Livestock abnormal feeding monitoring method based on vision and weight

Abstract

The invention belongs to the technical field of smart farms, and particularly relates to a method which combines deep learning with a weighing sensor, solves the problem of inaccurate data caused by single use of video to monitor livestock, processes video monitoring data through a deep learning algorithm, and realizes accurate identification and tracking of livestock behaviors. Meanwhile, the food intake of the livestock is accurately measured in combination with data of a weighing sensor, and the accuracy and reliability of abnormal food intake behavior detection of the livestock are greatly improved through combined use of the weighing sensor and the livestock; the technical prejudice existing in a traditional livestock breeding monitoring method is overcome, that is, the feeding behavior of the livestock can be accurately monitored only through manual observation or installation of an expensive sensor, accurate monitoring and analysis of the feeding behavior of the livestock are achieved at low cost by combining the deep learning algorithm and the weighing sensor, and the accuracy of the livestock breeding monitoring method is improved. A novel and more efficient monitoring method is provided for the field of livestock breeding.

Description

Technical Field

[0001] This invention belongs to the field of smart farm technology, specifically relating to a method for monitoring abnormal feeding in livestock based on vision and weight. Background Technology

[0002] In existing livestock farming management, livestock feeding behavior is one of the important indicators for assessing their health status. Traditional monitoring methods mainly rely on manual observation and recording of livestock feeding behavior, or installing sensors on animals, or installing sensors in designated areas to record livestock weight. However, these methods have significant drawbacks: manual observation is inefficient and easily affected by human factors, leading to inaccurate data; installing sensors on animals is costly and difficult to apply on a large scale; while installing sensors in designated areas eliminates human factors, the detection accuracy is low and cannot meet the needs of large-scale farm farming. Summary of the Invention

[0003] The purpose of this invention is to provide a vision- and weight-based method for monitoring abnormal feeding in livestock, in order to solve the problems mentioned in the background art.

[0004] To achieve the above-mentioned technical objectives, the technical solution adopted by the present invention is as follows: A method for monitoring abnormal feeding in livestock based on vision and weight includes the following steps: S1: Video monitoring, which monitors the daily behavior and activities of livestock in real time through surveillance cameras and records the monitoring content to computer storage in real time; S2: Weight recording. A highly sensitive weight is installed under the feed trough, and the data is imported into the computer in real time through IoT hardware devices.

[0005] S3: Target tracking involves extracting frames from surveillance images, labeling and segmenting the data, importing these frames into the model, training, validating, and testing the model. The optimized model is then fed continuous video data, and the tracking trajectory is obtained by further adjustments to the detection results. S4: Abnormal feeding behavior analysis. The feeding trough weight is recorded in real time by a weighing device, and the feeding trough range is delineated to track the feeding status of livestock. Then, the feeding range is selected based on the farm's feeding range and historical experience. Finally, the feeding behavior is judged to be abnormal based on the feeding trough weight, feeding status and feeding range.

[0006] The specific operation of step S1 is as follows: S11: Installation of equipment, including surveillance cameras, recording devices, switches, routers, servers, and monitors; S12: Data acquisition, recording the video from the surveillance footage; S13: Data transmission, transferring the recorded data to the computer; S14: Data storage, storing the transmitted data on the server.

[0007] The specific operation of step S2 is as follows: S21: Equipment installation, placing the weighing device under the feed trough; S22: Data acquisition, recording the weight of the feeding trough; S23: Data transmission, transmitting the trough weight data to the server; S24: Data storage, storing the transmitted data in the memory.

[0008] The specific operation of step S3 is as follows: S31: Data processing, extracting frames from the surveillance video to form individual images, and then using these images for data annotation to mark the location of the feeding trough and livestock; S32: Data partitioning. The labeled data is partitioned into an 8:1:1 ratio for the training set, validation set, and test set, respectively. S33: Model building and training. The model is built based on deep learning for target detection. The dataset is fed into the model for training to obtain a livestock target detection model. S34: Model validation and analysis, put the test set data into the model for training, analyze whether the model meets the requirements, and optimize it according to the requirements; S35: Target tracking results verification and analysis.

