A smart breeding monitoring method and system based on big data

Abstract

The present invention relates to the field of breeding monitoring technology, and specifically to a smart breeding monitoring method and system based on big data. After receiving the breeding data summarized by the breeding farm terminal, the smart breeding management platform uses natural language processing technology to extract key parameters; classifies the key parameters, analyzes the key parameters of the same type, and outputs a first evaluation report; generates a breeding plan recommendation based on the first evaluation report, transmits it to the terminal and the feed-to-weight ratio prediction model, and obtains the first feed-to-weight ratio; after the breeding farm starts to execute the breeding plan recommendation displayed by the breeding farm terminal, it receives new breeding data and outputs a second evaluation report; at the same time, calculates the similarity between the first evaluation report and the second evaluation report; extracts the feed weight from the new breeding data, and calculates the second feed-to-weight ratio based on the feed weight and the received breeding weight; outputs the similarity, the ratio of the second feed-to-weight ratio to the first feed-to-weight ratio to the first evaluation report, and improves product quality and market qualification rate.

Description

Technical Field

[0001] The present invention relates to the technical field of aquaculture monitoring, and in particular to a smart aquaculture monitoring method and system based on big data. Background Art

[0002] In the process of agricultural modernization, the transformation and upgrading of the livestock and poultry farming industry has brought about a dual improvement in efficiency and quality, but its high cost has become an obstacle to popularization. For some franchised breeders and small-scale farmers, the equipment investment, operating expenses and professional technical requirements required for modern breeding are beyond their ability to bear. In addition, the conditions of the farms are limited, and it is difficult to continuously maintain the best environment for poultry breeding, which further affects the breeding effect. At the same time, differences in geography, region and input costs, coupled with different concerns and investment capabilities for environmental protection, have led to a diversity of breeding plans adopted by different individual breeders. However, due to problems such as insufficient professional breeding knowledge and poor overall market data analysis, these plans often have major defects, which increase the uncertainty of breeding and affect the feed-to-weight ratio and the qualified rate of market output.

[0003] In response to the above problems, this application proposes a smart farming monitoring method based on big data, aiming to establish a comprehensive evaluation and management platform to solve the difficulties faced by individual farmers. The platform will comprehensively consider the actual situation of farmers, farming environment, market demand and environmental protection requirements, and provide farmers with personalized farming plan judgment and guidance through data analysis and model prediction, helping to improve feed conversion rate and feed-to-weight ratio, reduce farming costs, improve product quality and qualified rate of marketed products, and help farmers achieve stable income in a complex and changing farming environment. Summary of the invention

[0004] The purpose of the present invention is to provide a smart breeding monitoring method and system based on big data, in order to address the problem that individual breeders currently have major defects in the breeding process due to their own lack of professional breeding knowledge, resulting in a relatively large weight of the final feed when the animals are sold, affecting individual income. The embodiments of the present application provide a smart breeding monitoring method and system based on big data, provide more accurate breeding plans and management suggestions for individual breeders, help individual breeders to adjust feeding conditions in a timely manner, and achieve increased income.

[0005] In order to achieve the above-mentioned object of the invention, the present invention provides the following technical solutions:

[0006] A smart breeding monitoring method based on big data, comprising:

[0007] The farming data of the farm is collected through the IoT sensor. After the smart farming management platform receives the farming data summarized by the farm terminal, it uses natural language processing technology to extract key parameters; the platform classifies the key parameters, analyzes the key parameters of the same type, and outputs a first evaluation report;

[0008] The platform generates a breeding plan recommendation according to the first evaluation report and transmits it to the breeding farm terminal; at the same time, the platform outputs the breeding plan recommendation to the feed-to-weight ratio prediction model to obtain a first feed-to-weight ratio predicted by the feed-to-weight ratio prediction model according to the breeding plan recommendation;

[0009] After the farm starts to implement the breeding plan recommendation displayed on the farm terminal, the platform receives the new breeding data and outputs a second evaluation report; at the same time, the platform calculates the similarity between the first evaluation report and the second evaluation report;

[0010] When the farmed animals are released to market, the platform extracts the feed weight from the new farming data, and calculates the second feed-to-weight ratio based on the feed weight and the received farmed animal release weight; the platform outputs the similarity and the ratio of the second feed-to-weight ratio to the first feed-to-weight ratio to the first evaluation report, which is used to guide the generation of the next farming plan recommendation.

[0011] As a preferred technical solution of the present application, the classification of the key parameters includes: the platform classifies and stores the key parameters by name, and obtains an overall evaluation report after analyzing the key parameters; the first evaluation report includes the overall evaluation report.

[0012] As a preferred technical solution of the present application, the classification of the key parameters includes: the platform extracts environmental data from the breeding data; the platform classifies the farms according to the geographical information of the farms and the environmental data; the key parameters of the farms of the same category are classified and stored by name, and after analyzing the key parameters of multiple farms, multiple local evaluation reports are obtained; the first evaluation report includes the local evaluation report.

[0013] As a preferred technical solution of the present application, the first evaluation report includes an overall evaluation report and a local evaluation report, a first evaluation weight is set for the overall evaluation report, and a second evaluation weight is set for the local evaluation report; the first evaluation report is obtained by fusing the key parameters corresponding to the overall evaluation report and the local evaluation report.

