WO2021104007A1 - 一种动物状态监测方法、装置、电子设备及存储介质 - Google Patents

一种动物状态监测方法、装置、电子设备及存储介质 Download PDF

Info

Publication number
WO2021104007A1
WO2021104007A1 PCT/CN2020/127766 CN2020127766W WO2021104007A1 WO 2021104007 A1 WO2021104007 A1 WO 2021104007A1 CN 2020127766 W CN2020127766 W CN 2020127766W WO 2021104007 A1 WO2021104007 A1 WO 2021104007A1
Authority
WO
WIPO (PCT)
Prior art keywords
animal
information
determining
state
key point
Prior art date
Application number
PCT/CN2020/127766
Other languages
English (en)
French (fr)
Inventor
苏睿
Original Assignee
京东数科海益信息科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 京东数科海益信息科技有限公司 filed Critical 京东数科海益信息科技有限公司
Publication of WO2021104007A1 publication Critical patent/WO2021104007A1/zh

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
    • A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
    • A61B5/1126—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb using a particular sensing technique
    • A61B5/1128—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb using a particular sensing technique using image analysis
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B2503/00—Evaluating a particular growth phase or type of persons or animals
    • A61B2503/40—Animals

Definitions

  • This application relates to the field of image processing, and in particular to an animal state monitoring method, device, electronic equipment and storage medium.
  • the feed conversion rate occupies a relatively high proportion of the breeding cost.
  • the key to effective production is to maintain the optimal growth rate and feed conversion rate through continuous monitoring.
  • the body condition of farmed animals is an important factor that affects the optimal growth rate and feed conversion rate.
  • Body condition measurement is an important part of production management. It can provide an important basis for evaluating animal nutrition, growth environment, and sanitary conditions.
  • Evaluating the body condition of an animal generally requires analyzing the entire skeleton of the animal, judging the body condition through the empirical knowledge summarized by some experts, and judging the health of the animal, such as judging whether the animal has hoofs by analyzing the animal's body posture Sickness etc. It is of great significance to the quality control and sustainable development of farms.
  • embodiments of the present application provide an animal condition monitoring method, device, electronic equipment, and storage medium.
  • an animal state monitoring method including:
  • the body pose information of the animal is recognized from the side image
  • the animal state corresponding to the animal is determined according to the posture information.
  • the acquiring at least two consecutively shot side images of the same animal includes:
  • the body posture information includes key point information of the animal in the side image
  • the key points include at least two of the following:
  • the determining the animal state corresponding to the animal according to the posture information includes:
  • the determining the animal state corresponding to the animal according to the posture information includes:
  • the state of the animal is determined according to the vector change information.
  • the method further includes:
  • the determining the animal state corresponding to the animal according to the posture information includes:
  • the state of the animal is determined according to the body condition parameter.
  • the method further includes:
  • the method further includes:
  • the sample image and the annotation information are input into a preset depth convolutional neural network for training, and the body pose detection model is obtained.
  • an animal condition monitoring device including:
  • the acquisition module is used to acquire at least two consecutively photographed side images of the same animal
  • a recognition module configured to recognize the animal's body pose information from the side image according to a pre-trained body pose detection model
  • the determining module is used to determine the animal state corresponding to the animal according to the posture information.
  • an embodiment of the present application provides an electronic device, including: a processor, a communication interface, a memory, and a communication bus, where the processor, the communication interface, and the memory complete mutual communication through the communication bus;
  • the memory is used to store computer programs
  • the processor is used to implement the steps of the above method when the computer program is executed.
  • an embodiment of the present application provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method steps are implemented.
  • the animal body posture information is detected from the image through the pre-trained model, and the animal state is analyzed according to the posture information. Without manual monitoring, the abnormality of the animal’s body condition can be found timely and accurately, which improves the monitoring efficiency and reduces the labor cost. Time costs. In addition, the health of animals is ensured and the efficiency of breeding is improved.
  • FIG. 1 is a flowchart of an animal condition monitoring method provided by an embodiment of the application
  • Figure 2 is a schematic diagram of the key points of the bovine body provided by an embodiment of the application.
  • FIG. 3 is a flowchart of an animal condition monitoring method according to another embodiment of the application.
  • FIG. 4 is a schematic diagram of the key points of the cow body obtained by identifying the body pose detection model according to the embodiment of the application;
  • FIG. 5 is a schematic diagram of key points of a cow body obtained by recognizing a body pose detection model according to another embodiment of the application;
  • FIG. 6 is a schematic diagram of key points of a cow body obtained by recognizing a body pose detection model according to another embodiment of the application.
  • FIG. 7 is a schematic diagram of key points of a cow body obtained by recognizing a body pose detection model according to another embodiment of the application.
  • FIG. 8 is a schematic diagram of key points of a cow body obtained by recognizing a body pose detection model according to another embodiment of the application.
