Feed feeding control method and system for sheep raised in mountainous region

By establishing a numbered database for mountain sheep farming and utilizing machine vision and deep learning technologies, precise feed feeding control was achieved, solving the shortcomings of feed management in mountain sheep farming and improving feed conversion rate and breeding efficiency.

CN118915536AInactive Publication Date: 2024-11-08GUIZHOU BREEDING LIVESTOCK & POULTRY GERMPLASM TESTING CENT
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410973240.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-19
Publication Date
2024-11-08
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In existing technologies, it is difficult to achieve precise management of feed feeding for sheep raised in mountainous areas, which fails to meet the precise nutritional needs of sheep and results in low breeding efficiency.

Method used

By assigning a unique number to each sheep and establishing a number database, machine vision is used to identify the sheep. Combined with deep learning and neural network analysis of the sheep's breeding information and feeding videos, precise feed control is achieved, including nutritional requirement prediction and mental state assessment, to ensure optimal feed conversion rate.

Benefits of technology

It enables intelligent feed management for mountain sheep flocks, improves feed conversion rate, reduces breeding costs, and provides precise feed feeding strategies to promote the healthy growth of sheep.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118915536A_ABST
    Figure CN118915536A_ABST
Patent Text Reader

Abstract

The invention provides a feed feeding control method and system for sheep raised in a mountainous region, and the method comprises the following steps: wearing a unique number for each sheep in a sheep flock raised in the mountainous region, building a number database, and inputting the breeding information of each sheep into the number database; machine vision is used for identifying the serial number, and breeding information of the sheep corresponding to the current serial number is analyzed; feed and drinking water with corresponding weight and proportion are provided for the current ingestion sheep; acquiring and analyzing a sheep ingestion video, acquiring a mental state of a current ingestion sheep, and inputting a mental state analysis result into a serial number database; and monitoring the ingestion data of the current ingestion sheep, obtaining the growth data of the current ingestion sheep in combination with the mental state analysis result, and collecting and inputting the growth data into the numbering database to complete the accurate control of the feed feeding for the sheep raised in the mountainous region. According to the invention, through precise feeding of the sheep flock bred in the mountainous region, the growth condition of the sheep is monitored in real time, a personalized breeding scheme is provided for breeding, and the feed conversion rate is ensured.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent breeding, and in particular relates to a feed feeding control method and system for mountain sheep breeding. Background Art

[0002] Raising sheep in mountainous environments faces many challenges, among which feed feeding control is one of the key factors to ensure the healthy growth of sheep and improve breeding efficiency. Accurate feed feeding control can not only meet the nutritional needs of sheep, but also promote the growth and development of goats and reduce breeding costs. However, in the existing technology, there are many deficiencies in the feeding control of sheep feed, making it difficult to achieve refined feed management for sheep. Summary of the invention

[0003] In view of the shortcomings of the prior art, the present invention provides a feed feeding control method and system for mountain sheep farming. Through intelligent feeding of mountain sheep flocks, the growth status of sheep is monitored in real time, personalized feeding plans are provided for farming, and feed conversion rate is guaranteed.

[0004] To achieve the above object, the present invention provides the following solutions:

[0005] A feed feeding control method for mountain sheep farming comprises the following steps:

[0006] Each sheep in the mountain breeding flock is given a unique number, and a number database is established, and the breeding information of each sheep is entered into the number database;

[0007] Using machine vision to identify the number, and analyzing the breeding information of the sheep corresponding to the current number based on the number identification result;

[0008] Based on the results of the breeding information analysis, provide the sheep with the corresponding amount and proportion of feed and drinking water;

[0009] Acquire and analyze the sheep feeding video to obtain the mental state of the sheep currently feeding, and enter the mental state analysis results into the numbering database;

[0010] The feeding data of the current feeding sheep is monitored, and the growth data of the current feeding sheep is obtained in combination with the mental state analysis result, and the growth data is collected and entered into the numbering database to complete the feed feeding control for mountain sheep farming.