[0014] As a preferred technical solution of the present application, generating a breeding plan recommendation according to the first evaluation report includes:

[0015] The platform has preset expert weights; the platform sends the first evaluation report to multiple aquaculture experts. After the aquaculture experts formulate a breeding plan based on the first evaluation report, the platform assigns the corresponding expert weights to the received breeding plan, merges the multiple breeding plans with the expert weights, and outputs the breeding plan recommendation.

[0016] As a preferred technical solution of the present application, the breeding data includes feeding data, environmental data, drug data and health status data; when the platform outputs the breeding plan recommendation to the feed-to-weight ratio prediction model, it also includes:

[0017] The platform extracts a feeding management vector of the farm according to the feeding data; extracts an environmental control vector of the farm according to the environmental data; and extracts a disease prevention and control vector of the farm according to the drug data and the health status data;

[0018] The platform uses the feeding management vector, the environmental control vector and the disease prevention and control vector as influencing factors of the breeding plan recommendation, and also outputs them to the feed-to-weight ratio prediction model to obtain the first feed-to-weight ratio.

[0019] As a preferred technical solution of the present application, the platform performs statistical analysis on the feeding data and the environmental data, and after combining the statistical analysis results of multiple data, obtains the corresponding feeding management vector and the environmental control vector.

[0020] As a preferred technical solution of the present application, the method of extracting the disease prevention and control vector of the farm according to the drug data and the health status data includes:

[0021] After statistically analyzing the drug data, combining the analysis results of a plurality of the drug data into a drug vector;

[0022] Collecting weight data, eating behavior data and exercise data, performing statistical analysis on the weight data and the eating behavior data to obtain corresponding weight vectors and eating behavior vectors; performing image analysis on the exercise data to obtain exercise vectors;

[0023] The drug vector, the body weight vector, the dietary behavior vector and the exercise vector are combined into the disease prevention and control vector.

[0024] As a preferred technical solution of the present application, the motion data includes motion trajectory, neck image data, and head image data; the image analysis of the motion data includes:

[0025] Calculating the movement trajectory according to the movement position and movement distance of the poultry;

[0026] Extracting texture features and color features of neck feathers from the neck image data through a pre-trained neural network, and calculating the area of ​​the neck feathers according to the pixels occupied by the neck feathers;

[0027] The eyelid closure features of the eyes and the texture features of the poultry skirt are extracted from the head image data through the neural network to obtain the eyelid closure vector and the skirt extension vector, and the modulus of the eyelid closure vector and the skirt extension vector is calculated.

[0028] This application also provides a smart farming monitoring system based on big data, including the following modules:

[0029] An analysis module is used to receive the farming data collected by the IoT sensor from the farm terminal, extract and output key parameters using natural language processing technology; classify the key parameters, analyze the key parameters of the same type, and output a first evaluation report;

[0030] an application module, configured to generate a breeding plan recommendation according to the first evaluation report, and transmit the result to the breeding farm terminal; and output the breeding plan recommendation to a feed-to-weight ratio prediction model, and receive a first feed-to-weight ratio predicted by the feed-to-weight ratio prediction model according to the breeding plan recommendation;

[0031] A comparison module, after the farm starts to execute the breeding plan recommendation displayed on the farm terminal, is used to receive the new breeding data and output a second evaluation report; calculate and output the similarity between the first evaluation report and the second evaluation report;

[0032] The update module is used to extract the feed weight from the new breeding data when the farmed animals are released, and calculate the second feed-to-weight ratio based on the feed weight and the received farmed animal release weight; the similarity and the ratio of the second feed-to-weight ratio to the first feed-to-weight ratio are output to the first evaluation report to guide the generation of the next breeding plan recommendation.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] 1. By statistically analyzing the farm data and sending the first evaluation report to multiple aquaculture experts, the data specificity problem caused by different environments and different cost inputs of the farms is effectively solved, and the overall control of the data of multiple farms where the platform is located is achieved, which is conducive to comprehensive analysis of the data differences and average levels of individual farmers from a global perspective, and provides different data entry angles for aquaculture experts to specify aquaculture plans. By integrating the aquaculture plans output by multiple aquaculture experts, the final aquaculture plan recommendation is obtained, which realizes the integration of professional knowledge and experience of different experts, reduces the decision-making risks caused by a single point of view or lack of experience, and is conducive to forming a more comprehensive aquaculture plan and improving the adaptability of the aquaculture plan.

[0035] 2. By performing statistical analysis on the real-time data after the farm executes the breeding plan recommendation, and comparing the obtained second evaluation report with the first evaluation report corresponding to the historical data, the execution process of the breeding plan recommendation can be evaluated. By comparing the similarity between the historical breeding plan recommendation and the current breeding plan recommendation, the changes in the breeding data can be intuitively observed, which is conducive to directly evaluating the execution effect of the breeding plan recommendation and providing data support for further optimization of the breeding plan.

