a and s are the first and second estimated distribution parameters under two quality levels respectively; N ( ) is Gaussian distribution; m is the mean parameter of the Gaussian distribution.

[0017] In step S2, when each cage of broiler chickens is divided into three quality levels, it is as follows:

[0018] Firstly, the beta distribution is used according to the weight of each cage of broiler chickens in each quality grade to obtain the weight distribution of each cage of broiler chickens before grading and the weight distribution before and after the two weight dividing lines of the three quality grades after grading. The weight difference before and after the two weight dividing lines after grading is obtained according to the weight distribution before and after the two weight dividing lines of the three quality grades after grading. Then, the weight difference before and after the two weight dividing lines after grading is multiplied with the Gaussian distribution with the two weight dividing lines of the three quality grades as the mean, and then the distribution is linearly superimposed with the weight distribution of each cage of broiler chickens before grading to establish the redistribution weight measurement model of the chicken population under the three quality grades.

[0019] The chicken group redistribution weight measurement model under the three quality levels is as follows:

[0020] y 3~ f 2( m ´, a,a,s )= Beta ( m 1≤ m ´≤ m 2│ α )- a 1[ Beta ( m ´│ α )- Beta (2 m 1- m ´│ α )] N ( m ´≤ m 2│ m 1, s 1 2 )- a 2[ Beta ( m ´│ α )- Beta (2 m 2- m ´│ α )] N ( m 1≤ m ´│ m 2, s 2 2 )

[0021] α =( α 1, α 2), a =( a 1, a 2), s =( s 1, s 2)

[0022] in, y 3 is the weight distribution of broilers in each cage after three-level quality classification; f 2( ) is the weight distribution of broiler chickens after classification under three quality grades; m ´ is the weight of broilers in each cage m The scaled weight, m ´= m / 1.6; α , a and s are the first, second and third estimated distribution parameter vectors under three quality levels respectively; Beta ( ) is Beta distribution; m 1 and m 2 are the weight dividing lines between the second and third grades and between the first and second grades under the three-grade quality level; α 1. α 2. a 1. a 2. s 1 and s 2 are the first, second, third, fourth, fifth and sixth estimated distribution parameters under three quality levels respectively; N ( ) is a Gaussian distribution.

[0023] In step S3, when each cage of broiler chickens is divided into two quality levels, the weight of each cage of broiler chickens is first constructed as a weight data set, and then Gibbs sampling is used to obtain the posterior probability of the estimated distribution parameters under the two quality levels.

[0024] In the step S3, the posterior probabilities of the estimated distribution parameters at the two quality levels are iterated using Markov Monte Carlo to obtain the estimated distribution parameters at the two quality levels.

[0025] In the step S3, when each cage of broiler chickens is divided into three quality levels, the weight of each cage of broiler chickens is first constructed as a weight data set, and then the posterior probability of the estimated distribution parameters under the three quality levels is obtained by using the Bayesian theorem; the posterior probability of the estimated distribution parameters under the three quality levels is iterated using the Metropolis-Hastings algorithm to obtain the estimated distribution parameters under the three quality levels.

[0026] In step S4, the weight distribution of the graded broilers is preprocessed to obtain the average weight, standard deviation, uniformity, etc. The average weight, standard deviation, uniformity, etc. of the weight distribution of the graded broilers can be used to further guide poultry production and breeding and adjust breeding strategies.

[0027] The beneficial effects of the present invention are:

[0028] The invention uses a chicken group weight distribution measurement method that only relies on poultry grading and weighing data. It can reduce the number of broiler weighing times in actual production. After the chickens are graded and caged according to certain weight standards, the graded small groups are weighed as a whole and the number and weight are recorded. Based on this graded weighing data, the performance data of the broiler group, such as the average weight, standard deviation, and uniformity, can be analyzed, thereby achieving high-precision automated analysis of the weight distribution of the broiler group, so as to further guide poultry production and breeding. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 is an overall flow chart of the method of the present invention;

[0030] Figure 2 A schematic diagram of a handheld device of the present invention;