[0009] The specific operation of step S4 is as follows: S41: Target trajectory statistics, delineate the feeding trough range, divide feeding time according to the farm's historical experience, record the livestock's feeding status, mark the state of entering the marked range as feeding, and mark the state of not entering the marked range as not feeding; S42: Abnormal trajectory analysis. By identifying livestock feeding behavior and combining it with the weight of the feeding trough, livestock that eat within this time range from the start of entering the feeding area are considered normal, while those that do not eat within this time range are considered abnormal. When the same number of livestock eat within the same time period, if the weight of the feeding trough exceeds the relative weight of the previous feeding trough, it indicates that the livestock's feeding is abnormal and may require closer attention or the dispatch of personnel for inspection.

[0010] A system for detecting abnormal feeding behavior in livestock based on a combination of vision and weight analysis includes a video monitoring module, a weighing module, a target tracking module, and an abnormal feeding analysis module.

[0011] The video monitoring module is used to collect videos of livestock's daily feeding and living behaviors; the weighing module is used to record the weight of the feeding trough; the target tracking module is used to track the daily behavior of livestock; and the abnormal feeding analysis module is used to analyze and determine whether the livestock's feeding behavior is abnormal.

[0012] A computer device includes a memory and a processor, the memory storing a computer program that, when executed by the processor, implements the steps of the above-described method.

[0013] A computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method.

[0014] An information data processing terminal is used to implement a livestock abnormal feeding behavior detection system based on a combination of vision and weight.

[0015] This invention provides a method and system for detecting abnormal feeding behavior in livestock based on a combination of vision and weight. It uses monitoring equipment as the foundation, deep learning as the core, and a weighing sensor as a supplement to achieve the detection of abnormal feeding behavior in livestock.

[0016] This invention solves the problem of inaccuracy in manual detection by using deep learning algorithms and visual monitoring to achieve precise tracking and analysis of livestock feeding behavior; it also solves the problem of not being able to observe livestock at all times, enabling 24 / 7 uninterrupted monitoring; and it also solves the problem of the high cost of installing sensors on livestock, avoiding expensive costs by installing a weighing device under the feeding trough to indirectly measure the amount of feed consumed by livestock.

[0017] Second, considering the technical solution as a whole or from a product perspective, the technical effects and advantages of the technical solution to be protected by this invention are specifically described as follows: This invention combines deep learning with weighing sensors to solve the problem of inaccurate data caused by using video surveillance of livestock alone. By processing video surveillance data through deep learning algorithms, it achieves accurate identification and tracking of livestock behavior. At the same time, combined with data from weighing sensors, it accurately measures the amount of food consumed by livestock. The combined use of the two greatly improves the accuracy and reliability of detecting abnormal feeding behavior in livestock.

[0018] This invention can also realize real-time monitoring and early warning of livestock feeding behavior. Through real-time analysis of monitoring videos and weighing data, the system can promptly detect abnormal feeding behavior of livestock and send early warning information to farmers, helping them to take timely measures to prevent the deterioration of livestock health problems.

[0019] The system described in this invention can significantly improve the efficiency and effectiveness of livestock farming. By monitoring and issuing early warnings of abnormal feeding behavior in livestock in real time, the system can help farmers to promptly identify and address livestock health problems, reducing economic losses caused by these issues. At the same time, the system's automation and intelligence features also reduce labor costs and improve the level of intelligence in livestock management, thus possessing broad market prospects and commercial value.

[0020] Compared with traditional livestock breeding monitoring methods, the system described in this invention can provide more accurate and comprehensive data on livestock feeding behavior, providing farmers with a more scientific basis for decision-making, thereby further improving breeding efficiency.

[0021] This invention is the first in the field of detecting abnormal feeding behavior in livestock to combine deep learning algorithms with weighing sensors, enabling precise tracking and analysis of livestock feeding behavior and filling a technological gap in this field both domestically and internationally.

[0022] This invention overcomes the technical bias in traditional livestock farming monitoring methods, which assume that only manual observation or the installation of expensive sensors can accurately monitor livestock feeding behavior. By combining deep learning algorithms and weighing sensors, this invention achieves accurate monitoring and analysis of livestock feeding behavior at a lower cost, providing a new and more efficient monitoring method for the livestock farming field. Attached Figure Description

[0023] The present invention can be further illustrated by the non-limiting embodiments given in the accompanying drawings.