[0011] Preferably, the breeding information includes sheep breed, age, gender, vaccination status, feed requirements, historical growth data and environmental information.

[0012] Preferably, the method of using machine vision to identify the number is:

[0013] Collecting numbered images of the sheep currently eating, and enhancing the numbered images to obtain enhanced numbered images;

[0014] Based on a multi-layer region recommendation network, multi-scale target detection is performed on the enhanced numbered image, and based on a clustering algorithm, optimal anchor box parameters on each detection layer of the multi-layer region recommendation network are obtained;

[0015] Based on the optimal anchor frame parameters, a GIoU-based border regression loss function is calculated to complete the construction of the serial number detection model;

[0016] The number detection model is trained based on the existing training samples, and the number of the current feeding sheep is identified based on the trained number detection model to obtain a number recognition result.

[0017] Preferably, based on the results of the breeding information analysis, the method of providing the current feeding sheep with feed and drinking water in corresponding amounts and proportions is:

[0018] Preprocessing the breeding information, filling in missing values, and obtaining the preprocessed breeding information;

[0019] Based on the preprocessed breeding information, feature classification is performed, and the classified features are uniquely encoded to obtain the classification features of the breeding information;

[0020] Using the classification features, a nutritional demand prediction model for sheep is constructed based on a deep convolutional neural network and a long short-term memory neural network;

[0021] Based on the nutritional demand prediction model, predict the nutritional demand of the current feeding sheep;

[0022] Based on the pre-processed breeding information and the nutritional requirements, evaluating the health status of the current feeding sheep;

[0023] Based on the nutritional requirements and health status assessment results, provide the current feeding sheep with feed and drinking water in corresponding amounts and proportions.

[0024] Preferably, the method for obtaining the mental state of the current feeding sheep is:

[0025] Collect video data of current sheep eating;

[0026] Based on the spatial stream network, the existing sheep feeding video data is processed to learn the static features of the sheep feeding images;

[0027] Based on the temporal flow network, the existing sheep feeding video data is processed to learn the dynamic features of the sheep feeding video;

[0028] Based on a dual-scale neural network, the static features of the image and the dynamic features of the video are integrated to obtain the spatiotemporal fusion features of sheep feeding;

[0029] Based on the spatiotemporal fusion characteristics, the mental state levels of the sheep currently feeding are divided.

[0030] Preferably, the feeding data of the current feeding sheep include feed intake, water intake, feed surplus and drinking water surplus; the surplus feed and drinking water are recovered in time.

[0031] The present invention also provides a feed feeding control system for mountain sheep farming, which is used to implement the method described, including:

[0032] A numbering module is used to give each sheep in the mountain breeding flock a unique number, establish a numbering database, and enter the breeding information of each sheep into the numbering database;

[0033] A number recognition module, used to recognize the number using machine vision, and analyze the breeding information of the sheep corresponding to the current number based on the number recognition result;

[0034] The feeding module is used to provide the current sheep with feed and drinking water in the appropriate amount and proportion based on the results of breeding information analysis;

[0035] A feeding monitoring module is used to obtain and analyze the sheep feeding video, obtain the mental state of the current feeding sheep, and enter the mental state analysis results into the numbering database;

[0036] The numbered database update module is used to monitor the feeding data of the current feeding sheep, and in combination with the mental state analysis results, obtain the growth data of the current feeding sheep, and collect and enter the growth data into the numbered database to complete the feed feeding control for mountain sheep farming.

[0037] Preferably, the number identification module includes:

[0038] An image enhancement unit, used for collecting numbered images of the sheep currently eating, enhancing the numbered images, and obtaining enhanced numbered images;

[0039] An object detection unit, configured to perform multi-scale object detection on the enhanced numbered image based on a multi-layer region recommendation network, and obtain optimal anchor frame parameters on each detection layer of the multi-layer region recommendation network based on a clustering algorithm;

[0040] A model building unit, used to calculate a GIoU-based frame regression loss function based on the optimal anchor frame parameters to complete the construction of a serial number detection model;

[0041] The number recognition unit is used to train the number detection model based on the existing training samples, and recognize the number of the current feeding sheep based on the trained number detection model to obtain a number recognition result.