[0031] Figure 3 is a convergence trend diagram of the estimated distribution parameters in the two-stage scheme of the present invention, wherein, Figure 3 (a) is the estimated distribution parameter under the two-level scheme of the present invention a The convergence trend diagram of Figure 3 (b) is the estimated distribution parameters under the two-level scheme of the present invention s Convergence trend diagram of

[0032] Figure 4 is a convergence trend diagram of the estimated distribution parameters in the three-level scheme of the present invention, wherein, Figure 4 (a) is the estimated distribution parameter of the three-level scheme of the present invention α 1's convergence trend graph, Figure 4 (b) is the estimated distribution parameter of the three-level scheme of the present invention α 2's convergence trend diagram, Figure 4 (c) is the estimated distribution parameter of the three-level scheme of the present invention a 1's convergence trend graph, Figure 4(d) is the estimated distribution parameter of the three-level scheme of the present invention s 1's convergence trend graph, Figure 4 (e) is the estimated distribution parameter of the three-level scheme of the present invention a 2's convergence trend diagram, Figure 4 (f) is the estimated distribution parameter in the three-level scheme of the present invention s 2's convergence trend graph;

[0033] Figure 5 The data set histogram and corresponding quantile-quantile diagram of the actual distribution and estimated distribution of data sets from different sources of the present invention, wherein: Figure 5 (a) is the histogram of the actual data set and the corresponding quantile-quantile plot using the two-level scheme under virtual sampling. Figure 5 (b) is the histogram of the estimated data set and the corresponding quantile-quantile plot using the two-level scheme under virtual sampling. Figure 5 (c) is the histogram of the actual data set using the three-level scheme under virtual sampling and the corresponding quantile-quantile plot. Figure 5 (d) is the histogram of the estimated data set and the corresponding quantile-quantile plot using the three-level scheme under virtual sampling. Figure 5 (e) is the actual data set histogram and corresponding quantile-quantile plot of a broiler sales platform using a two-level solution. Figure 5 (f) is the histogram of the estimated data set and the corresponding quantile-quantile plot using the two-level scheme under a broiler sales platform data;

[0034] In the figure: 1. display area, 2. input area, 3. device hand-held part, 4. device housing. DETAILED DESCRIPTION

[0035] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0036] like Figure 1 As shown, the present invention takes yellow-feathered broiler chickens as an example to further illustrate the implementation process of the chicken group redistribution weight measurement method based on poultry grading weighing of the present invention, which is specifically as follows:

[0037] Step S1: Weigh each broiler in the chicken farm's flock in each cage, and divide it into several quality grades after quality grading. Divide each broiler in the chicken farm's flock evenly according to the number and put it into N cages, then weigh the broilers in each cage, and divide each cage of broilers into two or three quality grades from high to low according to the weight of the broilers in each cage, so as to complete the quality grading. The two-level quality grade division is a two-level scheme, which means that the chicken farm divides the chicken flock into two grades, namely, grade Ⅰ & Ⅱ and grade Ⅲ, and the three-level quality grade division is a three-level scheme, which means that the chicken farm divides the chicken flock into three grades, namely, grade Ⅰ, grade Ⅱ and grade Ⅲ; the first level in the two-level scheme includes grades Ⅰ & Ⅱ in the three-level scheme. As shown in Table 1 below, it is the grading standard for the live quality of yellow-feathered broilers, in g.

[0038] Table 1

[0039]

[0040] Enter the weight of each cage of broilers in each quality class into Figure 2 The weight of each cage of broilers in each quality grade is processed in the handheld distribution weighing device shown in the figure, and the weight distribution of each cage of broilers before grading and the weight difference before and after the weight dividing line after grading are obtained. Then, the distribution linear superposition is performed to establish the redistribution weighing model of the chicken population.