  • FIG. 9 is a schematic diagram of key points of a cow body obtained by recognizing a body pose detection model according to another embodiment of the application.
  • FIG. 10 is a block diagram of an animal condition monitoring device provided by an embodiment of this application.
  • FIG. 11 is a schematic structural diagram of an electronic device provided by an embodiment of the application.
  • the embodiment of this application uses computer vision to identify body posture information from the side image of an animal based on a pre-trained neural network model, and analyzes the animal’s state according to the posture information, such as body condition score (BCS), Synthetic fattening index (Synthetic Fattening Index), fattening index FI or R index, growth and development, whether there is hoof disease, etc.
  • body condition score BCS
  • Synthetic fattening index Synthetic Fattening Index
  • FI or R index fattening index
  • growth and development whether there is hoof disease, etc.
  • Fig. 1 is a flowchart of an animal condition monitoring method provided by an embodiment of the application. As shown in Figure 1, the method includes the following steps:
  • Step S11 acquiring at least two consecutively shot side images of the same animal
  • Step S12 according to the pre-trained body pose detection model, the body pose information of the animal is recognized from the side image;
  • Step S13 Determine the animal state corresponding to the animal according to the posture information.
  • the animal body posture information is detected from the image through the pre-trained model, and the animal state is analyzed according to the posture information, without manual monitoring, and the abnormality of the animal body condition can be found in time and accurately, and the monitoring efficiency can be improved. , Reduce labor cost and time cost. In addition, the health of animals is ensured and the efficiency of breeding is improved.
  • step S11 includes: acquiring a video obtained by photographing the side of the animal; and extracting each frame of image in the video as a side image.
  • the body pose detection model in step S12 may be obtained by training based on a preset deep convolutional neural network, such as a DeepLabCut network.
  • a preset deep convolutional neural network such as a DeepLabCut network.
  • the body pose information of the animal can be identified from the side image.
  • the posture information includes animal key point information in the side image.
  • Figure 2 is a schematic diagram of the key points of the bovine body provided by an embodiment of the application. As shown in Figure 2, taking a cow as an example, the key points on the cow's body can include at least two of the following: right eye, left eye, mouth, spine head, right front leg root, left front leg root, right front knee, left front leg Knee, left front ankle, right front ankle, clavicle, hip joint, left hind knee, right hind knee, left rear ankle, and right rear ankle.
  • step S13 can be implemented in at least one of the following ways to determine the animal state according to the posture information.
  • the first way is to determine whether the animal is normal or not by changing the position of the key point.
  • Step S13 includes: determining the position change information of the key point according to the key point information corresponding to the at least two side images; and determining the animal state according to the position change information.
  • the second way is to determine whether the animal is normal or not through the change of the key point composition vector.
  • connection of any two key points can form a vector.
  • the change of the vector such as the change of the direction of the vector, the change of the angle between the vector and the vector, will also show a certain trend. If this trend is not met, it means that the animal is abnormal.
  • step S13 includes: obtaining a vector based on the connection of the key points; determining the vector change information of the vector according to the key point information corresponding to the at least two side images; and determining the animal state according to the vector change information.
  • connection of key points 6 and 8 results in vector 1
  • the connection of key points 8, 10 results in vector 2.
  • the angle change between vector 1 and vector 2 has a different trend from that of a normal cow. Therefore, the state of the cow can be determined through the vector change information.
  • the method further includes: determining the key points corresponding to the position change information and/or vector change information that do not meet the preset conditions; and determining the abnormal part according to the key points.
  • the preset condition may be the position change trend or the vector change trend of the key point in the normal state of the animal.
  • the abnormality in addition to determining the abnormal state of the animal, the abnormality can also be further determined. For example, if the key points that do not meet the preset conditions are 6, 8, 10, it can be determined that the abnormality is the right front hoof of the cow. In this way, when reminding the abnormal state of the animal, you can also add the abnormal part of the animal, so that the breeder can quickly determine the problem of the animal, and can treat the animal in time or adjust the breeding strategy, etc., to further improve the animal The monitoring efficiency can ensure the health of animals and improve the efficiency of breeding.
  • the above step S13 includes: determining the body condition parameter corresponding to the animal according to the key point information; and determining the animal state according to the body condition parameter.
  • the body condition parameter may be the animal's body condition score, synthetic fattening index, fattening index FI or R index, growth and development index, and so on. Based on the key point information, the animal's height, volume, weight and other indicators can be estimated. Based on these indicators, the animal's body condition parameters can be calculated separately or combined with other indicator data, and then the animal's physical health, growth and development, etc. can be determined through the body condition parameters. status.