[0036] 3. This application not only analyzes all the data collected by the platform from a global perspective, but also realizes the analysis of the benchmark level of the aquaculture industry in the management area corresponding to the platform, and provides a basis for comparative analysis from other perspectives. Affected by geographical conditions, different farms may have small differences in environmental data such as temperature and humidity. Therefore, this application also classifies farms according to geographic information, and analyzes the data collected by the platform from the perspective of geographic information, so that the data differences of farms of the same type are reduced, and the attention to feeding capacity and disease prevention and control capabilities is increased. It is also conducive to the refined and differentiated analysis of different types of farms, so as to realize the adaptation of aquaculture methods to local conditions.

[0037] 4. By setting expert weights for the plans designated by different aquaculture experts, the professional qualities and practical experience of aquaculture experts can be more comprehensively and objectively reflected. With the help of multiple expert attributes, expert weights are determined from different dimensions, which helps to ensure that experts who have in-depth research and practical application of aquaculture data can have greater voice in the process of aquaculture data analysis and aquaculture plan formulation, thereby improving the scientificity and practicality of aquaculture plan recommendations.

[0038] 5. Analyze the weight changes of poultry before and after they become ill, the changes in their eating and drinking water, the amount of exercise during routine behaviors, and the changes in their mental state. This enables the use of the medication effects of poultry to reflect the disease prevention and control capabilities of farmers, which is conducive to forming an objective evaluation of the disease prevention and control capabilities of farmers and improving the accuracy of feed-to-weight ratio predictions.

[0039] 6. Although the breeding plan recommendation issued by the platform has been guided and suggested by experts, the implementation of the breeding plan recommendation and the final effect still depend on the input of the breeder. Therefore, the feed-to-weight ratio predicted by the breeding plan recommendation should be subject to the breeder's feeding management ability, environmental control ability and disease prevention and control ability. Through the analysis of historical data, it is helpful to form the evaluation of the breeder's ability to execute the breeding plan recommendation, improve the accuracy of the prediction results of the feed-to-weight ratio prediction model, and then facilitate targeted implementation guidance for the breeder. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a flowchart of the smart farming monitoring method;

[0041] Figure 2 This is a structural diagram of the smart breeding monitoring system. DETAILED DESCRIPTION

[0042] To make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be described clearly and completely in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them.

[0043] Therefore, the following detailed description of the embodiments of the present invention is not intended to limit the scope of the invention claimed for protection, but merely represents some embodiments of the present invention. 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.

[0044] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features and technical solutions in the embodiments may be combined with each other.

[0045] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0046] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "upper", "lower", etc. is based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the invention product is usually placed when in use, or the orientation or positional relationship commonly understood by those skilled in the art. Such terms are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0047] This application takes poultry farming as an example, and uses the Internet of Things technology to achieve real-time monitoring and data collection of the poultry farming environment. Sensors are used to monitor the temperature, humidity, light, air quality and other environmental parameters in the poultry farm in real time, and the data is transmitted to the platform. The platform analyzes and processes a large amount of real-time data to extract valuable information, such as trends and abnormal conditions of environmental parameters. Based on these data, the platform can provide farmers with more accurate breeding plans and management suggestions, help farmers adjust feeding conditions in time, and optimize the farm environment. The preferred method of data collection in this application is to collect relevant data of all poultry in the farm. For some large farms, due to the large number of poultry and the large amount of data, relevant data can also be collected by sampling.

[0048] Example 1: See Figure 1 As shown, this embodiment provides a smart farming monitoring method based on big data, including:

[0049] The farming data of the farm is collected through the IoT sensors. After the smart farming management platform receives the farming data summarized by the farm terminal, it uses natural language processing technology to extract key parameters. The farming data includes but is not limited to feeding data, environmental data, drug data and health status data.

[0050] Furthermore, the feeding data of the farm is obtained, including feed data and drinking water data. The feed data includes but is not limited to the type of feed, quality data, number of feedings, feeding time, and weight of feed for each feeding. The quality data is determined by checking the proportion of trace elements in the feed weight in the ingredient list. The drinking water data includes but is not limited to the amount of drinking water, number of drinking water, and drinking water time. At the same time, the feeding data is divided into normal feeding data and sick feeding data according to whether the animal is in the process of illness.

[0051] Furthermore, environmental data of the farm is obtained, including temperature data, humidity data, light data, and harmful gas data. The temperature data includes but is not limited to the lowest temperature, the highest temperature, the average temperature, and the maximum temperature difference between day and night, and is recorded as daily data, monthly average data, and quarterly average data. The light data includes but is not limited to the light time and light intensity. The harmful gas data includes the name, concentration, and source of harmful gases that affect meat growth, and the harmful gases include but are not limited to ammonia, hydrogen sulfide, and carbon dioxide.

[0052] Furthermore, the drug data of the farm is obtained, including but not limited to the name of the vaccine and / or drug, drug dosage, frequency of use, and duration.