[0041] When each cage of broilers is divided into two quality grades, firstly, the weight distribution of each cage of broilers before grading and the weight distribution before and after the weight dividing line of the two quality grades after grading are obtained using Gaussian distribution according to the weight of each cage of broilers in each quality grade. The weight distribution of each cage of broilers before grading into two quality grades is as follows:

[0042] y 1~ N ( m │ m , s 2 )=exp(-( m - m ) 2 / 2 s 2 ) / (2π) 1 / 2 s

[0043] in, y 1 is the weight distribution of broilers in each cage before two-level quality classification; ~ is approximate; N ( ) is a Gaussian distribution, m is the weight of broilers in each cage, mis the mean parameter of the Gaussian distribution, m =( α 1-1) / ( α 1+ α 2-2)=1.4kg; s is the second estimated distribution parameter under two quality levels.

[0044] According to the weight distribution before and after the weight dividing line of the two quality grades after classification, the weight difference before and after the weight dividing line after classification is obtained, and the quality dividing line between the first grade and the second grade in the two-grade scheme is m 1, use Gaussian distribution to obtain the weight distribution, and then obtain the weight difference before and after the graded weight dividing line△ m a as follows:

[0045] △ m a =N (2 m 1- m 1│ m , s 2 )- N ( m 2│ m , s 2 )

[0046] in, m 1 and m 2 are the body weights of broilers per cage in the first and second quality levels of the two-level scheme, respectively.

[0047] Then the weight distribution of broilers in each cage before grading and the weight difference before and after the weight dividing line after grading are linearly superimposed as follows:

[0048] y 2~ f 1( m , a , s )= aN ( m │ m , s 2 )+(1- a )[ N (2 m 1- m │ m , s 2 )- N ( m │ m , s2 )]

[0049] = aN ( m │ m , s 2 )+(1- a )[ N ( m │2 m 1- m , s 2 )- N ( m │ m , s 2 )]

[0050] =(2 a -1) N ( m │ m , s 2 )+(1- a ) N ( m │0.6 m , s 2 )

[0051] in, y 2 is the weight distribution of broiler chickens in each cage after two-level quality grade classification.

[0052] Finally, a chicken group redistribution weight measurement model under two quality levels was established as follows:

[0053] f 1( m , a , s )=(2 a -1) N ( m │ m , s 2 )+(1- a ) N ( m │0.6 m , s 2 )

[0054] in, f 1( ) is the weight distribution of broiler chickens after classification under two quality grades; m is the body weight of broilers in each cage; a and sare the first and second estimated distribution parameters under two quality levels respectively; N ( ) is Gaussian distribution; m is the mean parameter of the Gaussian distribution.

[0055] Taking hens as an example, based on the weight classification standards for broilers of different growth types in Table 1, the original Beta distribution in the (0,1) interval is scaled to the (0,1.6) interval, that is, the proportion of hens weighing more than 1.6 kg in the Beta distribution is ignored.

[0056] When each cage of broilers is divided into three quality grades, the weight distribution of each cage of broilers before grading and the weight distribution before and after the two weight dividing lines of the three quality grades after grading are first obtained using Beta distribution according to the weight of each cage of broilers in each quality grade. The weight distribution of each cage of broilers before grading into the three quality grades is as follows:

[0057] y 3~ Beta ( m ´│ α 1, α 2)=Γ( α 1+ α 2) m ´ α1-1 (1- m ´) α2-1 / Γ( α 1)Γ( α 2)

[0058] in, y 3 is the weight distribution of broilers in each cage before the three-level quality classification; Beta ( ) is Beta distribution; m ´ is the weight of broilers in each cage m The scaled weight, m ´= m / 1.6; α 1 and α 2 are the first and second estimated distribution parameters under three quality levels respectively; Γ( ) is the gamma function, which is Euler's second integral.

[0059] According to the weight distribution before and after the two weight dividing lines of the three-level quality grade after classification, the weight difference before and after the two weight dividing lines after classification is obtained, and the quality dividing line between grade II and grade III in the three-level scheme is m 1, use Beta distribution to obtain the weight distribution, and then obtain the weight difference before and after the weight dividing line of level II and level III m b as follows:

[0060] △ m b =Beta (2 m 1- m 3│ α 1, α 2)- Beta ( m 2│ α 1, α 2)

[0061] Quality dividing line between level I and level II in the three-level scheme m 2, use the Beta distribution to obtain the weight distribution, and then obtain the weight difference before and after the weight dividing line of level I and level II. m c as follows:

[0062] △ m c =Beta (2 m 2- m 4│ α 1, α 2)- Beta ( m 3│ α 1, α 2)

[0063] in, m 4. m 3 and m 2Respectively, the body weight of broilers in each cage in level I, level II and level III of the three-level scheme.