[0042] Compared with the prior art, the beneficial effects of the present invention are as follows: for the feed feeding control method of mountain sheep farming, a sheep flock database is established by numbering, and the feeding data and growth data of each sheep are recorded. Through intelligent analysis of the data, the feed intake of each sheep is controlled to ensure the optimal feed conversion rate. The technical solution of the present invention realizes the intelligent and precise management and feeding of feed for mountain sheep farming, and timely recovers the feed to reduce costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0044] Figure 1 The present invention is a flow chart of a feed feeding control method for mountain sheep farming according to an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0047] Embodiment 1

[0048] like Figure 1 As shown, a feed feeding control method for mountain sheep farming includes the following steps:

[0049] S1: Each sheep in the mountainous sheep flock is given a unique number, and a number database is established, and the breeding information of each sheep is entered into the number database; a further implementation method is that the breeding information includes the sheep breed, age, gender, vaccination status, feed requirements, historical growth data and environmental information. The environmental information is the mountainous environmental information.

[0050] S2: Using machine vision to identify the number, and analyzing the breeding information of the sheep corresponding to the current number based on the number identification result;

[0051] A further implementation method is that the method of using machine vision to identify the number is:

[0052] The numbered images of the currently eating sheep are collected, and the numbered images are enhanced to obtain enhanced numbered images; in this embodiment, the collected numbered images are enhanced in consideration of the brightness of mountain weather and the fact that the sheep are mostly in a non-stationary state. Specifically, the Retinex algorithm is used to estimate the illumination in the collected original numbered images, obtain an illumination image, and decompose a reflection image, and the product of the illumination image and the reflection image is used to represent the numbered image, and the signal-to-noise ratio of the numbered image is calculated. When the signal-to-noise ratio is lower than a preset threshold, the numbered image is enhanced.

[0053] Based on the multi-layer region recommendation network, multi-scale target detection is performed on the enhanced numbered image, and based on the clustering algorithm, the optimal anchor frame parameters on each detection layer of the multi-layer region recommendation network are obtained; in this embodiment, a hierarchical multi-scale region recommendation network of FPN is used, and the region recommendation network selects the ResNetl01 network as the backbone network, and the numbered features are extracted layer by layer from the bottom to the top. The number of channels of the four scale feature maps of the encoded image is uniformly reduced to 256 through a 1x1 convolution operation, and the high-level feature map with strong semantic information is fused with the high-resolution feature map of the previous layer by two times upsampling. The fusion method is pixel-by-pixel addition, which not only utilizes the strong semantic features of the high-level feature map, but also utilizes the high-resolution information of the low-level feature map. The optimal anchor frame parameters include size, aspect ratio and quantity.

[0054] Based on the optimal anchor box parameters, the border regression loss function based on GIoU is calculated to complete the construction of the number detection model. In this embodiment, GIoU obtains IoU, the minimum closure area, the proportion of the closure area that does not belong to the calibrated bounding box and the anchor box area in the closure area by calculating the two-point coordinates of the anchor box and the two-point coordinates of the calibrated bounding box, and finally subtracts the proportion from IoU to obtain GIoU.

[0055] The number detection model is trained based on the existing training samples, and the number of the current feeding sheep is identified based on the trained number detection model to obtain a number recognition result.