  • the body condition parameter may be a score of the animal's body condition, for example, a 5-point system is adopted, with 1 point indicating excessive weight loss and 5 points indicating excessive obesity.
  • the body condition parameter is the growth and development stage of the animal.
  • a 4-point scale is adopted, with 1 point representing calves, 2 points representing mature cattle, 3 points representing young cattle, and 4 points representing adult cows.
  • body condition parameters can take many forms, so I won't repeat them here.
  • the method further includes: obtaining identification information corresponding to the animal; and performing a preset reminder operation according to the identification information.
  • the preset reminding operation may include: sending the identification and animal status of the animal to the preset terminal in the form of information, or controlling the electronic collar corresponding to the identification information to emit light of a specific color to remind the staff, and so on.
  • the abnormal part can also be added to the reminder message, so that the staff can learn about the specific abnormality of the animal.
  • the method further includes a training process of the body pose detection model.
  • FIG. 3 is a flowchart of an animal state monitoring method according to another embodiment of the application. As shown in Figure 3, the method also includes the following steps:
  • Step S21 obtaining a sample image of the side of the animal
  • Step S22 Determine the label information corresponding to the sample image, and the label information includes key point information of the animal;
  • Step S23 Input the sample image and the annotation information into the preset depth convolutional neural network for training to obtain a body pose detection model.
  • the sample image and the labeled key point information can be input into the DeepLabCut network for training, and when the loss converges, a trained body pose detection model is obtained.
  • DeepLabCut is a deep convolutional neural network that combines two important technologies in target detection and image semantic segmentation: the ResNet network pre-trained by ImageNet and the deconvolution layer, and the results of ResNet are upsampled to decode through deconvolution. Key point information. Using the DeepLabCut network, you can use fewer sample images (about 200) for training to complete the posture detection of animals.
  • a large number of sample images of the side of the cow can be collected, and the above-mentioned 16 key points of the cow can be marked in these images.
  • Figures 4 to 9 are schematic diagrams of key points of a cow body obtained by recognizing a body pose detection model according to an embodiment of the application. As shown in Figure 4-9, these 6 images are obtained by continuously shooting the same cow. Inputting these images into the pose detection model can accurately identify the 16 key points of the cow in each image. Subsequent analysis of the changes in the position of the key points or the changes in the vector composition of the key points can determine whether there is an abnormality in the cow's body.
  • the animal body pose information can be subsequently detected based on the model, and the animal state can be analyzed based on the body pose information.
  • an abnormal animal body condition can be detected in a timely and accurate manner. Monitoring efficiency, reducing labor costs and time costs. In addition, the health of animals is ensured and the efficiency of breeding is improved.
  • Fig. 10 is a block diagram of an animal condition monitoring device provided by an embodiment of the application.
  • the device can be implemented as part or all of an electronic device through software, hardware, or a combination of the two.
  • the animal condition monitoring device includes:
  • the obtaining module 31 is used to obtain at least two consecutively shot side images of the same animal;
  • the recognition module 32 is used for recognizing and obtaining the body pose information of the animal from the side image according to the pre-trained body pose detection model;
  • the determining module 33 is used to determine the animal state corresponding to the animal according to the posture information.
  • the electronic device may include: a processor 1501, a communication interface 1502, a memory 1503, and a communication bus 1504, wherein the processor 1501, the communication interface 1502, and the memory 1503 pass through The communication bus 1504 completes mutual communication.
  • the memory 1503 is used to store computer programs
  • the processor 1501 is configured to execute the computer program stored in the memory 1503 to implement the steps of the above method embodiment below.
  • the communication bus mentioned in the above electronic device may be a Peripheral Component Interconnect (P C I) bus or an Extended Industry Standard Architecture (EISA) bus.
  • P C I Peripheral Component Interconnect
  • EISA Extended Industry Standard Architecture
  • the communication bus can be divided into address bus, data bus, control bus and so on. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.
  • the communication interface is used for communication between the above-mentioned electronic device and other devices.
  • the memory may include random access memory (Random Access Memory, RAM), and may also include non-volatile memory (Non-Volatile Memory, NVM), such as at least one disk storage.
  • NVM non-Volatile Memory
  • the memory may also be at least one storage device located far away from the foregoing processor.
  • the above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (Network Processor, NP), etc.; it can also be a digital signal processor (Digital Signal Processing, DSP), a dedicated integrated Circuit (Application Specific Integrated Circuit, ASIC), Field-Programmable Gate Array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.
  • CPU central processing unit
  • NP Network Processor
  • DSP Digital Signal Processing
  • ASIC Application Specific Integrated Circuit
  • FPGA Field-Programmable Gate Array
  • FPGA Field-Programmable Gate Array
  • the present application also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the above-mentioned method embodiment below.

