WO2023130804A1 - 一种肉鸭生理生长信息巡检方法及系统 - Google Patents

一种肉鸭生理生长信息巡检方法及系统 Download PDF

Info

Publication number
WO2023130804A1
WO2023130804A1 PCT/CN2022/126503 CN2022126503W WO2023130804A1 WO 2023130804 A1 WO2023130804 A1 WO 2023130804A1 CN 2022126503 W CN2022126503 W CN 2022126503W WO 2023130804 A1 WO2023130804 A1 WO 2023130804A1
Authority
WO
WIPO (PCT)
Prior art keywords
meat duck
meat
information
duck
model
Prior art date
Application number
PCT/CN2022/126503
Other languages
English (en)
French (fr)
Inventor
肖德琴
招胜秋
刘又夫
黄一桂
殷建军
刘俊彬
卞智逸
Original Assignee
华南农业大学
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 华南农业大学 filed Critical 华南农业大学
Publication of WO2023130804A1 publication Critical patent/WO2023130804A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/02—Agriculture; Fishing; Forestry; Mining
    • A—HUMAN NECESSITIES
    • A01—AGRICULTURE; FORESTRY; ANIMAL HUSBANDRY; HUNTING; TRAPPING; FISHING
    • A01D—HARVESTING; MOWING
    • A01D45/00—Harvesting of standing crops
    • A—HUMAN NECESSITIES
    • A01—AGRICULTURE; FORESTRY; ANIMAL HUSBANDRY; HUNTING; TRAPPING; FISHING
    • A01K—ANIMAL HUSBANDRY; AVICULTURE; APICULTURE; PISCICULTURE; FISHING; REARING OR BREEDING ANIMALS, NOT OTHERWISE PROVIDED FOR; NEW BREEDS OF ANIMALS
    • A01K67/00—Rearing or breeding animals, not otherwise provided for; New or modified breeds of animals
    • A01K67/02—Breeding vertebrates
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01G—WEIGHING
    • G01G17/00—Apparatus for or methods of weighing material of special form or property
    • G01G17/08—Apparatus for or methods of weighing material of special form or property for weighing livestock
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01J—MEASUREMENT OF INTENSITY, VELOCITY, SPECTRAL CONTENT, POLARISATION, PHASE OR PULSE CHARACTERISTICS OF INFRARED, VISIBLE OR ULTRAVIOLET LIGHT; COLORIMETRY; RADIATION PYROMETRY
    • G01J5/00—Radiation pyrometry, e.g. infrared or optical thermometry
    • G01J5/0022—Radiation pyrometry, e.g. infrared or optical thermometry for sensing the radiation of moving bodies
    • G01J5/0025—Living bodies
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10016—Video; Image sequence
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10024—Color image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]

Definitions

  • the invention relates to the technical field of meat duck breeding, in particular to a method and system for inspecting physiological growth information of meat ducks.
  • the traditional method is that the staff regularly inspect the duck houses one by one to judge, but manual inspections have disadvantages such as poor real-time performance, labor-intensive, and irritating odors in the duck houses are harmful to human health.
  • ducks are sensitive poultry, and it is easy for staff to cause stress reactions in meat ducks during inspections, which does not meet the needs of animal welfare. Therefore, at present, there has been a technology of monitoring meat ducks by collecting information through machines, but the current monitoring method can only monitor the breeding of meat ducks according to the environmental conditions of the duck house, and cannot monitor the actual growth and physiological conditions of meat ducks.
  • Chinese patent application CN201610246433.4 discloses an online intelligent monitoring system for white muscovy duck houses, including an environmental indicator monitoring module, an environmental activity video monitoring module, a No. 1 wireless transmission module, a central processing unit, and a No. 2 wireless transmission module. module, intelligent management module and intelligent control module.
  • the intelligent management module includes servers, desktop computers, notebooks and mobile APPs.