[0053] The farm terminal summarizes the breeding data in chronological order and transmits it to the platform, which is used to receive, store and process the breeding data of multiple different poultry farms. The platform extracts key parameters from the breeding data through natural language processing technologies such as keyword matching and TextRank, classifies the key parameters, and performs statistical analysis on multiple breeding data with similar key parameters to obtain a first evaluation report. The platform sends the first evaluation report to multiple breeding experts. The breeding experts formulate their own breeding plans based on the first evaluation report and send them back to the platform. The platform presets expert weights for the breeding experts, assigns the corresponding expert weights to the received breeding plans, and then integrates them, for example, multiplying the data in the breeding plan by the corresponding expert weight, and then summing all the product results, and finally outputs a breeding plan recommendation and transmits it to the farm terminal. The sum of the expert weights is 1.

[0054] By statistically analyzing the farm data and sending the first evaluation report to multiple aquaculture experts, the data specificity problem caused by different environments and different cost inputs in the farms is effectively solved, and the overall control of the data of multiple farms in the platform location is achieved, which is conducive to comprehensive analysis of the data differences and average levels of individual farmers from a global perspective, and provides different data entry angles for aquaculture experts to specify aquaculture plans. By integrating the aquaculture plans output by multiple aquaculture experts, the final aquaculture plan recommendation is obtained, which realizes the integration of professional knowledge and experience of different experts, reduces the decision-making risks caused by a single point of view or lack of experience, and is conducive to forming a more comprehensive aquaculture plan and improving the adaptability of the aquaculture plan.

[0055] The feed-to-weight ratio prediction model is established with the breeding data as input and the feed-to-weight ratio when the poultry reaches the standard weight for market release as output. The collected historical breeding data and the corresponding historical poultry market weight are divided into a training set and a test set according to the proportion. Since the breeding plan also includes the breeding data, it is added to the test set as verification data in this embodiment. The feed-to-weight ratio prediction model is built using a support vector regression model, and a polynomial kernel is selected as the kernel function. The selected parameters are the regularization constant C and the insensitive loss function e, and the wolf pack algorithm is used to adjust the values ​​of the parameters C and e. The breeding plan recommendation is input into the feed-to-weight ratio prediction model to predict the first feed-to-weight ratio when the farmed animal is marketed. In this way, the execution effect evaluation of the breeding plan recommendation can be realized before application, which is conducive to timely feedback and adjustment of the breeding plan recommendation. The feed-to-weight ratio is the ratio of the amount of feed ingested by the farmed animal per unit time to its weight gain.

[0056] After the farm terminal receives the breeding plan recommendation, the breeder executes the breeding process under the guidance of the breeding plan recommendation, and the breeding process is the process in which the breeder starts to execute the breeding plan recommendation until the poultry reaches the qualified weight for market release. At the same time, the new breeding data generated during the breeding process is collected through sensors, and the farm terminal classifies and summarizes the new breeding data. The platform performs statistical analysis on the new breeding data and outputs a second evaluation report.

[0057] The platform converts the text of the first evaluation report into a first vector and the text of the second evaluation report into a second vector through existing technologies such as one-hot encoding and bag-of-words model, and calculates the similarity between the first vector and the second vector using existing technologies such as cosine similarity.

[0058] By performing statistical analysis on the real-time data after the farm executes the breeding plan recommendation, and comparing the obtained second evaluation report with the first evaluation report corresponding to the historical data, the execution process of the breeding plan recommendation can be evaluated. By comparing the similarity between the historical breeding plan recommendation and the current breeding plan recommendation, the changes in the breeding data can be intuitively observed, which is conducive to directly evaluating the execution effect of the breeding plan recommendation and providing data support for further optimization of the breeding plan.

[0059] When the farmed animals are ready for market, the platform extracts the feed weight from the new farming data, and calculates a second feed-to-weight ratio based on the feed weight and the weight of the received poultry when it reaches the qualified weight for market; the platform outputs the similarity and the ratio of the second feed-to-weight ratio to the first feed-to-weight ratio to the first evaluation report, which is used to guide the generation of the next farming plan recommendation.

[0060] The similarity can be used to compare the degree of deviation of the breeding data before and after the execution of the breeding plan recommendation. Combined with the second feed-to-weight ratio when the poultry is actually marketed, it can be used to evaluate the breeding plan recommendation and the execution effect of the breeding plan recommendation by the breeder. The smaller the similarity, the greater the degree of deviation and the lower the degree of execution of the breeding plan recommendation by the breeder. The lower the second feed-to-weight ratio, the more poultry meat is produced per unit feed, indicating that the breeding effect of the breeder is better and the feed utilization rate is higher. If the similarity decreases, but the feed-to-weight ratio decreases, that is, the breeder's execution degree is low, but the breeding effect is good; or if the similarity increases, but the feed-to-weight ratio also increases, that is, the breeder's execution degree is high, but the breeding effect is poor; it means that the breeding plan recommendation generated this time is of low quality.

[0061] Furthermore, the key parameters are classified, including: the platform classifies and stores the key parameters by name, and statistically analyzes the breeding data of similar information to obtain an overall evaluation report, the content of which includes but is not limited to the mean value and standard deviation of the breeding data.

[0062] Furthermore, the classification of the key parameters includes: the platform extracts the geographic information and the environmental data of the farm to obtain the corresponding geographic feature vector and environmental feature vector. The geographic feature vector and the environmental feature vector are spliced ​​into a first vector, and the similarity of the first vectors of multiple farms is calculated. Multiple similarity ranges are preset, and the farms whose similarities meet the similarity ranges are classified into the same type of farms. The key parameters of the farms of the same category are classified and stored by name, and after statistical analysis of the key parameters of multiple farms of the same category is performed, multiple local evaluation reports are obtained.