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Veterinary Medicine (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physiology (AREA)
  • Public Health (AREA)
  • Animal Behavior & Ethology (AREA)
  • Surgery (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Artificial Intelligence (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Medical Informatics (AREA)
  • Molecular Biology (AREA)
  • Evolutionary Computation (AREA)
  • Signal Processing (AREA)
  • Mathematical Physics (AREA)
  • Psychiatry (AREA)
  • Fuzzy Systems (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Dentistry (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Image Analysis (AREA)

Abstract

一种动物状态监测方法、装置、电子设备及存储介质，该方法包括：获取同一动物的至少两张连续拍摄的侧面图像（S11）；根据预先训练的体姿检测模型，从侧面图像中识别得到动物的体姿信息（S12）；根据体姿信息确定动物对应的动物状态（S13）。基于计算机视觉方式，通过预先训练的模型从图像中检测动物体姿信息，根据体姿信息分析动物状态，无需人工监测，可以及时、准确地发现动物体况异常，提高监测效率，降低人力成本和时间成本。

Description

一种动物状态监测方法、装置、电子设备及存储介质
本申请要求于2019年11月26日提交中国专利局、申请号为201911176736.3、发明名称为“一种动物状态监测方法、装置、电子设备及存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及图像处理领域，尤其涉及一种动物状态监测方法、装置、电子设备及存储介质。
背景技术
我国畜牧业发展不平衡，既有一定规模的现代化大型养殖场，又有生产力低下的小型养殖场，畜牧业总体水平同国外相比有很大差距。这主要体现在生产管理、环境条件等各个方面。
对于生产管理来说，饲料转化率在养殖成本中所占比例较高，有效生产的关键就是通过连续监测来维持最优生长率和饲料转化率。养殖动物体况则是影响最优生长率和饲料转化率的重要因素，体况测量是生产管理中的一项重要环节，它可以为评价动物的营养，生长环境，卫生条件提供重要的依据。
对评估动物体况一般需要对动物的整个骨架进行分析，通过一些专家总结的经验知识对体况加以判断，以及对动物的健康情况进行判断，比如通过分析动物的身体姿态来判断动物是否有蹄病等。对于养殖场的质量控制及可持续发展有着极其重要的意义。
但是，通过专家人工观察的方式评估动物体况，效率较低，无法及时发现动物体况异常，且耗费大量的人力成本和时间成本。
发明内容
为了解决上述技术问题或者至少部分地解决上述技术问题，本申请实施例提供了一种动物状态监测方法、装置、电子设备及存储介质。
第一方面，本申请实施例提供了一种动物状态监测方法，包括：
获取同一动物的至少两张连续拍摄的侧面图像；
根据预先训练的体姿检测模型，从所述侧面图像中识别得到所述动物的体姿信息；
根据所述体姿信息确定所述动物对应的动物状态。
可选的，所述获取同一动物的至少两张连续拍摄的侧面图像，包括：
获取对所述动物侧面拍摄得到的视频；
提取所述视频中的每一帧图像作为所述侧面图像。
可选的，所述体姿信息包括在所述侧面图像中所述动物的关键点信息；
所述关键点包括以下至少两个：
右眼、左眼、嘴、脊柱头、右前腿根部、左前腿根部、右前腿膝部、左前腿膝部、左前脚踝、右前脚踝、锁骨、髋关节、左后腿膝部、右后腿膝部、左后脚踝和右后脚踝。
可选的，所述根据所述体姿信息确定所述动物对应的动物状态，包括：
根据至少两张所述侧面图像对应的所述关键点信息确定所述关键点的位置变化信息；
根据所述位置变化信息确定所述动物状态；
和/或，
所述根据所述体姿信息确定所述动物对应的动物状态，包括：
基于所述关键点的连线得到向量；
根据至少两张所述侧面图像对应的所述关键点信息确定所述向量的向量变化信息；
根据所述向量变化信息确定所述动物状态。
可选的，当所述动物状态为异常时，所述方法还包括：
确定不符合预设条件的所述位置变化信息和/或向量变化信息对应的关键点；
根据所述关键点确定异常部位。
可选的，所述根据所述体姿信息确定所述动物对应的动物状态，包括：
根据所述关键点信息确定所述动物对应的体况参数；
根据所述体况参数确定所述动物状态。
可选的，当所述动物状态为异常时，所述方法还包括：
获取所述动物对应的标识信息；
根据所述标识信息执行预设提醒操作。
可选的，所述方法还包括：
获取动物侧面的样本图像；
确定所述样本图像对应标注信息，所述标注信息包括所述动物的关键点信息；
将所述样本图像及所述标注信息输入预设深度卷积神经网络进行训练，得到所述体姿检测模型。
第二方面，本申请实施例提供了一种动物状态监测装置，包括：
获取模块，用于获取同一动物的至少两张连续拍摄的侧面图像；
识别模块，用于根据预先训练的体姿检测模型，从所述侧面图像中识别得到所述动物的体姿信息；
确定模块，用于根据所述体姿信息确定所述动物对应的动物状态。
第三方面，本申请实施例提供了一种电子设备，包括：处理器、通信接口、存储器和通信总线，其中，处理器，通信接口，存储器通过通信总线完成相互间的通信；
所述存储器，用于存放计算机程序；
所述处理器，用于执行计算机程序时，实现上述方法步骤。