  • the intelligent control module includes dehumidifiers, fans, heating equipment, lighting adjustment devices, water pump control devices and sound and light alarms.
  • the indicator monitoring module and the environmental activity video monitoring module are connected to the central processor through the No. 1 wireless transmission module, and the central processor is connected to the server and the intelligent control module through the No. 2 wireless transmission module, and the server is connected to the desktop computer, notebook and mobile APP.
  • This patent can only monitor the environmental condition of the duck tongue, but cannot monitor the actual growth physiological condition of the meat duck.
  • the purpose of the present invention is to provide a meat duck physiological growth information inspection method and system that can be comprehensively monitored and have high accuracy in health evaluation.
  • the invention provides a method for inspection of physiological growth information of meat ducks, comprising the following steps:
  • the physiological information of meat ducks includes body temperature and weight; among them, the body temperature information of meat ducks is collected and processed as follows:
  • the infrared thermal imaging image of the meat duck is input into the temperature measurement model of the meat duck.
  • the temperature measurement model of the meat duck is a convolutional neural network model.
  • the meat duck temperature measurement model includes a thermal infrared meat duck head detection model and a meat duck head temperature determination model.
  • the collected infrared thermal imaging images are input into the thermal infrared meat duck head detection model, and the thermal infrared meat duck head detection model The model outputs the position information of the meat duck head in the image to determine the position of the meat duck head in the image.
  • the thermal infrared meat duck head detection model specifically outputs the meat duck head ROI area, and the meat duck head temperature determination model extracts Meat duck head temperature matrix data in this image, and output the highest temperature wherein, as the head temperature of meat duck;
  • the concrete conversion step of the temperature matrix information of meat duck head temperature determination model comprises: 1, to infrared heat
  • the image and its temperature and width bars are grayscaled to obtain the grayscale image of the infrared thermal image and its temperature and width bars, where the temperature and width bars include the R component, G component and B component of each pixel; 2.
  • the meat duck head temperature determination model connects the meat duck head ROI region with the temperature matrix of the infrared thermal imaging image, and extracts the highest temperature value in the temperature matrix in the meat duck head ROI region as the temperature of the meat duck;
  • step S3 Input the growth cycle and gender information obtained in step S1 and the information collected in step S2 into the health evaluation model, and carry out abnormal warning and growth status scoring for the current physiological state of the meat duck.
  • the collection and processing of the weight information of the meat duck is as follows: the RGB-D image of the meat duck is obtained, and the obtained RGB-D image is input into the weight estimation model of the meat duck.
  • the weight estimation model of the meat duck is a convolutional neural network model.
  • the duck weight estimation model first determines the projected area of the meat duck in the RGB-D image, and outputs the estimated weight of the meat duck according to the mapping relationship between the projected area of the meat duck and the weight of the meat duck in the neural network model.
  • the behavior information collection and processing of the meat duck is as follows: obtain the color image of the meat duck, input the color image into the meat duck detection model and the meat duck behavior model respectively, and the meat duck detection model and the meat duck behavior model are neural networks Model, output the target position information of the meat duck in the color image through the meat duck detection model, input the target position information of the meat duck in the color image into the meat duck behavior model, and output the meat duck corresponding to the target position through the meat duck behavior model behavioral information.
  • the present invention also provides an inspection system based on the inspection method for physiological growth information of meat ducks, including a big data platform, an inspection robot, a data processing device, a communication device, and a plurality of information collection devices.