[0024] Figure 1 This is a flowchart of a method for monitoring abnormal feeding in livestock based on vision and weight according to the present invention; Figure 2 This is a schematic diagram illustrating the content of video monitoring, a specific implementation of the present invention. Detailed Implementation

[0025] To enable those skilled in the art to better understand the present invention, the technical solution of the present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0026] like Figure 1-2 As shown, the present invention provides a method for monitoring abnormal feeding in livestock based on vision and weight, comprising the following steps: S1: Video monitoring, which monitors the daily behavior and activities of livestock in real time through surveillance cameras and records the monitoring content to computer storage in real time; S2: Weight recording. A highly sensitive weight is installed under the feed trough, and the data is imported into the computer in real time through IoT hardware devices.

[0027] S3: Target tracking involves extracting frames from surveillance images, labeling and segmenting the data, importing these frames into the model, training, validating, and testing the model. The optimized model is then fed continuous video data, and the tracking trajectory is obtained by further adjustments to the detection results. S4: Abnormal feeding behavior analysis. The feeding trough weight is recorded in real time by a weighing device, and the feeding trough range is delineated to track the feeding status of livestock. Then, the feeding range is selected based on the farm's feeding range and historical experience. Finally, the feeding behavior is judged to be abnormal based on the feeding trough weight, feeding status and feeding range.

[0028] The specific operation of step S1 is as follows: S11: Installation of equipment, including surveillance cameras, recording devices, switches, routers, servers, and monitors; S12: Data acquisition, recording the video from the surveillance footage; S13: Data transmission, transferring the recorded data to the computer; S14: Data storage, storing the transmitted data on the server.

[0029] The specific operation of step S2 is as follows: S21: Equipment installation, placing the weighing device under the feed trough; S22: Data acquisition, recording the weight of the feeding trough; S23: Data transmission, transmitting the trough weight data to the server; S24: Data storage, storing the transmitted data in the memory.

[0030] The specific operation of step S3 is as follows: S31: Data processing, extracting frames from the surveillance video to form individual images, and then using these images for data annotation to mark the location of the feeding trough and livestock; S32: Data partitioning. The labeled data is partitioned into an 8:1:1 ratio for the training set, validation set, and test set, respectively. S33: Model building and training. The model is built based on deep learning for target detection. The dataset is fed into the model for training to obtain a livestock target detection model. S34: Model validation and analysis, put the test set data into the model for training, analyze whether the model meets the requirements, and optimize it according to the requirements; S35: Target tracking results verification and analysis.

[0031] The specific operation of step S4 is as follows: S41: Target trajectory statistics, delineate the feeding trough range, divide feeding time according to the farm's historical experience, record the livestock's feeding status, mark the state of entering the marked range as feeding, and mark the state of not entering the marked range as not feeding; S42: Abnormal trajectory analysis. By identifying livestock feeding behavior and combining it with the weight of the feeding trough, livestock that eat within this time range from the start of entering the feeding area are considered normal, while those that do not eat within this time range are considered abnormal. When the same number of livestock eat within the same time period, if the weight of the feeding trough exceeds the relative weight of the previous feeding trough, it indicates that the livestock's feeding is abnormal and may require closer attention or the dispatch of personnel for inspection.

[0032] A system for detecting abnormal feeding behavior in livestock based on a combination of vision and weight, the system comprising: The video monitoring module is used to collect images of livestock's daily feeding and living behaviors, providing image data for model building. The weighing module is used to record the weight of the feed trough, in preparation for subsequent analysis of abnormal feeding in livestock; The target tracking module is used to track the daily behavior of livestock and prepare for the analysis of abnormal feeding behavior in livestock; The abnormal feeding analysis module is used to analyze abnormal feeding behavior in livestock. Within a specified time range, a specified feeding range, and a relative weight range of the feed trough, it analyzes and judges abnormal feeding behavior in livestock based on the livestock's condition, feeding interval, feeding range, and feed trough weight.

[0033] Another object of the present invention is to provide a computer device, the computer device including a memory and a processor, the memory storing a computer program, which, when executed by the processor, enables the processor to perform the steps of the method for detecting abnormal feeding behavior in livestock based on a combination of vision and weight.

[0034] Another object of the present invention is to provide a computer-readable storage medium storing a computer program that, when executed by a processor, enables the processor to perform the steps of the method for detecting abnormal feeding behavior in livestock based on a combination of vision and weight.

[0035] Another objective of this invention is to provide an information data processing terminal for implementing the aforementioned livestock abnormal feeding behavior detection system based on a combination of vision and weight.