[0064] Then, the weight difference before and after the two weight dividing lines after grading is multiplied by the Gaussian distribution with the two weight dividing lines of the three quality levels as the mean, and then the weight distribution of broilers in each cage before grading is linearly superimposed to establish the weight measurement model of the chicken group redistribution under the three quality levels, as follows:

[0065] y 3~ f 2( m ´, a,a,s )= Beta ( m 1≤ m ´≤ m 2│ α )- a 1[ Beta ( m ´│ α )- Beta (2 m 1- m ´│ α )]N ( m ´≤ m 2│ m 1, s 1 2 )- a 2[ Beta ( m ´│ α )- Beta (2 m 2- m ´│ α )] N ( m 1≤ m ´│ m 2, s 2 2 )

[0066] α =( α 1, α 2), a =( a 1, a 2), s =( s 1, s 2)

[0067] in, y 3 is the weight distribution of broilers in each cage after three-level quality classification; f 2( ) is the weight distribution of broiler chickens after classification under three quality grades; m ´ is the weight of broilers in each cage m The scaled weight, m ´= m / 1.6; α , a and s are the first, second and third estimated distribution parameter vectors under three quality levels respectively; Beta ( ) is Beta distribution; m 1 and m 2 are the weight dividing lines between the second and third grades and between the first and second grades under the three-grade quality level; α 1. α 2. a 1. a 2. s 1 and s 2 are the first, second, third, fourth, fifth and sixth estimated distribution parameters under three quality levels respectively; N ( ) is a Gaussian distribution.

[0068] Assume that the dividing line between level I and level II m 2=0.9×1.6=1.44kg, assuming the dividing line between level II and level III m 1=0.7×1.6=1.12kg, for the Beta distribution, the above assumption has interval symmetry, that is ( m 2- m ) / (1.6- m )=( m - m 1) / m .

[0069] When each cage of broiler chickens is divided into two quality levels, the weight of each cage of broiler chickens is first constructed as a weight data set, and then Gibbs sampling is used to obtain the posterior probability of the estimated distribution parameters under the two quality levels. The posterior probability of the estimated distribution parameters under the two quality levels is iterated using Markov Monte Carlo to obtain the estimated distribution parameters under the two quality levels. Markov Monte Carlo is used to obtain the estimated distribution parameters under the two quality levels through sampling conditional posterior probabilities in the iterative process. p ( a │ s , y N ) Estimate the posterior probability of the distribution parameters p ( a , s │ y N ), y N is the weight data set. In each iteration, the univariate conditional probability of each estimated distribution parameter needs to be sampled in turn, and the distribution parameters that have not been updated use the values ​​of the last sample, as follows:

[0070] a i ~ p ( a │ s i-1 , y N )

[0071] s i ~ p ( s │ a i-1 , y N )

[0072] a i+1 ~ p ( a │ s i , y N )

[0073] s i+1 ~ p ( s │ a i , y N )

[0074] in, a i , a i+1 and a i-1 Respectively i , No. i +1 and i -1st estimated distribution parameters at two quality levels for iteration; s i , s i+1 and s i-1 Respectively i , No. i +1 and i The second estimated distribution parameters at two quality levels for -1 iteration.

[0075] like Figure 3 (a) and Figure 3 As shown in (b), the distribution parameters of the final two-level scheme converge to: a =0.94, s =0.12.

[0076] When each cage of broiler chickens is divided into three quality levels, the weight of each cage of broiler chickens is first constructed as a weight data set, and then the posterior probability of the estimated distribution parameters under the three quality levels is obtained by using the Bayesian theorem.