[0063] This application not only analyzes all the data collected by the platform from a global perspective, but also realizes the analysis of the benchmark level of the aquaculture industry in the management area corresponding to the platform, and provides a basis for comparative analysis from other perspectives. Affected by geographical conditions, different farms may have small differences in environmental data such as temperature and humidity. Therefore, this application also classifies farms according to geographic information, and analyzes the data collected by the platform from the perspective of geographic information, so that the data differences of farms of the same type are reduced, and the attention to feeding management capabilities and disease prevention and control capabilities is improved. It is also conducive to the refined and differentiated analysis of different types of farms, so as to realize the adaptation of aquaculture methods to local conditions.

[0064] Furthermore, the key parameters of the same type are analyzed to output a first evaluation report, including:

[0065] The first evaluation report includes an overall evaluation report and / or a local evaluation report.

[0066] When the first evaluation report includes the overall evaluation report and the local evaluation report, the overall evaluation report and the local evaluation report are integrated. For example, the data of the overall evaluation report and the local evaluation report are simply combined and sent to the aquaculture experts for comparative analysis; or a first evaluation weight is set for the overall evaluation report, and a second evaluation weight is set for the local evaluation report, the sum of the first evaluation weight and the second evaluation weight is 1, and the weighted evaluation report data is added to obtain the first evaluation report, which is then sent to the aquaculture experts.

[0067] Furthermore, the platform presets expert weights for the aquaculture experts, including: the expert weights are determined according to the expert attributes. The data of expert attributes are extracted, the data represented quantitatively is not processed, and the evaluation scores are set for the data represented qualitatively. The quantitative data values ​​and the qualitative evaluation scores are calculated, such as summed and normalized, to obtain the expert weights. The expert attributes include, but are not limited to: the expert's years of work experience, the number of papers published, the number of papers related to aquaculture data, the number of field surveys of farms participated in, the professional title level, and the education level. The number of papers published related to aquaculture data is used to indicate the degree of attention paid by experts to the relevant aquaculture data. The more the number, the higher the degree of attention and the greater the professional voice. Therefore, a first coefficient is set for the number of papers published related to aquaculture data; a second coefficient is set for the number of papers published; and the first coefficient is greater than the second coefficient.

[0068] By setting expert weights for the plans designated by different aquaculture experts, the professional qualities and practical experience of aquaculture experts can be more comprehensively and objectively reflected. With the help of multiple expert attributes, expert weights are determined from different dimensions, which helps to ensure that experts who have in-depth research and practical application of aquaculture data can have greater voice in the process of aquaculture data analysis and aquaculture plan formulation, thereby improving the scientificity and practicality of aquaculture plan recommendations.

[0069] In the actual breeding industry, due to the large and diversified group of farmers, different farmers have significant differences in breeding professional level, input cost and experience accumulation, resulting in differences in the degree of execution of the breeding plan recommendation by farmers, making the feed-to-weight ratio of poultry at the time of market release quite different. In this embodiment, the factors affecting the degree of execution include the farmers' feeding management ability, environmental control ability and disease prevention and control ability. Farmers with rich breeding experience can more accurately grasp the growth cycle, nutritional needs, feeding time and behavioral habits of poultry, so as to more effectively implement the feeding strategies in the recommendation and improve breeding efficiency. On the contrary, inexperienced farmers may cause the poultry to be too large due to excessive feeding, or because the hardware facilities are difficult to meet the guidance requirements of the recommendation, resulting in too high feed-to-weight ratio, affecting the final benefit. Good environmental control can reduce the occurrence of diseases. However, the gap in investment, maintenance management and awareness of environmental control facilities among different farmers leads to uneven execution of environmental regulation suggestions in the recommendation, resulting in immunosuppression of poultry and susceptibility to pathogen infection. Therefore, disease prevention and control capabilities are the last guarantee to save the farms. Farmers with professional prevention and control knowledge and skills can respond quickly, effectively control the spread of the epidemic, and ensure the safety of farming. On the contrary, if farmers have a weak awareness of disease prevention and control and backward technical means, even with detailed prevention and control guidelines, it is difficult to effectively implement them, thereby increasing farming risks and leading to an increase in feed-to-weight ratio.

[0070] Therefore, further, the platform uses the feeding management vector, the environmental control vector and the disease prevention and control vector as the influencing factors of the breeding plan recommendation, and also outputs them to the feed-to-weight ratio prediction model to obtain the first feed-to-weight ratio. By evaluating the feeding management ability, environmental control ability and disease prevention and control ability of the farm in a result-oriented manner, the evaluation dimensions are unified, which is conducive to reducing the interference of complex factors in the intermediate process and reducing the difficulty of evaluating the farm under the influence of multiple factors.

[0071] The platform calculates the standard deviation of the quantitative indicators in a breeding cycle based on the specific values ​​of the quantitative indicators collected in the breeding data, such as quality data, feeding times, feed weight for each feeding, water intake, and drinking times, and combines the standard deviations of multiple quantitative indicators into the breeding management vector to represent the breeding management capabilities of the farm.