[0036] Once the system is started and running, it can monitor the feeding behavior of livestock in real time and automatically identify and warn of abnormal feeding behavior of livestock by analyzing video trajectory and feed trough weight data. Operating steps: 1. System deployment and initialization; Install surveillance cameras on the farm to ensure coverage of the livestock's feeding area, and connect the cameras to network devices such as recording equipment, switches, and routers to achieve real-time transmission of video data.

[0037] A highly sensitive weighing device is installed under the feeding trough, and the weighing device is connected to the server via IoT hardware to ensure that real-time data on the weight of the feeding trough can be transmitted to the server.

[0038] Deploy the video surveillance module, weighing module, target tracking module, and abnormal eating analysis module to the server and perform initial configuration.

[0039] 2. Data Acquisition and Processing: Surveillance cameras collect real-time videos of livestock's daily behavior and activities, and transmit the video data to a server for storage.

[0040] The weighing device records the weight data of the feeding trough in real time and transmits the data to the server for storage.

[0041] The target tracking module extracts frames from the surveillance video to form image data, and then performs data annotation and segmentation for model training.

[0042] 3. Model training and anomaly detection: The deep learning model is trained using a labeled dataset to obtain a livestock target detection model.

[0043] The trained model is used for target tracking, tracking livestock in surveillance videos to obtain their feeding status and trajectory.

[0044] The abnormal feeding analysis module determines whether the livestock's feeding behavior is abnormal based on the weight data of the feed trough, the feeding status and trajectory of the livestock, as well as the farm's feeding area and historical experience, and sends an early warning message when an abnormality is detected.

[0045] The embodiments of the present invention have achieved some positive results during the research and development or use process, and indeed have great advantages compared with the prior art; In the experiment, we compared the livestock abnormal feeding behavior detection method based on the combination of vision and weight provided by this invention with traditional monitoring methods (manual observation and sensors installed on animals).

[0046] Cost-benefit analysis: Manual observation methods require a large amount of human resources and are inefficient, resulting in high costs.

[0047] Methods of installing sensors on animals: Although the detection accuracy is high, the sensor cost and maintenance cost are both high.

[0048] The method of this invention: By installing a weighing device under the feeding trough, the amount of feed consumed by livestock is indirectly measured, avoiding the cost of expensive sensors. At the same time, the automation and intelligence features of the system also reduce labor costs and improve the level of intelligence in breeding management.

[0049] This invention provides a wealth of data visualization tools, enabling livestock farmers to intuitively understand livestock feeding behavior and health status. Through charts and reports, farmers can clearly see livestock feeding trajectories, feed intake, and the occurrence of abnormal feeding behaviors, providing a scientific basis for livestock management decisions.

[0050] Practical application of this application in beef cattle farms 1. System Deployment: High-definition surveillance cameras and weighing devices are installed in every feeding area of ​​the beef cattle farm to ensure comprehensive coverage.

[0051] Configure servers and network equipment to establish a real-time transmission channel for video and weight data.

[0052] Deploy the video surveillance module, weighing module, target tracking module, and abnormal eating analysis module to the server and perform initial configuration.

[0053] 2. Data Acquisition and Processing: Surveillance cameras collect real-time videos of the daily activities of beef cattle, while weighing devices record the weight data of the feed troughs in real time.

[0054] Video and weight data are transmitted to the server in real time via the network and stored in the video database and weight database respectively.

[0055] The target tracking module performs frame extraction on the video data to form an image sequence, and then performs data annotation and segmentation.

[0056] 3. Model training and anomaly detection: The deep learning model was trained using a labeled image dataset to obtain a beef cattle target detection model.

[0057] The trained model is applied to real-time video data to achieve real-time tracking of the feeding status of beef cattle.

[0058] The abnormal feeding analysis module uses trough weight data and the cattle's feeding status, combined with the farm's feeding plan and historical experience, to determine whether the cattle's feeding behavior is abnormal, and sends early warning information to the farmers via SMS or APP push when an abnormality is detected.

[0059] This application includes a comparative experiment conducted at a black goat farm; 1. Experimental Design: Two similar feeding areas were selected within the black goat farm as the experimental group and the control group.

[0060] The experimental group used the detection method based on the combination of vision and weight provided by this invention, while the control group used the traditional manual observation method.

[0061] 2. Experimental procedure: The experimental group followed the steps in Example 1 to deploy the system, collect and process data, train the model, and detect anomalies.

[0062] The control group was assigned a dedicated person to observe and record the feeding behavior of the black goats at regular intervals.