第四方面，本申请实施例提供了一种计算机可读存储介质，其上存储有计算机程序，该计算机程序被处理器执行时实现上述方法步骤。
本申请实施例提供的上述技术方案与现有技术相比具有如下优点：
基于计算机视觉方式，通过预先训练的模型从图像中检测动物体姿信息，根据体姿信息分析动物状态，无需人工监测，可以及时、准确地发现动物体况异常，提高监测效率，降低人力成本和时间成本。并且，保证动物健康，提高养殖效益。
附图说明
此处的附图被并入说明书中并构成本说明书的一部分，示出了符合本发明的实施例，并与说明书一起用于解释本发明的原理。
为了更清楚地说明本发明实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，对于本领域普通技术人员而言，在不付出创造性劳动性的前 提下，还可以根据这些附图获得其他的附图。
图1为本申请实施例提供的一种动物状态监测方法的流程图；
图2为本申请实施例提供的牛体关键点示意图；
图3为本申请另一实施例提供的一种动物状态监测方法的流程图；
图4为本申请实施例通过对体姿检测模型识别得到的牛体关键点的示意图；
图5为本申请另一实施例通过对体姿检测模型识别得到的牛体关键点的示意图；
图6为本申请另一实施例通过对体姿检测模型识别得到的牛体关键点的示意图；
图7为本申请另一实施例通过对体姿检测模型识别得到的牛体关键点的示意图；
图8为本申请另一实施例通过对体姿检测模型识别得到的牛体关键点的示意图；
图9为本申请另一实施例通过对体姿检测模型识别得到的牛体关键点的示意图；
图10为本申请实施例提供的一种动物状态监测装置的框图；
图11为本申请实施例提供的一种电子设备的结构示意图。
具体实施方式
为使本申请实施例的目的、技术方案和优点更加清楚，下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例是本申请的一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动的前提下所获得的所有其他实施例，都属于本申请 保护的范围。
本申请实施例通过计算机视觉的方式，基于预先训练的神经网络模型从动物的侧面图像中识别出体姿信息，根据体姿信息分析动物状态，如体况评分(body condition score，简称BCS)、合成育肥指数(Synthetic Fattening Index)、育肥指数FI或R指数、生长发育情况、是否存在蹄病等等。
下面首先对本发明实施例所提供的一种动物状态监测方法进行介绍。
图1为本申请实施例提供的一种动物状态监测方法的流程图。如图1所示，该方法包括以下步骤：
步骤S11，获取同一动物的至少两张连续拍摄的侧面图像；
步骤S12，根据预先训练的体姿检测模型，从侧面图像中识别得到动物的体姿信息；
步骤S13，根据体姿信息确定动物对应的动物状态。
本实施例中，基于计算机视觉方式，通过预先训练的模型从图像中检测动物体姿信息，根据体姿信息分析动物状态，无需人工监测，可以及时、准确地发现动物体况异常，提高监测效率，降低人力成本和时间成本。并且，保证动物健康，提高养殖效益。
可选的，步骤S11包括：获取对动物侧面拍摄得到的视频；提取视频中的每一帧图像作为侧面图像。
以牛为例，可以在牛场通道处布置拍摄装置，当牛经过时，对牛体侧面录制视频。如果通道较长，可以在入口和出口处分别布置拍摄装置，录制视频，分别从两段视频中分别截取拍摄角度符合要求的视频部分，且要保证时间连续性，之后再进行侧面图像的提取。
可选的，步骤S12中的体姿检测模型可以基于预设深度卷积神经网络训练得到，如DeepLabCut网络。
通过体姿检测模型，可以从侧面图像中识别到动物的体姿信息。可选的，体姿信息包括在侧面图像中动物关键点信息。图2为本申请实施例提供的牛体关键点示意图。如图2所示，以牛为例，牛体上的关键点可以包括以下至少两个：右眼、左眼、嘴、脊柱头、右前腿根部、左前腿根部、右前腿膝部、左前腿膝部、左前脚踝、右前脚踝、锁骨、髋关节、左后腿膝部、右后腿膝部、左后脚踝和右后脚踝。
可选的，步骤S13根据体姿信息确定动物状态可以通过以下至少一种方式实现。
方式一，通过关键点的位置变化确定动物是否正常。
由于正常动物运动过程中，其身体上的某个关键点位置变化会呈现一定的趋势，如果不符合该趋势，则说明动物出现异常。
步骤S13包括：根据至少两张侧面图像对应的关键点信息确定关键点的位置变化信息；根据位置变化信息确定动物状态。
方式二，通过关键点组成向量的变化情况确定动物是否正常。
任意两个关键点连线可以形成一个向量。在正常动物运动过程中，该向量的变化，如向量方向变化、向量和向量之间的角度变化等也会呈现一定的趋势，如果不符合该趋势，则说明动物出现异常。
因此，步骤S13包括：基于关键点的连线得到向量；根据至少两张侧面图像对应的关键点信息确定向量的向量变化信息；根据向量变化信息确定动物状态。
例如，如图2所示，关键点6、8连线得到向量1，关键点8、10连线得到向量2。如果牛蹄出现疾病，如坡脚，则该向量1和向量2之间的角度变化情况与正常牛呈现不同的趋势，因此，通过向量变化 信息可以确定牛的状态。
可选的，当动物状态为异常时，该方法还包括：确定不符合预设条件的位置变化信息和/或向量变化信息对应的关键点；根据关键点确定异常部位。
其中，预设条件可以为动物正常状态下关键点的位置变化趋势或向量变化趋势。
本实施例中，除了确定动物状态异常，还可以进一步确定出现异常的部位，例如，不符合预设条件的关键点为6、8、10，则可确定异常部位为牛的右前蹄。这样，在对动物的异常状态进行提醒时，还可以加入动物出现异常的部位，以便饲养人员可以快速确定动物所出现的问题，可以及时对动物进行治疗或调整饲养策略等等，进一步提高对动物的监测效率，保证动物健康，提高养殖效益。
可选的，上述步骤S13包括：根据关键点信息确定动物对应的体况参数；根据体况参数确定动物状态。
其中，体况参数可以为动物的体况评分、合成育肥指数、育肥指数FI或R指数、生长发育指数等等。基于关键点信息，可以估算出动物的高度、体积、重量等指标，基于这些指标可以单独计算或结合其他指标数据计算动物体况参数，进而通过体况参数可以确定动物的身体健康、生长发育等状态。