[0072] The platform calculates the standard deviation of the quantitative indicators collected in the environmental data, such as temperature, humidity, light intensity, and concentration of harmful gases, in a breeding cycle, and combines the standard deviations of multiple quantitative indicators into the environmental control vector to represent the environmental control capability of the farm.

[0073] The platform calculates the standard deviation of these quantitative indicators in a breeding cycle based on the specific values ​​of the quantitative indicators collected in the drug data, such as drug dosage, frequency of use, and duration, and combines the standard deviations of multiple quantitative indicators into a drug vector to represent the drug use of the farm, which can evaluate the farm's ability to actively intervene after the poultry becomes sick. However, the disease prevention and control ability is not only reflected by the farm's active intervention ability, but also the health status of the poultry after the farm uses the drug should be evaluated to reflect the farm's drug application ability.

[0074] Therefore, further, the weight data, eating behavior data, and exercise data of the poultry in the farm before and after the disease are obtained to indicate the changes in weight, changes in food and drinking water, and changes in exercise and mental state of the poultry before and after the disease.

[0075] The weight data can be monitored by using an automatic weighing system, such as the automatic group weight monitoring system for flat-raised meat poultry provided in the paper DEVELOPMENT OF AN AUTOMATIC WEIGHING PLATFORM FOR MONITORING BODY WEIGHT OF BROILER CHICKENS IN COMMERCIAL PRODUCTION. The weight data is converted into a weight vector. For example, the average weight of the poultry in the corresponding breeding cycle is obtained from the platform, and the weight ratio of the weight data of the currently monitored poultry and the average weight is calculated; or the standard deviation of the weight of the poultry before and after the disease is calculated to reflect the weight change of the poultry. The weight ratio or the standard deviation of the weight is used as a weight vector. In this way, the growth of the poultry under the influence of the disease can be reflected.

[0076] The dietary behavior data includes pecking amount, pecking duration, and number of pecking times per unit time. The pecking amount is obtained by subtracting the remaining amount of feed in the feed tray from the amount of feed fed. The number of pecking times is obtained by recording the number of times the bird's head falls into the feed tray during the feeding time. Similarly, by calculating the standard deviation, the standard deviation of the pecking amount, the standard deviation of the pecking duration, and the standard deviation of the number of pecking times per unit time are combined into a dietary behavior vector. The standard deviation can be used to reflect the degree of fluctuation of the data, which helps to evaluate the appetite and nutritional intake of birds.

[0077] The motion data uses motion sensors and image sensors to collect and process data based on computer vision monitoring technology, such as poultry health monitoring imaging technology, to record and capture motion data and image data during non-sleep time. The motion data includes motion trajectory, neck image data, and head image data.

[0078] The movement trajectory can be tracked by computer image tracking technology for each bird in the poultry group, or by recording the movement position of the bird at a certain time through a motion sensor set for each bird, and calculating the movement distance of the bird in a certain time.

[0079] The neck image data is processed by a pre-trained convolutional neural network, the texture features and color features of the feathers are extracted, the neck feather part is identified, the pixel area occupied by the neck feather part in the image is calculated, and the pixel area of ​​the neck feather and the image scale are calculated to obtain the area of ​​the neck feather on one side of each bird. The feather state of the bird is an important indicator for judging its health. The compactness, flatness and brightness of the feathers indicate that the bird is in good health; while the messy, broken and falling feathers may indicate that the bird has a disease problem, causing the bird to be listless and unwilling to move. Birds in good health will raise their heads, and the neck and body of the bird will be far away, so that the neck feathers of the bird will stretch and the neck feather area will increase; while birds in poor health will curl up their bodies, and the neck and body of the bird will be close, so that the neck feathers of the bird will gather and overlap, and the neck feather area will decrease. Therefore, by identifying the area of ​​the neck feathers of the bird, the movement state of the bird can be reflected. If the neck feather area of ​​the bird is less than the neck feather threshold of the bird in a healthy state, and remains unchanged in the time window, it indicates that the bird has a health problem.

[0080] The head image data is processed by a pre-trained convolutional neural network to extract the eyelid closure features of the eye and the texture features of the poultry flesh skirt, and the eyelid closure features are converted into eye closure vectors, for example, the RGB channels of the eyelid closure features are sequentially spliced ​​into a one-dimensional vector, and the modulus of the vector is used to represent the degree of eye closure; the texture features of the poultry flesh skirt are converted into a flesh skirt extension vector, for example, the grayscale values ​​of pixels are compared using a local binary pattern algorithm to obtain a one-dimensional or multi-dimensional vector, and the modulus of the vector is used to represent the degree of extension of the poultry flesh skirt. The eye closure vector and the flesh skirt extension vector are spliced ​​to obtain a facial feature vector, which can be used to quantitatively represent the facial mental state of the poultry. The mental state of a poultry in poor health is depressed, which is intuitively manifested as half-closed or fully closed eyes; at the same time, when curling up, the poultry head droops, causing the flesh skirt of the poultry chin to appear wrinkled or bent when it contacts the poultry neck. If the facial feature vector is continuously less than the facial feature vector threshold in a healthy state in the time window, it indicates that the poultry has health problems. The facial feature vector threshold is the facial feature vector when the bird's eyelids are fully opened and the bird's flesh skirt is fully extended.