[0063] 3. Experimental Results: The experimental group successfully detected multiple instances of abnormal feeding behavior in black goats (such as refusal to eat and overeating) and promptly sent early warning information to the breeders, effectively preventing the deterioration of the black goats' health problems.

[0064] In the control group, inaccurate data recording due to human factors prevented the timely detection of all abnormal feeding behaviors, leading to a worsening of health problems in some black goats.

[0065] Cost-benefit analysis shows that although the initial investment in the experimental group was higher (mainly for equipment purchase and system deployment), in the long run it significantly reduced labor costs and economic losses caused by the health problems of black goats, and had a higher cost-effectiveness.

[0066] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. A method of monitoring abnormal feeding of livestock based on vision and weight, characterized by, Comprise the following steps: S1: video monitoring, through the monitoring camera real-time monitoring of livestock daily behavior activities, the monitoring content real-time recording to computer storage; S2: weight recorder record, through the installation of high sensitivity under the trough weight recorder, through the Internet of Things hardware equipment data real-time import to the computer; S3: target tracking, through the monitoring data image data frame, data annotation, data division import model, the model is trained, verified, tested, to the debugged model into the coherent video data, in the detection result debugging gets track S4: abnormal eating behavior analysis, through the weight recorder real-time record of trough weight, and enclose the trough range, the livestock eating state tracking, then according to the farm feeding interval and historical experience selection eating interval, finally according to the trough weight, eating state and eating interval to judge whether the eating behavior is abnormal.

2. A method of monitoring abnormal feeding of livestock based on vision and weight according to claim 1, characterized in that, The specific operation of step S1 is as follows: S11: installation of equipment, including monitoring camera, video equipment, switch, router, server and display; S12: data acquisition, record the video in monitoring; S13: data transmission, the recorded data to the computer; S14: data storage, the data transmission is stored in the server.

3. A visual and weight based livestock abnormal feeding monitoring method according to claim 1, wherein, The specific operation of step S2 is as follows: S21: equipment installation, the weight recorder is placed under the trough; S22: data acquisition, record the weight of the trough; S23: data transmission, the data of the trough weight is transmitted to the server; S24: data storage, the data transmission is stored in the storage.

4. A visual and weight based livestock abnormal feeding monitoring method according to claim 1, wherein, The specific operation of step S3 is as follows: S31: data processing, frame the monitoring video, form a picture, and then use the picture for data annotation, mark the position of the trough and livestock; S32: data division, divide the annotated data into 8:1:1, which are training set, validation set and test set respectively; S33: model building and training, the model is built based on deep learning target detection, the data set is transmitted to the model for training, and the livestock target detection model is obtained; S34: model verification and analysis, put the test set data into the model for training, analyze whether the model meets the requirements, and optimize according to the requirements; S35: target tracking result verification and analysis.

5. A visual and weight based livestock abnormal feeding monitoring method according to claim 1, wherein, The specific operation of step S4 is as follows: S41: target trajectory statistics, enclose the trough range, divide the eating time according to the historical experience of the farm, record the eating state of the livestock, enter the marked range state, mark as eating, and mark as not eating if not entering the marked range; S42: abnormal trajectory analysis, through identifying the livestock eating behavior combined with the weight of the trough, mark as normal from the beginning of entering the eating interval in this time range, mark as abnormal if not eating in this time, when eating in the same time period of the same number of livestock, if the weight of the trough exceeds the relative weight of the past trough, it means that the livestock eating is abnormal, which may need to be paid more attention or arrange personnel to check.

6. A livestock abnormal feeding behavior detection system based on a combination of vision and weight, characterized by, It comprises a video monitoring module, a weight module, a target tracking module and an abnormal eating analysis module.

7. The system of claim 6, wherein, The video monitoring module is used for collecting daily feeding and living behavior videos of the livestock; the weight counting module is used for recording the weight of the feeding trough; the target tracking module is used for tracking the daily behavior of the livestock; and the abnormal feeding analysis module is used for analyzing and judging whether the feeding behavior of the livestock is abnormal. 8.A computer device, comprising a memory and a processor, and characterized in that, The memory stores a computer program, and the computer program is executed by the processor to realize the steps of the method in any one of claims 1 to 5.

9. A computer readable storage medium storing a computer program, characterized in that, The computer program is executed by the processor to realize the steps of the method in any one of claims 1 to 5.

10. An information data processing terminal, characterized by A livestock abnormal feeding behavior detection system based on vision and weight combination is used to realize the method in claim 6 or 7.

Images