可选的，体况参数可以为对动物体况的评分，如采用5分制，1分表示过度消瘦，5分表示过度肥胖。或者，体况参数为动物的生长发育阶段，如采用4分制，1分表示犊牛，2分表示育成牛，3分表示青年牛，4分表示成年母牛。
根据实际需要，体况参数可以有多种形式，在此不再赘述。
可选的，当动物状态为异常时，方法还包括：获取动物对应的标 识信息；根据标识信息执行预设提醒操作。
该预设提醒操作可以包括：将动物的标识和动物状态以信息方式发送到预设终端，或者，可以控制该标识信息对应的电子项圈发出特定颜色的光以提示工作人员，等等。
可选的，还可以将异常部位加入到提醒信息中，以使得工作人员可以获知具体的动物异常情况。
在一个可选实施例中，该方法还包括体姿检测模型的训练过程。
图3为本申请另一实施例提供的一种动物状态监测方法的流程图。如图3所示，该方法还包括以下步骤：
步骤S21，获取动物侧面的样本图像；
步骤S22，确定样本图像对应标注信息，标注信息包括动物的关键点信息；
步骤S23，将样本图像及标注信息输入预设深度卷积神经网络进行训练，得到体姿检测模型。
本实施例中，可以将样本图像以及标注好的关键点信息输入DeepLabCut网络进行训练，当loss收敛后，得到训练好的体姿检测模型。
DeepLabCut是一个深度卷积神经网络，合并了目标检测和图像语义分割中的两个重要技术：通过ImageNet预训练的ResNet网络和反卷积层，通过反卷积对ResNet的结果进行上采样来解码关键点信息。采用DeepLabCut网络，可以使用较少的样本图像(约200张)训练，即可完成对于动物的体姿检测。
以牛为例，可以采集大量牛侧面样本图像，在这些图像中标注出牛的上述16个关键点。将样本图像及关键点信息输入DeepLabCut网 络进行训练。
图4-图9为本申请实施例通过对体姿检测模型识别得到的牛体关键点的示意图。如图4-图9所示，这6张图像为对同一只牛连续拍摄得到的，将这些图像输入体姿检测模型，可以准确识别到每张图像中牛体的16个关键点。后续通过对关键点位置变化情况或关键点组成向量的变化情况进行分析，可以确定牛体是否出现异常。
本实施例中，通过对体姿检测模型进行训练，使得后续可以基于该模型检测动物体姿信息，根据体姿信息分析动物状态，无需人工监测，可以及时、准确地发现动物体况异常，提高监测效率，降低人力成本和时间成本。并且，保证动物健康，提高养殖效益。
下述为本申请装置实施例，可以用于执行本申请方法实施例。
图10为本申请实施例提供的一种动物状态监测装置的框图，该装置可以通过软件、硬件或者两者的结合实现成为电子设备的部分或者全部。如图10所示，该动物状态监测装置包括：
获取模块31，用于获取同一动物的至少两张连续拍摄的侧面图像；
识别模块32，用于根据预先训练的体姿检测模型，从侧面图像中识别得到动物的体姿信息；
确定模块33，用于根据体姿信息确定动物对应的动物状态。
本申请实施例还提供一种电子设备，如图11所示，电子设备可以包括：处理器1501、通信接口1502、存储器1503和通信总线1504，其中，处理器1501，通信接口1502，存储器1503通过通信总线1504完成相互间的通信。
存储器1503，用于存放计算机程序；
处理器1501，用于执行存储器1503上所存放的计算机程序时，实 现以下上述方法实施例的步骤。
上述电子设备提到的通信总线可以是外设部件互连标准(Peripheral Component Interconnect，P C I)总线或扩展工业标准结构(Extended Industry Standard Architecture，EISA)总线等。该通信总线可以分为地址总线、数据总线、控制总线等。为便于表示，图中仅用一条粗线表示，但并不表示仅有一根总线或一种类型的总线。
通信接口用于上述电子设备与其他设备之间的通信。
存储器可以包括随机存取存储器(Random Access Memory，RAM)，也可以包括非易失性存储器(Non-Volatile Memory，NVM)，例如至少一个磁盘存储器。可选的，存储器还可以是至少一个位于远离前述处理器的存储装置。
上述的处理器可以是通用处理器，包括中央处理器(Central Processing Unit，CPU)、网络处理器(Network Processor，NP)等；还可以是数字信号处理器(Digital Signal Processing，DSP)、专用集成电路(Application Specific Integrated Circuit，ASIC)、现场可编程门阵列(Field-Programmable Gate Array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。
本申请还提供一种计算机可读存储介质，其上存储有计算机程序，该计算机程序被处理器执行时实现以下上述方法实施例的步骤。
需要说明的是，对于上述装置、电子设备及计算机可读存储介质实施例而言，由于其基本相似于方法实施例，所以描述的比较简单，相关之处参见方法实施例的部分说明即可。
进一步需要说明的是，在本文中，诸如“第一”和“第二”等之类的关系术语仅仅用来将一个实体或者操作与另一个实体或操作区分开来，而不一定要求或者暗示这些实体或操作之间存在任何这种实际 的关系或者顺序。而且，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者设备所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括所述要素的过程、方法、物品或者设备中还存在另外的相同要素。
以上所述仅是本发明的具体实施方式，使本领域技术人员能够理解或实现本发明。对这些实施例的多种修改对本领域的技术人员来说将是显而易见的，本文中所定义的一般原理可以在不脱离本发明的精神或范围的情况下，在其它实施例中实现。因此，本发明将不会被限制于本文所示的这些实施例，而是要符合与本文所申请的原理和新颖特点相一致的最宽的范围。