  • the communication device and the data processing device are mounted on the inspection robot, the inspection robot is used to move in the breeding area, the information collection device is connected to the data processing device by communication, and the data processing device uses In order to analyze and make decisions on the information collected by the information collection device, the information collection device, the data processing device and the big data platform are connected in communication with the communication device, and the communication device is used to transfer the The data of the information collection device and the data processing device are transmitted to the big data platform;
  • the data processing device includes a meat duck positioning module, a meat duck head position determination module, a meat duck head temperature determination module, and a mask segmentation module , weight estimation module, meat duck behavior module, exercise amount information collection module, feed intake and drinking water information module, and meat duck positioning module is used to receive the acquired meat duck image and obtain the
  • the inspection robot includes a fuselage, a moving wheel and a suspension, the moving wheel is connected to the fuselage through the suspension, and a power supply device, a driving device and a control device are arranged in the fuselage , the driving device and the control device are connected to the power supply device, the driving device is connected to the moving wheel to drive the moving wheel, a laser radar is connected to the top of the fuselage, and the Anti-collision radars are evenly arranged in the circumferential direction, the laser radar and the anti-collision radar are in communication connection with the control device, and the control device is in communication connection with the driving device.
  • the information collection device includes an RGB camera, an infrared thermal imaging camera and a binocular stereo camera, and a pan/tilt is arranged above the inspection robot, and the pan/tilt is connected to the inspection via a telescopic support rod. connected with the robot, the pan-tilt is rotationally connected with the support rod, the binocular stereo camera is connected above the pan-tilt, and the RGB camera and the infrared thermal imaging camera are respectively connected to two sides of the pan-tilt. side.
  • the patrol robot is also provided with a duck foot ring base station, an RFID receiver and an environmental sensor
  • the duck foot ring base station is used to receive the information of the duck foot ring installed on the meat duck
  • the RFID receiver The sensor is used to receive the information of the monitoring device of the feed bucket and the drinking bucket and the label information of the operation point
  • the environmental sensor is used to detect the environment
  • the duck foot ring base station, the RFID receiver and the environmental sensor are respectively connected with the The communication connection of the data processing device.
  • the invention monitors the physiological information, behavior information, exercise information, feed intake and drinking water information of the meat ducks, and can detect various abnormalities of the meat ducks in time, and has a comprehensive monitoring of the growth of the meat ducks, ensuring that the meat ducks Breeding health, improve the slaughter rate.
  • the physiological indicators that healthy meat ducks should have can be obtained through the growth cycle and gender information. If the collected physiological information is not within the range of indicators, it is abnormal; at the same time, if the meat ducks are injured, it will directly affect The mental state, appetite and behavioral performance of meat ducks. Therefore, it is possible to judge whether there is any abnormality in meat ducks based on behavioral information, exercise information, feed intake and drinking water information.
  • the health evaluation model of the present invention has various indicators of healthy meat ducks, which can be compared with the collected information to give scores, and at the same time give early warning when the scores are too low, which can make it easier for staff Intuitively grasp the growth of meat ducks.
  • Fig. 1 is a flow chart of a meat duck physiological growth information inspection method according to an embodiment of the present invention.
  • Fig. 2 is a flow chart of information acquisition in step S2 of the embodiment of the present invention.
  • Fig. 3 is a flow chart of information processing in step S2 of the embodiment of the present invention.
  • Fig. 4 is a flow chart of the health assessment in step S3 of the embodiment of the present invention.
  • Fig. 5 is a flow chart of the inspection robot according to the embodiment of the present invention.
  • Fig. 6 is a structural schematic diagram of a first viewing angle of an inspection robot according to an embodiment of the present invention.
  • Fig. 7 is a schematic structural diagram of a second viewing angle of an inspection robot according to an embodiment of the present invention.
  • connection should be understood in a broad sense, for example, it can be a fixed connection or a detachable connection. Connected, or integrally connected; it can be mechanically connected or electrically connected; it can be directly connected or indirectly connected through an intermediary, and it can be the internal communication of two components. Those of ordinary skill in the art can understand the specific meanings of the above terms in the present invention in specific situations.
  • a meat duck physiological growth information inspection method includes the following steps:
  • step S3 Input the growth cycle and gender information obtained in step S1 and the information collected in step S2 into the health evaluation model, and carry out abnormal warning and growth status scoring for the current physiological state of the meat duck.