[0081] The moving distance, the neck feather area and the facial feature vector are combined into a motion vector to reflect the activity state of the poultry. The weight vector, the eating behavior vector and the motion vector are combined to obtain a health state vector to reflect the health state of the poultry. The drug vector and the health state vector are combined to obtain a disease prevention and control vector to represent the disease prevention and control capability of the farm.

[0082] By analyzing the changes in weight of poultry before and after they become ill, changes in their eating and drinking water intake, and changes in their exercise and mental state during routine behaviors, we have achieved the goal of using the effects of medication on poultry to reflect the disease prevention and control capabilities of farmers. This is conducive to forming an objective evaluation of the disease prevention and control capabilities of farmers and improving the accuracy of feed-to-weight ratio predictions.

[0083] Although the breeding plan recommendations issued by the platform have been guided and suggested by experts, the implementation of the breeding plan recommendations and the final results still depend on the input of the farmers. Therefore, the feed-to-weight ratio predicted by the breeding plan recommendations should be subject to the farmers' feeding management capabilities, environmental control capabilities, and disease prevention and control capabilities. Through the analysis of historical data, it is helpful to form farmers' evaluation of their ability to execute breeding plan recommendations, improve the accuracy of the prediction results of the feed-to-weight ratio prediction model, and then provide targeted implementation guidance to farmers.

[0084] Embodiment 2: This embodiment provides a smart farming monitoring system based on big data, as well as a storage medium and device for storing and executing the method. Figure 2 As shown, the system includes the following modules:

[0085] The analysis module is used to receive the breeding data collected by the Internet of Things sensor from the farm terminal, extract and output key parameters using natural language processing technology; classify the key parameters, analyze the key parameters of the same type, and output a first evaluation report.

[0086] The application module is used to generate a breeding plan recommendation based on the first evaluation report and transmit it to the farm terminal; at the same time, the breeding plan recommendation is output to the feed-to-weight ratio prediction model, and the first feed-to-weight ratio is received by the feed-to-weight ratio prediction model based on the breeding plan recommendation.

[0087] The control module is used to receive the new breeding data and output a second evaluation report after the breeding farm starts to execute the breeding plan recommendation displayed on the breeding farm terminal; calculate and output the similarity between the first evaluation report and the second evaluation report.

[0088] An updating module is used to extract feed weight from the new breeding data, calculate a second feed-to-weight ratio based on the feed weight and the received slaughter weight of the breeding products; output the similarity and the ratio of the second feed-to-weight ratio to the first feed-to-weight ratio to the first evaluation report, so as to guide the generation of the next breeding plan recommendation.

[0089] A smart farming monitoring device based on big data, the device comprising: a sensor for collecting various data related to the farm environment and poultry; at least one processor capable of processing various data related to the farm environment and poultry collected by the sensor; at least one memory for storing various data related to the farm environment and poultry collected by the sensor and at least one program; when the at least one program is executed by the at least one processor, the at least one processor implements the smart farming monitoring method based on big data, the processor and the memory can be connected by a bus or other means for storing at least one program; the memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network and a combination thereof, when the at least one program is executed by the at least one processor, the at least one processor implements the smart farming monitoring method based on big data.

[0090] A computer storage medium stores a program executable by a processor, wherein the program executable by the processor is used to implement the smart breeding monitoring method based on big data when executed by the processor.

[0091] It will be appreciated by those skilled in the art that all or some of the steps and systems in the disclosed method above may be implemented as software, firmware, hardware and appropriate combinations thereof. Some physical components or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor or a microprocessor, or may be implemented as hardware, or may be implemented as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or a non-transitory medium) and a communication medium (or a temporary medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that may be used to store desired information and may be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0092] The above embodiments are only used to illustrate the present invention and are not intended to limit the technical solutions described in the present invention. Although the present invention has been described in detail with reference to the above embodiments, the present invention is not limited to the above specific implementation methods. Therefore, any modification or equivalent replacement of the present invention; and all technical solutions and improvements thereof that do not depart from the spirit and scope of the invention are included in the scope of the claims of the present invention.

Claims

1. A smart breeding monitoring method based on big data, characterized in that: include: The farming data of the farm is collected through the IoT sensors. After the smart farming management platform receives the farming data summarized by the farm terminal, it uses natural language processing technology to extract key parameters; The platform classifies the key parameters, analyzes the key parameters of the same type, and outputs a first evaluation report; The platform generates a breeding plan recommendation based on the first evaluation report and transmits it to the breeding farm terminal; At the same time, the platform outputs the breeding plan recommendation to the feed-to-weight ratio prediction model to obtain a first feed-to-weight ratio predicted by the feed-to-weight ratio prediction model according to the breeding plan recommendation; After the farm starts to implement the breeding plan recommendation displayed on the farm terminal, the platform receives the new breeding data and outputs a second evaluation report; at the same time, the platform calculates the similarity between the first evaluation report and the second evaluation report; When the farmed animals are released from the market, the platform extracts the feed weight from the new farming data, and calculates the second feed-to-weight ratio based on the feed weight and the received weight of the farmed animals; the platform outputs the similarity and the ratio of the second feed-to-weight ratio to the first feed-to-weight ratio to the first evaluation report for guiding the generation of the next farming plan recommendation.