[0077] The posterior probability of the estimated distribution parameters under the three quality levels is as follows:

[0078] p ( a,a,s │ y N )∝ p ( y 1│ a,a,s )… p ( y n │ a,a,s ) p ( α )p ( a ) p ( s )

[0079] ∝ f 2( y 1, a,a,s )… f 2( y n , a,a,s ) p ( α ) p ( a ) p ( s )

[0080] in, y 1 and y n The weight data sets are y N The 1st and 2nd n Weight of broiler chickens in cages.

[0081] The posterior probability of the estimated distribution parameters under three quality levels is iterated using the Metropolis-Hastings algorithm to obtain the estimated distribution parameters under three quality levels.

[0082] When obtaining the estimated distribution parameters at the third quality level, the Metropolis-Hastings algorithm is used to construct a posterior probability p ( a,a,s │ y N ) is a Markov chain with limiting distribution. First, assume that the number of iterations is t When the Markov chain is located at the coordinates in the parameter space ( α t ,a t ,s t ), and then, by applying a small perturbation to the Markov chain, it moves to the new coordinates ( α t ´ ,a t ´ ,s t ´), if the posterior probability at the new coordinate is higher than the posterior probability at the original coordinate, p ( α t ´ ,a t ´ ,s t ´│ y N )> p ( a,a,s │ y N ), then after this iteration the Markov chain will move directly to the new coordinates, ( α t+1 ,a t+1 ,s t+1 )=( α t ´ ,a t ´ ,s t ´), otherwise the Markov chain will be q Advance to new coordinates, probability q The details are as follows: q = p ( α t ´ ,a t ´ ,s t ´│ y N ) / p ( α t ,a t ,s t │ y N ), probability q represents the degree of reduction of the new posterior probability compared to the original posterior probability; if the move is rejected, the Markov chain will stay at the original coordinates, ( α t+1 ,a t+1 ,s t+1 )=( α t ,a t ,s t ).

[0083] according to Figure 4 (a) Figure 4 (b) Figure 4 (c) Figure 4 (d) Figure 4 (e) and Figure 4 As shown in (f), the distribution parameters of the three-level scheme converge to: α 1=14.2, α 2=2.89, a 1=0.05, a 2=0.02, s1=0.25, s 2=0.06.

[0084] The estimated distribution parameters are input into the chicken group redistribution weight measurement model for processing. After processing, the chicken group redistribution weight measurement model outputs the weight distribution of the graded broiler chickens, such as Figure 5 As shown in (b), the weight estimation distribution of the two-level scheme can be obtained, and the actual value sampling is substituted to obtain Figure 5 The actual weight distribution is shown in (a), the estimated weight distribution and the actual weight distribution of the three-level scheme are shown as Figure 5 (d) and Figure 5 As shown in (c), the above data constitutes a virtual sampling data set. In addition, in order to verify the effectiveness of the inference model under non-ideal conditions, the actual weight distribution of broiler chickens after grading in a batch was obtained from a broiler sales platform. Figure 5 As shown in (e), the distribution parameters are estimated by the two-level scheme and substituted into the model to obtain the estimated weight distribution as follows Figure 5 As shown in (f).

[0085] In terms of qualitative Figure 5 The actual weight distribution in the left column and the estimated weight distribution in the right column have intuitive similarities; quantitatively, the Kolmogorov-Smirnov test (KS test) is used to statistically test whether the two distributions come from the same distribution. It is found that for the data set derived from virtual sampling p The values ​​are all greater than 0.05, indicating that there is no significant difference between the two sides of the distribution at the 95% confidence level. Further analysis depends on the quantile-quantile plot (qq plot). Figure 5 (a) Figure 5 (c) and Figure 5 As shown in the qq diagram of (e), the actual distribution is taken as the theoretical distribution and the estimated distribution is drawn as the sample distribution, and the result is the 45° reference line; Figure 5 (b) Figure 5 (d) and Figure 5 As shown in the qq diagram of (f), the actual distribution is taken as the theoretical distribution and the estimated distribution is taken as the sample distribution, and the result is the corresponding quantile line. Figure 5 It can be found that the quantile lines of the data sets derived from virtual sampling are all located near the reference line. Although the quantile lines of the data sets derived from the broiler sales platform have large fluctuations locally, they are still located near the reference line as a whole. Therefore, although the actual distribution and estimated distribution of the data sets derived from the broiler sales platform cannot pass the KS test, the two still show a certain statistical correlation at the QQ graph level, indicating that the method and device of the present invention are basically effective under non-ideal conditions; as shown in Table 2, the KS test results of the actual distribution and estimated distribution of data sets from different sources are shown.