  • the physiological indicators that healthy meat ducks should have can be obtained through the growth cycle and gender information. If the collected physiological information is not within the range of indicators, it is abnormal; at the same time, if the meat ducks are injured, it will directly affect The mental state, appetite and behavioral performance of meat ducks. Therefore, it is possible to judge whether there is any abnormality in meat ducks based on behavioral information, exercise information, feed intake and drinking water information.
  • the health evaluation model has various indicators of healthy meat ducks, which can be compared with the collected information and given a score. Simple and intuitive grasp of the growth of meat ducks.
  • the physiological information of the meat duck includes body temperature and body weight.
  • the body temperature information of the meat duck is collected and processed as follows: the infrared thermal imaging image of the meat duck is obtained, and the obtained infrared thermal imaging image is input into the temperature measurement model of the meat duck, and the temperature measurement model of the meat duck is a convolutional neural network model , the meat duck temperature measurement model first determines the position of the meat duck head in the thermal imaging image, and then combines the head position of the meat duck in the infrared thermal imaging image with the temperature matrix of the infrared thermal imaging image to obtain the current temperature of the duck head area Matrix value, the highest temperature value in the current area is used as the temperature value of the meat duck head.
  • an infrared thermal imaging image of a meat duck is acquired by an infrared thermal imaging camera.
  • the meat duck temperature measurement model includes a thermal infrared meat duck head detection model and a meat duck head temperature determination model.
  • the collected infrared thermal imaging images are input into the thermal infrared meat duck head detection model, and the thermal infrared meat duck head detection model outputs The position information of the head of the meat duck in the image determines the position of the head of the meat duck in the image, and the thermal infrared meat duck head detection model in this embodiment specifically outputs the ROI area of the head of the meat duck.
  • the meat duck head temperature determination model extracts the meat duck head temperature matrix data in the image, and outputs the highest temperature among them as the meat duck head temperature.
  • the specific conversion steps of the temperature matrix information of the meat duck head temperature determination model include: 1. Grayscale processing of the infrared thermal image and its temperature and width bars, and obtaining the grayscale image of the infrared thermal image and its temperature and width bars , wherein the temperature-width bar includes the R component, G component and B component of each pixel; 2. Since the temperature value corresponding to each pixel in the grayscale image of the temperature-width bar is known, it can be processed according to the corresponding temperature value The interpolation processing of each pixel on the infrared thermal image further obtains the temperature value of each pixel on the infrared thermal image. Then, the meat duck head temperature determination model combines the meat duck head ROI area with the temperature matrix of the infrared thermal imaging image, and extracts the highest temperature value in the temperature matrix in the meat duck head ROI area as the temperature of the meat duck.
  • the establishment of the thermal infrared meat duck head detection model is obtained through the retraining of the meat duck detection model.
  • the meat duck detection model is used to locate the meat duck in the image.
  • the establishment method of meat duck detection model comprises: 1, collects training data, and described training data comprises the RGB color image of meat duck group of limit bar; 2, manually marks meat duck position information; 3, uses training data as model input, will Corresponding to the position of the target meat duck as the model output, the initial model is trained based on the deep learning method to obtain the meat duck detection model.
  • the thermal infrared meat duck head detection model determines the head of the meat duck with a known position.
  • the establishment method of the thermal infrared meat duck head detection model is as follows: 1.
  • Collect the thermal infrared meat duck image training set 2. Manually mark the position of the meat duck head in the thermal infrared meat duck image data; 3. Mark the thermal infrared meat duck
  • the training data set is used as the input of the thermal infrared meat duck head detection model, and the thermal infrared meat duck head detection model is trained, and then the trained thermal infrared meat duck head detection model is obtained.
  • the body weight information collection and processing of the meat duck of the present embodiment are as follows: the body weight information collection and processing of the meat duck are as follows: obtain the RGB-D image of the meat duck, input the RGB-D image obtained into the meat duck weight estimation model, and the meat duck estimate
  • the heavy model is a convolutional neural network model.