[0056] S3: Based on the results of the breeding information analysis, feed and drinking water of corresponding amounts and proportions are provided to the current feeding sheep; a further implementation method is that the method of providing feed and drinking water of corresponding amounts and proportions to the current feeding sheep is:

[0057] Preprocess the breeding information, fill in the missing values, and obtain the preprocessed breeding information;

[0058] Based on the preprocessed breeding information, feature classification is performed, and the classified features are uniquely encoded to obtain the classification features of the breeding information; specifically, the breeding information is classified into time series information vectors, spatial distribution information vectors, and combination information;

[0059] Among them, the time series information includes continuous feed quantity records, from which the time characteristics of breeding information, i.e., timestamp, date, etc., are extracted;

[0060] Spatial distribution information includes temperature and humidity in mountainous environments.

[0061] Combined information average feed amount and feed change rate etc.

[0062] Using classification features, a nutritional demand prediction model for sheep is constructed based on a deep convolutional neural network (CNN) and a long short-term memory neural network (LSTM); in this embodiment, two fully connected layers are set to connect the output results of the deep convolutional neural network and the long short-term memory neural network, the activation function of the first fully connected layer is set to ReLu, and the activation function of the second fully connected layer is set to Linear. Specifically, the classification features are input into an embedding layer of CNN to obtain a feature vector of fixed length. The output of the embedding layer is merged with the numerical features of the breeding information, and then input into one or more fully connected layers to obtain a first output. The time series data in the first output is input into the LSTM layer, and then the output of the LSTM is merged with the first output to obtain a merged feature vector. Finally, the merged feature vector is input into two fully connected layers to generate the final nutritional demand prediction.

[0063] Based on the nutritional demand prediction model, predict the nutritional needs of the current feeding sheep;

[0064] Evaluate the health status of the current feeding sheep based on the pre-processed breeding information and nutritional requirements;

[0065] Based on the nutritional needs and health status assessment results, provide the sheep currently eating with appropriate amounts and proportions of feed and drinking water.

[0066] S4: Obtain and analyze the sheep feeding video, obtain the mental state of the current feeding sheep, and enter the mental state analysis results into the numbering database;

[0067] A further implementation method is that the method for obtaining the mental state of the current feeding sheep is:

[0068] Collect video data of current sheep eating;

[0069] Based on the spatial stream network, the existing sheep feeding video data is processed to learn the static features of the sheep feeding images;

[0070] Based on the temporal stream network, the existing sheep feeding video data is processed to learn the dynamic characteristics of the sheep feeding video;

[0071] Based on a dual-scale neural network, the image static features and the video dynamic features are fused to obtain the spatiotemporal fusion features of sheep feeding; in this embodiment, the dual-scale neural network is the time scale and the space scale. Among them, the RGB feature generation network (spatial flow network) and the Flow feature generation network (temporal flow network) are time domain segmentation networks or three-dimensional convolutional neural networks. In the cross-correlation fusion network, the cross-correlation module is used to fuse the image static features and video dynamic features generated by the sub-network, and then the residual module is used to enhance the spatiotemporal representation ability of the fused features, and it is used as the input of the fusion module of the next layer.

[0072] Based on the spatiotemporal fusion characteristics, the mental state levels of the current feeding sheep are divided.

[0073] S5: Monitor the feeding data of the current sheep, and combine it with the results of the mental state analysis to obtain the growth data of the current sheep, and collect the growth data into the numbering database to complete the feed feeding control for mountain sheep farming. Continuously update the growth data to achieve accurate feed feeding.

[0074] A further implementation method is that the current feeding data of the sheep includes feed intake, water intake, feed surplus and drinking water surplus; the surplus feed and drinking water are recycled in time, and the water quality is monitored in real time to prevent drinking water pollution.

[0075] This embodiment also includes drawing a growth curve of the sheep based on the collected sheep growth data; analyzing the optimal growth cycle and feeding cycle of the sheep based on the growth curve, and generating a feed adjustment strategy to achieve precision feeding.