Claims (11)

  1. 一种动物状态监测方法，其特征在于，包括：
    获取同一动物的至少两张连续拍摄的侧面图像；
    根据预先训练的体姿检测模型，从所述侧面图像中识别得到所述动物的体姿信息；
    根据所述体姿信息确定所述动物对应的动物状态。
  2. 根据权利要求1所述的方法，其特征在于，所述获取同一动物的至少两张连续拍摄的侧面图像，包括：
    获取对所述动物侧面拍摄得到的视频；
    提取所述视频中的每一帧图像作为所述侧面图像。
  3. 根据权利要求1所述的方法，其特征在于，所述体姿信息包括在所述侧面图像中所述动物的关键点信息；
    所述关键点包括以下至少两个：
    右眼、左眼、嘴、脊柱头、右前腿根部、左前腿根部、右前腿膝部、左前腿膝部、左前脚踝、右前脚踝、锁骨、髋关节、左后腿膝部、右后腿膝部、左后脚踝和右后脚踝。
  4. 根据权利要求3所述的方法，其特征在于，所述根据所述体姿信息确定所述动物对应的动物状态，包括：
    根据至少两张所述侧面图像对应的所述关键点信息确定所述关键点的位置变化信息；
    根据所述位置变化信息确定所述动物状态；
    和/或，
    所述根据所述体姿信息确定所述动物对应的动物状态，包括：
    基于所述关键点的连线得到向量；
    根据至少两张所述侧面图像对应的所述关键点信息确定所述向量的向量变化信息；
    根据所述向量变化信息确定所述动物状态。
  5. 根据权利要求4所述的方法，其特征在于，当所述动物状态为异常时，所述方法还包括：
    确定不符合预设条件的所述位置变化信息和/或向量变化信息对应的关键点；
    根据所述关键点确定异常部位。
  6. 根据权利要求3所述的方法，其特征在于，所述根据所述体姿信息确定所述动物对应的动物状态，包括：
    根据所述关键点信息确定所述动物对应的体况参数；
    根据所述体况参数确定所述动物状态。
  7. 根据权利要求1所述的方法，其特征在于，当所述动物状态为异常时，所述方法还包括：
    获取所述动物对应的标识信息；
    根据所述标识信息执行预设提醒操作。
  8. 根据权利要求1所述的方法，其特征在于，所述方法还包括：
    获取动物侧面的样本图像；
    确定所述样本图像对应标注信息，所述标注信息包括所述动物的关键点信息；
    将所述样本图像及所述标注信息输入预设深度卷积神经网络进行 训练，得到所述体姿检测模型。
  9. 一种动物状态监测装置，其特征在于，包括：
    获取模块，用于获取同一动物的至少两张连续拍摄的侧面图像；
    识别模块，用于根据预先训练的体姿检测模型，从所述侧面图像中识别得到所述动物的体姿信息；
    确定模块，用于根据所述体姿信息确定所述动物对应的动物状态。
  10. 一种电子设备，其特征在于，包括：处理器、通信接口、存储器和通信总线，其中，处理器，通信接口，存储器通过通信总线完成相互间的通信；
    所述存储器，用于存放计算机程序；
    所述处理器，用于执行所述计算机程序时，实现权利要求1-8任一项所述的方法步骤。
  11. 一种计算机可读存储介质，其上存储有计算机程序，其特征在于，该计算机程序被处理器执行时实现权利要求1-8任一项所述的方法步骤。
PCT/CN2020/127766 2019-11-26 2020-11-10 一种动物状态监测方法、装置、电子设备及存储介质 WO2021104007A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201911176736.3 2019-11-26
CN201911176736.3A CN111297367A (zh) 2019-11-26 2019-11-26 一种动物状态监测方法、装置、电子设备及存储介质

Publications (1)

Publication Number Publication Date
WO2021104007A1 true WO2021104007A1 (zh) 2021-06-03

Family

ID=71161792

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/127766 WO2021104007A1 (zh) 2019-11-26 2020-11-10 一种动物状态监测方法、装置、电子设备及存储介质

Country Status (2)

Country Link
CN (1) CN111297367A (zh)
WO (1) WO2021104007A1 (zh)

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111297367A (zh) * 2019-11-26 2020-06-19 北京海益同展信息科技有限公司 一种动物状态监测方法、装置、电子设备及存储介质
CN112036364B (zh) * 2020-09-14 2024-04-16 京东科技信息技术有限公司 跛行识别方法及装置、电子设备和计算机可读存储介质
CN112215105A (zh) * 2020-09-28 2021-01-12 北京海益同展信息科技有限公司 目标动物的身体异常监测方法和装置、设备及存储介质
US12026968B2 (en) 2020-11-12 2024-07-02 Sony Group Corporation Training machine learning-based models for animal feature detection
CN113974828B (zh) * 2021-09-30 2024-02-09 西安交通大学第二附属医院 一种手术参考方案生成方法及装置
CN113963298A (zh) * 2021-10-25 2022-01-21 东北林业大学 基于计算机视觉的野生动物识别跟踪与行为检测系统、方法、设备及存储介质
CN117173733A (zh) * 2022-05-26 2023-12-05 阿果明股份有限公司 动物生长辨识方法及其系统