  • the meat duck weight estimation model first determines the projected area of the meat duck in the RGB-D image, and outputs the weight of the meat duck according to the mapping relationship between the projected area of the meat duck in the neural network model and the weight of the meat duck. weight estimate.
  • the RGB-D image of the meat duck in this embodiment is collected by a binocular stereo camera.
  • the meat duck weight estimation model includes a mask segmentation sub-model and a weight estimation sub-model.
  • the mask segmentation sub-model is used to determine the meat duck outline information in the image, and the weight estimation sub-model is used to output the corresponding meat duck weight information according to the meat duck outline information.
  • the collected RGB-D image is input to the mask segmentation sub-model to obtain the segmented meat duck RGB-D mask image, and then the mask image is input into the weight estimation sub-model to obtain the estimated value of the meat duck weight.
  • the establishment method of the mask segmentation sub-model is as follows: 1. Collect the RGB-D image and the weight value of the meat duck under the image together to form a data set; 2. Manually mark the outline information of the meat duck in the RGB-D image; 3. Label the Input the data into the convolutional neural network model for training, and obtain the mask segmentation sub-model.
  • the establishment method of the weight estimation sub-model is as follows: input the marked meat duck outline information and the corresponding meat duck weight input information into the weight estimation sub-model for training, and finally obtain the trained meat duck weight estimation model.
  • the collection and processing of the behavior information of the meat duck in this embodiment are as follows: obtain the color image of the meat duck, input the obtained color image into the meat duck detection model and the meat duck behavior model respectively, and the meat duck detection model and the meat duck behavior model are both
  • the neural network model outputs the target position information of the meat duck in the color image through the meat duck detection model, inputs the target position information of the meat duck in the color image into the meat duck behavior model, and outputs the corresponding target position information through the meat duck behavior model Meat duck behavior information.
  • the establishment of the meat duck behavior model is based on the secondary training of the meat duck detection model, including: 1. Collecting training data, the training data includes color images of the meat duck behavior in the limit bar; 2.
  • the pre-trained meat duck behavior model is trained based on the deep convolutional cyclic neural network method to obtain the meat duck behavior model. It should be noted that in the model training stage, the collection of behavior image samples of meat ducks needs to cover the behaviors of eating, drinking, standing, and lying down, and then use these relatively complete behavior information as training samples, so that predictions can be made. A better model, and then when using the trained model and real-time collected meat duck behavior images to predict the behavior of meat ducks, a more accurate prediction effect can be obtained.
  • the exercise amount information collection in this embodiment is mainly to collect the walking steps of the meat duck. After obtaining the exercise data of the meat duck, count the steps of the duck per hour, calculate and analyze the exercise data of a single duck, mainly to analyze whether the number of steps per hour is too small or too much, and analyze the number of steps during the day and the number of steps at night, and analyze the number of steps during each time period during the day, and finally output the exercise data of the meat duck for that day.
  • the number of walking steps of the meat duck is collected through the duck foot ring worn on the meat duck.
  • the duck foot ring continuously records the step number information of the meat duck.
  • the step number information is in hours, that is, "how many steps per hour "Transfer to the health evaluation model.
  • Duck foot rings are the same as chicken foot rings.
  • the identity mark of the meat duck is set on the duck foot ring.
  • the meat duck is positioned according to the identity mark carried by the individual meat duck to obtain the corresponding information of a single meat duck. Growth cycle and sex.
  • physiological information, behavior information and exercise information it can correspond to the meat duck.
  • the identity of the meat duck includes its location information, so that when RGB color images, thermal infrared meat duck images and RGB-D images are acquired, the same The location can be acquired, and then the corresponding image of the same meat duck can be obtained, so as to link the physiological information and behavior information of the meat duck, and the identification includes the growth cycle and gender of the meat duck, and the acquired image can be related to the growth cycle and behavior information of the meat duck.