[0076] Specifically, the growth data of all sheep are cleaned, the cleaned growth data are analyzed, the data are classified according to the sheep breed, gender and feeding conditions, and a Logistic growth curve model is established;

[0077] The maximum likelihood method was used to fit the growth rate of the sheep and the most suitable body weight they could achieve in the Logistic growth curve model to obtain the fitting model.

[0078] Perform residual analysis on the fitted model to evaluate the fitting effect of the model. When the effect threshold is met, the optimal fitting model is obtained.

[0079] Based on the fitting model, draw the sheep growth prediction curve;

[0080] Analyze the growth prediction curve of sheep to obtain the growth rate of sheep at different time points. Based on the change trend of the growth rate, identify the time point when the growth rate begins to decline significantly. This time point is taken as the inflection point of the growth rate, indicating that the optimal growth cycle of the current sheep has ended. The time period from the birth of the sheep to the inflection point of the growth rate is defined as the optimal growth cycle.

[0081] The optimal growth cycle and feeding cycle of each sheep are obtained, and a separate feed adjustment strategy is formulated for each sheep. Automatic feeding is performed based on the feed adjustment strategy and the sheep information obtained in real time in steps S2-S5 to achieve precision feeding.

[0082] Embodiment 2

[0083] The present invention also provides a feed feeding control system for mountain sheep farming, and a method for implementing the control system includes:

[0084] The numbering module is used to give each sheep in the mountain breeding flock a unique number, establish a numbering database, and enter the breeding information of each sheep into the numbering database;

[0085] A number recognition module is used to use machine vision to recognize the number and analyze the breeding information of the sheep corresponding to the current number based on the number recognition result;

[0086] The feeding module is used to provide the current sheep with feed and drinking water in the appropriate amount and proportion based on the results of breeding information analysis;

[0087] The feeding monitoring module is used to obtain and analyze the sheep feeding video, obtain the mental state of the current feeding sheep, and enter the mental state analysis results into the numbering database;

[0088] The numbered database update module is used to monitor the feeding data of the current sheep, and combine it with the mental state analysis results to obtain the growth data of the current sheep, and collect the growth data into the numbered database to complete the feed feeding control for mountain sheep farming.

[0089] In a further embodiment, the number identification module comprises:

[0090] An image enhancement unit is used to collect the numbered images of the sheep currently eating, enhance the numbered images, and obtain enhanced numbered images;

[0091] The target detection unit is used to perform multi-scale target detection on the enhanced numbered image based on the multi-layer region recommendation network, and obtain the optimal anchor box parameters on each detection layer of the multi-layer region recommendation network based on the clustering algorithm;

[0092] The model building unit is used to calculate the GIoU-based bounding box regression loss function based on the optimal anchor box parameters to complete the construction of the number detection model;

[0093] The number recognition unit is used to train a number detection model based on existing training samples, and to recognize the number of the current feeding sheep based on the trained number detection model to obtain a number recognition result.

[0094] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for controlling feed feeding for mountain sheep farming, characterized in that: The following steps are involved: Each sheep in the mountain breeding flock is given a unique number, and a number database is established, and the breeding information of each sheep is entered into the number database; Using machine vision to identify the number, and analyzing the breeding information of the sheep corresponding to the current number based on the number identification result; Based on the results of the breeding information analysis, provide the sheep with the corresponding amount and proportion of feed and drinking water; Acquire and analyze the sheep feeding video to obtain the mental state of the sheep currently feeding, and enter the mental state analysis results into the numbering database; The feeding data of the current feeding sheep is monitored, and the growth data of the current feeding sheep is obtained in combination with the mental state analysis result, and the growth data is collected and entered into the numbering database to complete the feed feeding control for mountain sheep farming.

2. The feed feeding control method for mountain sheep farming according to claim 1, characterized in that: The breeding information includes sheep breed, age, gender, vaccination status, feed requirements, historical growth data and environmental information.