[0086] Table 2

[0087]

[0088] like Figure 2 As shown, when the weight distribution of the chicken group is measured, the handheld distribution weighing device is held by the handheld portion 3 of the device, and the weight of each cage of broilers in each quality grade is input into the handheld distribution weighing device through the input area 2 on the device shell 4 of the handheld distribution weighing device. After processing, the handheld distribution weighing device outputs the graded broiler weight distribution in its own display area 1, and the average weight, standard deviation and uniformity are obtained after preprocessing, and finally the distribution weighing of the chicken group is achieved. The average weight, standard deviation and uniformity of the graded broiler weight distribution can be used to further guide poultry production and breeding and adjust breeding strategies.

Claims

1. A method for measuring the weight distribution of a chicken group based on poultry grading weighing, characterized in that: include: Step S1: weighing each broiler in a chicken farm by cage, and grading the quality of the broilers into several quality grades; Step S2: inputting the weight of each cage of broilers in each quality grade into a handheld distribution weight measuring device for processing, obtaining the weight distribution of each cage of broilers before grading and the weight difference before and after the weight dividing line after grading according to the weight of each cage of broilers in each quality grade, and then establishing a chicken group redistribution weight measuring model after linear superposition of the distribution; Step S3: constructing the weight of broilers in each cage into a weight data set, performing sampling and iterative processing on the weight data set to obtain the posterior probability of the weight data set, and then obtaining the estimated distribution parameters according to the posterior probability; Step S4: inputting the estimated distribution parameters into the chicken group redistribution weight measurement model for processing, the chicken group redistribution weight measurement model outputs the graded broiler weight distribution after processing and displays it on the handheld distribution weight measurement device after preprocessing, and finally realizing the distribution weight measurement of the chicken group weight; In step S2, when each cage of broiler chickens is divided into two quality levels, it is as follows: Firstly, according to the weight of each cage of broiler chickens in each quality grade, the weight distribution of each cage of broiler chickens before grading and the weight distribution before and after the weight dividing line of the two quality grades after grading are obtained by using Gaussian distribution, and the weight difference before and after the weight dividing line after grading is obtained according to the weight distribution before and after the weight dividing line of the two quality grades after grading; then, the weight distribution of each cage of broiler chickens before grading and the weight difference before and after the weight dividing line after grading are linearly superimposed to establish a chicken group redistribution weight measurement model under two quality grades; The chicken group redistribution weight measurement model under the two-level quality level is as follows: f 1( m , a , σ )=(2 a -1) N ( m │ μ , σ 2 )+(1- a ) N ( m │0.6 μ , σ 2 ) in, f 1() is the weight distribution of broiler chickens after classification under two quality levels; m is the body weight of broilers in each cage; a and σ are the first and second estimated distribution parameters under two quality levels respectively; N () is Gaussian distribution; μ is the mean parameter of Gaussian distribution; The quality dividing line between the first and second levels in a two-level quality scale μ 1, use Gaussian distribution to obtain the weight distribution, and then obtain the weight difference before and after the graded weight dividing line△ m a as follows: △ m a =N (2 μ 1- m 1│ μ , σ 2 )- N ( m 2│ μ , σ 2 ) in, m 1 and m 2 are the weight of broilers in each cage in the first and second quality grades of the two quality grades respectively; In step S2, when each cage of broiler chickens is divided into three quality levels, it is as follows: Firstly, the weight distribution of each cage of broilers in each quality grade is obtained by using Beta distribution according to the weight of each cage of broilers in each quality grade before grading and the weight distribution before and after the two weight dividing lines of the three quality grades after grading, and the weight difference before and after the two weight dividing lines after grading is obtained according to the weight distribution before and after the two weight dividing lines of the three quality grades after grading; then the weight difference before and after the two weight dividing lines after grading is multiplied by a Gaussian distribution with the two weight dividing lines of the three quality grades as the mean, and then the weight distribution of each cage of broilers before grading is linearly superimposed to establish a chicken group redistribution weight measurement model under the three quality grades; The chicken group redistribution weight measurement model under the three quality levels is as follows: f 2( m ´, α,a,σ )= Beta ( μ 1≤ m ´≤ μ 2│ α )- a 1[ Beta ( m ´│ α )- Beta (2 μ 1- m ´│ α )] N ( m ´≤ μ 2│ μ 1, σ 1 2 )- a 2[ Beta ( m ´│ α )- Beta (2 μ 2- m ´│ α )] N ( μ 1≤ m ´│ μ 2, σ 2 2 ) α =( α 1, α 2)， a =( a 1, a 2)， σ =( σ 1, σ 2) in, f 2() is the weight distribution of broiler chickens after classification under three quality levels; m ´ is the weight of broilers in each cage m The scaled weight, m ´= m / 1.6; α , a and σ are the first, second and third estimated distribution parameter vectors under three quality levels respectively; Beta () is Beta distribution; μ 1 and μ 2 are the weight dividing lines between the second and third grades and between the first and second grades under the three-grade quality level; α 1. α 2. a 1. a 2. σ 1 and σ 2 are the first, second, third, fourth, fifth and sixth estimated distribution parameters under three quality levels respectively; N () is Gaussian distribution; The quality dividing line between Grade II and Grade III in the three-grade quality system μ 1, use Beta distribution to obtain the weight distribution, and then obtain the weight difference before and after the weight dividing line of level II and level III m b as follows: △ m b =Beta (2 μ 1- m 3│ α 1, α 2)- Beta ( m 2│ α 1, α 2) The quality dividing line between Grade I and Grade II in the three-grade quality system μ 2, use the Beta distribution to obtain the weight distribution, and then obtain the weight difference before and after the weight dividing line of level I and level II. m c as follows: △ m c =Beta (2 μ 2- m 4│ α 1, α 2)- Beta ( m 3│ α 1, α 2) in, Beta ( ) is Beta distribution; m 4. m 3 and m 2Respectively, it is the body weight of broilers in each cage in the three quality grades of grade I, grade II and grade III.