  • Gender correspondence in the health evaluation, it can conduct a comprehensive analysis of the growth cycle, gender, physiological information, behavior information and exercise information of a single meat duck to achieve accurate evaluation.
  • a bluetooth module is provided on the duck foot ring, which can carry out data communication during information collection, and a GPS positioning module can also be set on the duck foot ring and send position information when the image is acquired, so as to realize the positioning of a single meat duck .
  • the feed intake and drinking water information collection in this embodiment is determined by the weight of the feed bucket and drinking bucket in the duck cage where the meat duck is located, and is used to determine the average feed intake of the meat duck in the duck cage and average water intake.
  • the weight of the feed bucket and the drinking bucket is measured by the sensor installed on the feed bucket and the drinking bucket, and communicates through RFID. When the information is obtained, the data of the sensor is read through the RFID receiver.
  • step S2 the information collection process of step S2 is as follows: S2.1. Real-time shooting of color images of meat ducks, and the images are transmitted to the meat duck detection model, which outputs the corresponding target position information of meat ducks; S2.2. Duck target position information is input into the meat duck behavior model, and the model outputs the meat duck behavior information corresponding to the target position; S2.3, real-time collection of infrared thermal imaging image data, and the infrared thermal imaging image is transmitted to the thermal infrared meat duck head for detection model, the model outputs the corresponding meat duck head ROI area; S2.4, connect the meat duck head ROI area with the temperature matrix of the infrared thermal imaging image, and extract the highest temperature matrix in the meat duck head ROI area The temperature value is used as the temperature of the meat duck; S2.5, collect the RGB-D image of the meat duck in real time, input it to the mask segmentation sub-module to obtain the segmented meat duck RGB-D mask image, and then divide the mask image into Input the estimated weight sub-model to obtain the
  • the average feed intake (the weight of the current feed bucket - the weight of the previous feed bucket) ⁇ the number of meat ducks
  • the average drinking water (the weight of the current drinking bucket - the weight of the last drinking bucket) ⁇ the number of meat ducks.
  • the health evaluation model performs multi-data fusion analysis, and performs early warning and evaluation processing on the physiological growth status of meat ducks.
  • the health evaluation model in this embodiment includes a physiological state early warning model and a growth status scoring model.
  • the establishment methods of physiological state early warning model and growth state scoring model include but not limited to decision tree, random forest, support vector machine and BP neural network.
  • the physiological state early warning model provides early warning for meat ducks with abnormal physiological state. Input the collected meat duck head temperature data, meat duck behavior information, exercise information, feed intake and drinking water data into the physiological state early warning model.
  • the physiological state early warning model first detects whether the temperature of the meat duck head is in a healthy range value, when the temperature of the head of the meat duck is an abnormal value, the physiological state early warning model is based on the deviation degree of the head temperature of the meat duck, as well as the current behavior, exercise, feed intake, drinking water and other data of the meat duck. Meat ducks that meet the standards, such as abnormal body temperature and exercise, will issue an early warning of "abnormal health".
  • the growth status scoring model scores the growth status of meat ducks, which are divided into four grades: A, B, C, and D. Input the growth cycle, gender, weight data, feed intake and drinking water data of meat ducks into the growth status scoring model, and the growth status scoring model combines the current weight, growth cycle, gender, feed intake, drinking water, etc. of meat ducks, Score the current growth status of the ducks.
  • meat ducks with the same growth cycle and sex as a template compare the collected data, and perform arithmetic addition and subtraction in the range of change if the weight, feed intake, and drinking water exceed a certain range. This example is within the healthy range of weight Points can be added within the value, and points can be deducted if the value exceeds the healthy range.
  • the weight exceeds 0.1kg, add 5 points; if the weight exceeds 0.2kg, add 10 points; kg, 10 points will be deducted. Ratings are based on the duck's score.