3. The feed feeding control method for mountain sheep farming according to claim 1, characterized in that: The method of using machine vision to identify the number is: Collecting numbered images of the sheep currently eating, and enhancing the numbered images to obtain enhanced numbered images; Based on a multi-layer region recommendation network, multi-scale target detection is performed on the enhanced numbered image, and based on a clustering algorithm, optimal anchor box parameters on each detection layer of the multi-layer region recommendation network are obtained; Based on the optimal anchor frame parameters, a GIoU-based border regression loss function is calculated to complete the construction of the serial number detection model; The number detection model is trained based on the existing training samples, and the number of the current feeding sheep is identified based on the trained number detection model to obtain a number recognition result.

4. The feed feeding control method for mountain sheep farming according to claim 1, characterized in that: Based on the results of the breeding information analysis, the method of providing the current sheep with feed and drinking water in the corresponding amount and proportion is as follows: Preprocessing the breeding information, filling in missing values, and obtaining the preprocessed breeding information; Based on the preprocessed breeding information, feature classification is performed, and the classified features are uniquely encoded to obtain the classification features of the breeding information; Using the classification features, a nutritional demand prediction model for sheep is constructed based on a deep convolutional neural network and a long short-term memory neural network; Based on the nutritional demand prediction model, predict the nutritional demand of the current feeding sheep; Based on the pre-processed breeding information and the nutritional requirements, evaluating the health status of the current feeding sheep; Based on the nutritional requirements and health status assessment results, provide the current feeding sheep with feed and drinking water in corresponding amounts and proportions.

5. The feed feeding control method for mountain sheep farming according to claim 1, characterized in that: The method to obtain the mental state of the current feeding sheep is: Collect video data of current sheep eating; Based on the spatial stream network, the existing sheep feeding video data is processed to learn the static features of the sheep feeding images; Based on the temporal flow network, the existing sheep feeding video data is processed to learn the dynamic features of the sheep feeding video; Based on a dual-scale neural network, the static features of the image and the dynamic features of the video are integrated to obtain the spatiotemporal fusion features of sheep feeding; Based on the spatiotemporal fusion characteristics, the mental state levels of the currently feeding sheep are divided.

6. The feed feeding control method for mountain sheep farming according to claim 1, characterized in that: The current feeding data of the sheep include feed intake, water intake, feed surplus and drinking water surplus; the surplus feed and drinking water are recovered in time.

7. A feed feeding control system for mountain sheep farming, used to implement the method described in any one of claims 1 to 6, characterized in that: include: A numbering module is used to give each sheep in the mountain breeding flock a unique number, establish a numbering database, and enter the breeding information of each sheep into the numbering database; A number recognition module, used to recognize the number by machine vision, and analyze the breeding information of the sheep corresponding to the current number based on the number recognition result; The feeding module is used to provide the current sheep with feed and drinking water in the appropriate amount and proportion based on the results of breeding information analysis; A feeding monitoring module is used to obtain and analyze the sheep feeding video, obtain the mental state of the current feeding sheep, and enter the mental state analysis results into the numbering database; The numbered database update module is used to monitor the feeding data of the current feeding sheep, and in combination with the mental state analysis results, obtain the growth data of the current feeding sheep, and collect and enter the growth data into the numbered database to complete the feed feeding control for mountain sheep farming.

8. The feed feeding control system for mountain sheep farming according to claim 7, characterized in that: The number recognition module comprises: An image enhancement unit, used for collecting numbered images of the sheep currently eating, enhancing the numbered images, and obtaining enhanced numbered images; An object detection unit, configured to perform multi-scale object detection on the enhanced numbered image based on a multi-layer region recommendation network, and obtain optimal anchor frame parameters on each detection layer of the multi-layer region recommendation network based on a clustering algorithm; A model building unit, used to calculate a GIoU-based frame regression loss function based on the optimal anchor frame parameters to complete the construction of a serial number detection model; The number recognition unit is used to train the number detection model based on the existing training samples, and recognize the number of the current feeding sheep based on the trained number detection model to obtain a number recognition result.