2. According to the big data-based smart farming monitoring method of claim 1, it is characterized in that: The classifying of the key parameters includes: the platform classifies and stores the key parameters according to names, and obtains an overall evaluation report after analyzing the key parameters; and the first evaluation report includes the overall evaluation report.

3. According to the big data-based smart farming monitoring method of claim 1, it is characterized in that: The classification of the key parameters includes: the platform extracting environmental data from the breeding data; the platform classifying the breeding farms according to the geographical information of the breeding farms and the environmental data; classifying and storing the key parameters of the breeding farms of the same category by name, and obtaining multiple local evaluation reports after analyzing the key parameters of multiple breeding farms; the first evaluation report includes the local evaluation report.

4. The method for intelligent breeding monitoring based on big data according to claim 1, characterized in that: The first evaluation report includes an overall evaluation report and a local evaluation report. A first evaluation weight is set for the overall evaluation report, and a second evaluation weight is set for the local evaluation report. The first evaluation report is obtained by fusing the key parameters corresponding to the overall evaluation report and the local evaluation report.

5. The method for intelligent breeding monitoring based on big data according to claim 1, characterized in that: The step of generating a breeding plan recommendation according to the first evaluation report comprises: The platform presets expert weights; the platform sends the first evaluation report to multiple aquaculture experts. After the aquaculture experts formulate a breeding plan based on the first evaluation report, the platform assigns the corresponding expert weights to the received breeding plan, merges the multiple breeding plans with the expert weights set, and outputs the breeding plan recommendation.

6. The method for intelligent breeding monitoring based on big data according to claim 1, characterized in that: The breeding data includes feeding data, environmental data, drug data and health status data; when the platform outputs the breeding plan recommendation to the feed-to-weight ratio prediction model, it also includes: The platform extracts a feeding management vector of the farm according to the feeding data; extracts an environmental control vector of the farm according to the environmental data; and extracts a disease prevention and control vector of the farm according to the drug data and the health status data; The platform uses the feeding management vector, the environmental control vector and the disease prevention and control vector as influencing factors of the breeding plan recommendation, and also outputs them to the feed-to-weight ratio prediction model to obtain the first feed-to-weight ratio.

7. The intelligent breeding monitoring method based on big data according to claim 6 is characterized in that: The platform performs statistical analysis on the feeding data and the environmental data, and after combining the statistical analysis results of a plurality of data, obtains the corresponding feeding management vector and the environmental control vector.

8. The method for intelligent breeding monitoring based on big data according to claim 6, characterized in that: The extracting of the disease prevention and control vector of the farm according to the drug data and the health status data includes: After statistically analyzing the drug data, combining the analysis results of a plurality of the drug data into a drug vector; Collecting weight data, eating behavior data and exercise data, performing statistical analysis on the weight data and the eating behavior data to obtain corresponding weight vectors and eating behavior vectors; performing image analysis on the exercise data to obtain exercise vectors; The drug vector, the body weight vector, the dietary behavior vector and the exercise vector are combined into the disease prevention and control vector.

9. The method for intelligent breeding monitoring based on big data according to claim 8, characterized in that: The motion data includes motion trajectory, neck image data, and head image data; and the image analysis of the motion data includes: Calculating the movement trajectory according to the movement position and movement distance of the poultry; Extracting texture features and color features of neck feathers from the neck image data through a pre-trained neural network, and calculating the area of ​​the neck feathers according to the pixels occupied by the neck feathers; The eyelid closure features of the eyes and the texture features of the poultry skirt are extracted from the head image data through the neural network to obtain the eyelid closure vector and the skirt extension vector, and the modulus of the eyelid closure vector and the skirt extension vector is calculated.

10. A smart breeding monitoring system based on big data, characterized in that: Includes the following modules: The analysis module is used to receive the farming data collected by the IoT sensors from the farm terminal, and extract and output key parameters using natural language processing technology; Classifying the key parameters, analyzing the key parameters of the same type, and outputting a first evaluation report; an application module, configured to generate a breeding plan recommendation according to the first evaluation report, and transmit the result to the breeding farm terminal; and output the breeding plan recommendation to a feed-to-weight ratio prediction model, and receive a first feed-to-weight ratio predicted by the feed-to-weight ratio prediction model according to the breeding plan recommendation; A control module, after the breeding farm starts to implement the breeding plan recommendation displayed on the breeding farm terminal, is used to receive new breeding data and output a second evaluation report; Calculating and outputting the similarity between the first evaluation report and the second evaluation report; The update module is used to extract the feed weight from the new breeding data when the farmed animals are released, and calculate the second feed-to-weight ratio based on the feed weight and the received farmed animal release weight; the similarity and the ratio of the second feed-to-weight ratio to the first feed-to-weight ratio are output to the first evaluation report to guide the generation of the next breeding plan recommendation.

Images