  • the physiological growth status of meat ducks is displayed in real time during the inspection of the duck house, so that the staff can handle it on the spot.
  • the early warning and scoring results are uploaded to the Internet of Things platform, which is convenient for remote monitoring, and is conducive to the further supervision and investigation of the staff. For the relevant personnel to make further analysis and decision-making on the growth and physiological status of meat ducks.
  • the meat duck detection model, meat duck behavior model and thermal infrared meat duck head detection model of this embodiment are implemented based on the training of convolutional neural network models such as Faster R-CNN, SSD and YOLO;
  • the meat duck weight estimation model is Mask R-CNN convolutional neural network model.
  • this embodiment also collects the environmental information of the duck house, including the current ambient temperature, humidity, carbon dioxide concentration, ammonia concentration, and light intensity, so as to provide a comfortable environment for the growth of meat ducks.
  • step S2 the inspection robot collects the physiological information, behavior information, exercise information, feed intake and drinking water information of the meat duck.
  • the inspection robot is equipped with RGB camera, infrared thermal imaging camera, binocular stereo camera, duck foot ring base, RFID receiver and environmental sensing module.
  • the RGB camera is used to collect color images of meat ducks;
  • the infrared thermal imaging camera is used to collect infrared thermal imaging images of meat ducks;
  • the binocular stereo camera is used to collect RGB-D images;
  • the duck foot ring base station is used to receive the information of duck foot rings, The exercise data of the meat duck;
  • the RFID receiver is used to receive the sensor data of the feed bucket and the drinking water bucket, and is used to collect the data of the feed intake and water consumption of the meat duck;
  • the environmental sensing module is used to collect the ambient temperature and humidity, ammonia concentration, CO 2 concentration data.
  • the inspection steps in this embodiment are as follows: set up a channel for the inspection robot to move in the duck house, set up limit bars on both sides of the channel, and install RFID tags on the limit bars to mark each stop detection operation point.
  • the inspection robot relies on the lidar camera set on it to map the surrounding environment. After forming a planar map, the user sets the inspection path of the robot, and plans the operating points where the robot needs to stay and the location of the inspection robot on the inspection path.
  • the charging point and operating point are usually the center point of a certain limit bar on both sides of the channel.
  • the RFID tag of the limit bar is set at the limit bar. Before collecting data, the robot first reads the tag information of the limit bar, and starts to collect the meat duck data in the limit bar after corresponding to the limit bar information. .
  • the laser radar on the inspection robot can emit a laser beam.
  • the laser radar receives the returned beam and calculates the relative distance between the target and the laser radar according to the return time. It is used to accurately measure the relative distance between the edge of the object outline in the field of view and the equipment, and form a point cloud image based on the outline information and draw it into an environmental map to form a modeling map and further establish an inspection path.
  • the modeling map uses laser radar to carry out map modeling on a specific client and the industrial computer of the inspection robot, and then plans the path according to the built map, and sets the path stop points, and the inspection robot collects data according to the set path and stop points .
  • the inspection robot can plan the charging point according to the preset path, and can automatically return to the charging point for charging when the power of the inspection robot is low.
  • the charging point is set as the starting point and the end point, and the inspection robot performs an inspection operation every two hours, and stays for a period of time every time it passes through a column for data collection and analysis. After the inspection, the inspection robot automatically returns to the charging point and waits for the next inspection.
  • the inspection robot of this embodiment is equipped with an RGB camera, an infrared thermal imaging camera, a binocular stereo camera, an RFID receiver, a duck foot ring base station, and an environmental sensing module. Real-time monitoring of feed intake and environmental parameters is the most comprehensive meat duck inspection robot.