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130243308A1 (en) * 2012-03-17 2013-09-19 Sony Corporation Integrated interactive segmentation with spatial constraint for digital image analysis
CN109543607A (zh) * 2018-11-22 2019-03-29 广州小狗机器人技术有限公司 目标物异常状态检测方法、系统、监护系统及存储介质
CN109960986A (zh) * 2017-12-25 2019-07-02 北京市商汤科技开发有限公司 人脸姿态分析方法、装置、设备、存储介质以及程序
CN110188633A (zh) * 2019-05-14 2019-08-30 广州虎牙信息科技有限公司 人体体态指标预测方法、装置、电子设备及存储介质
CN110287923A (zh) * 2019-06-29 2019-09-27 腾讯科技（深圳）有限公司 人体姿态获取方法、装置、计算机设备及存储介质
CN110443150A (zh) * 2019-07-10 2019-11-12 思百达物联网科技(北京)有限公司 一种跌倒检测方法、装置、存储介质
CN110457999A (zh) * 2019-06-27 2019-11-15 广东工业大学 一种基于深度学习和svm的动物姿态行为估计与心情识别方法
CN111297367A (zh) * 2019-11-26 2020-06-19 北京海益同展信息科技有限公司 一种动物状态监测方法、装置、电子设备及存储介质

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190037800A1 (en) * 2017-08-03 2019-02-07 Vium Inc. Device and method of identification and classification of rodent cognition and emotion
CN107480651B (zh) * 2017-08-25 2019-12-17 清华大学深圳研究生院 异常步态检测方法及异常步态检测系统
CN108922622B (zh) * 2018-07-10 2023-10-31 平安科技（深圳）有限公司 一种动物健康监测方法、装置及计算机可读存储介质
CN109938737A (zh) * 2019-03-01 2019-06-28 苏州博慧智能科技有限公司 一种基于深度学习关键点检测的人体身型测量方法及装置

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130243308A1 (en) * 2012-03-17 2013-09-19 Sony Corporation Integrated interactive segmentation with spatial constraint for digital image analysis
CN109960986A (zh) * 2017-12-25 2019-07-02 北京市商汤科技开发有限公司 人脸姿态分析方法、装置、设备、存储介质以及程序
CN109543607A (zh) * 2018-11-22 2019-03-29 广州小狗机器人技术有限公司 目标物异常状态检测方法、系统、监护系统及存储介质
CN110188633A (zh) * 2019-05-14 2019-08-30 广州虎牙信息科技有限公司 人体体态指标预测方法、装置、电子设备及存储介质
CN110457999A (zh) * 2019-06-27 2019-11-15 广东工业大学 一种基于深度学习和svm的动物姿态行为估计与心情识别方法
CN110287923A (zh) * 2019-06-29 2019-09-27 腾讯科技（深圳）有限公司 人体姿态获取方法、装置、计算机设备及存储介质
CN110443150A (zh) * 2019-07-10 2019-11-12 思百达物联网科技(北京)有限公司 一种跌倒检测方法、装置、存储介质
CN111297367A (zh) * 2019-11-26 2020-06-19 北京海益同展信息科技有限公司 一种动物状态监测方法、装置、电子设备及存储介质

Also Published As

Publication number Publication date
CN111297367A (zh) 2020-06-19

Similar Documents

Publication Publication Date Title
WO2021104007A1 (zh) 一种动物状态监测方法、装置、电子设备及存储介质
Wurtz et al. Recording behaviour of indoor-housed farm animals automatically using machine vision technology: A systematic review
Chen et al. Behaviour recognition of pigs and cattle: Journey from computer vision to deep learning
Poppe et al. Exploration of variance, autocorrelation, and skewness of deviations from lactation curves as resilience indicators for breeding
Wang et al. ASAS-NANP SYMPOSIUM: Applications of machine learning for livestock body weight prediction from digital images
Chen et al. Recognition of feeding behaviour of pigs and determination of feeding time of each pig by a video-based deep learning method
Chen et al. A computer vision approach for recognition of the engagement of pigs with different enrichment objects
Zhang et al. Algorithm of sheep body dimension measurement and its applications based on image analysis
Ayadi et al. Dairy cow rumination detection: A deep learning approach
KR102141582B1 (ko) 영상분석을 활용한 어미돼지의 분만시점 예측 방법 및 장치
WO2023041904A1 (en) Systems and methods for the automated monitoring of animal physiological conditions and for the prediction of animal phenotypes and health outcomes
Zin et al. An automatic estimation of dairy cow body condition score using analytic geometric image features
Gu et al. A two-stage recognition method based on deep learning for sheep behavior
Gascoigne et al. Lamb growth rates and optimising production
Ojukwu et al. Development of a computer vision system to detect inactivity in group-housed pigs
Sun et al. Basic behavior recognition of yaks based on improved SlowFast network
Ronghua et al. Cow behavioral recognition using dynamic analysis
Xue et al. An analytical framework to predict slaughter traits from images in fish
Chang et al. Using deep learning to accurately detect sow vulva size in a group pen with a single camera
CN113052114A (zh) 死虾识别方法、终端设备及可读存储介质
Rahman et al. Prospect and scope of artificial neural network in livestock farming: a review
CN111523446A (zh) 图像识别方法及装置、存储介质
Yuan et al. Stress-free detection technologies for pig growth based on welfare farming: A review
Ding et al. Automatic recognition of diarrhea in weaned piglets based on machine vision.
CN110991235A (zh) 一种状态监测方法、装置、电子设备及存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20892851

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20892851

Country of ref document: EP

Kind code of ref document: A1