2. The chicken group weight distribution measurement method based on poultry grading weighing according to claim 1 is characterized in that: In the step S1, the broilers in the chicken farm are evenly divided according to the number and placed in N chicken cages, and then the broilers in each cage are weighed. According to the weight of the broilers in each cage, the broilers in each cage are divided into two or three quality grades from high to low according to the living quality grade, thereby completing the quality grade.

3. The chicken group weight distribution measurement method based on poultry grading weighing according to claim 1 is characterized in that: In step S3, when each cage of broiler chickens is divided into two quality levels, the weight of each cage of broiler chickens is first constructed as a weight data set, and then Gibbs sampling is used to obtain the posterior probability of the estimated distribution parameters under the two quality levels.

4. The chicken group weight distribution measurement method based on poultry grading weighing according to claim 3 is characterized in that: In the step S3, the posterior probabilities of the estimated distribution parameters at the two quality levels are iterated using Markov Monte Carlo to obtain the estimated distribution parameters at the two quality levels.

5. The chicken group weight distribution measurement method based on poultry grading weighing according to claim 1 is characterized in that: In the step S3, when each cage of broiler chickens is divided into three quality levels, the weight of each cage of broiler chickens is first constructed as a weight data set, and then the posterior probability of the estimated distribution parameters under the three quality levels is obtained by using the Bayesian theorem; the posterior probability of the estimated distribution parameters under the three quality levels is iterated using the Metropolis-Hastings algorithm to obtain the estimated distribution parameters under the three quality levels.

6. The chicken group weight distribution measurement method based on poultry grading weighing according to claim 1 is characterized in that: In step S4, the weight distribution of the graded broilers is preprocessed to obtain the average weight, standard deviation and uniformity.

Images