  • This embodiment provides a system based on the inspection method for meat duck physiological growth information in Embodiment 1, including a big data platform, an inspection robot, a data processing device 17, a communication device and a plurality of information collection devices, an information collection device, a communication
  • the device and the data processing device 17 are mounted on the inspection robot, the inspection robot is used to move in the breeding area, the information collection device is connected to the data processing device 17 in communication, and the data processing device 17 is used to analyze the information collected by the information collection device And make a decision, the information collection device, the data processing device 17 and the big data platform are connected in communication with the communication device, and the communication device is used to transmit the data of the information collection device and the data processing device 17 to the big data platform.
  • the data processing device 17 can pack all the collected data back to the big data platform, and can further process the data for relevant personnel to make further analysis and decision-making on the growth and physiological status of the meat duck.
  • the data processing device 17 of the present embodiment includes a meat duck positioning module, a meat duck head position determination module, a meat duck head temperature determination module, a mask segmentation module, a weight estimation module, a meat duck behavior module, an exercise amount information collection module, Food intake and drinking water information module, the meat duck positioning module is used to receive the acquired meat duck image and obtain the position of the meat duck in the image through the meat duck detection model; the meat duck head position determination module is used to receive the output of the meat duck positioning module The meat duck position information is obtained through the meat duck head detection model; the meat duck head temperature determination module is used to receive the meat duck head position output by the meat duck head position determination module and pass the meat duck head temperature The internal temperature determination model obtains the highest temperature of the head of the meat duck; the mask segmentation module is used to receive the acquired image and determine the outline information of the meat duck in the image through the mask segmentation sub-model; the weight estimation module is used to receive the output of the mask segmentation module The meat duck outline information and the corresponding meat duck weight information are obtained through the weight estimation
  • the data processing device 17 in this embodiment is a computer, and the data processing device 17 is installed on the inspection robot.
  • the communication device includes an antenna 18, and the antenna 18 is connected to the inspection robot.
  • the inspection robot of this embodiment includes a fuselage 1, a moving wheel 2 and a suspension 3, the moving wheel 2 is connected to the fuselage 1 through the suspension 3, and the fuselage 1 is provided with a power supply device, a driving device and a control device , the driving device and the control device are connected with the power supply device, the driving device is connected with the moving wheel 2 to drive the moving wheel 2, the laser radar 4 is connected above the fuselage 1, and the anti-collision radar 5 is uniformly arranged on the circumference of the fuselage 1, The laser radar 4 and the anti-collision radar 5 are connected in communication with the control device, and the control device is connected in communication with the driving device.
  • the suspension 3 includes a shock absorber, which can play a shock-absorbing role for the inspection robot.
  • the laser radar 4 emits a laser beam. When the laser beam travels forward and encounters an obstacle, it will return.
  • the laser radar 4 receives the reentrant beam and calculates the relative distance between the target and the laser radar 4 according to the time of return to accurately measure the field of view.
  • the relative distance between the edge of the outline of the object and the inspection robot is calculated, and the point cloud image is composed according to the outline information and drawn into an environment map to form a modeling map and further establish an inspection path.
  • the modeling map uses the laser radar 4 to carry out map modeling on a specific client and control device, and then plans the path according to the built map, and sets the path stop points, and the inspection robot collects data according to the set path and stop points.
  • the inspection robot can plan the charging point according to the preset path, and can automatically return to the charging point for charging when the power of the inspection robot is low.
  • the anti-collision radar 5 is used to detect whether there are obstacles around when the robot is cruising, so as to adjust the walking direction and route of the inspection robot in time.
  • the fuselage 1 includes a shell and a chassis, and the shell cover is arranged on the chassis so that the fuselage 1 is shaped like a quadrangular platform, and the power supply device, driving device and control are all arranged in the shell.
  • the control device in this embodiment is an industrial computer.
  • the driving device includes a servo steering gear, a servo motor and a differential gear.
  • the fuselage 1 of the present embodiment is connected with four moving wheels 2, and each moving wheel 2 is connected to a servo steering gear adjacent to it with a steering gear bracket and a servo steering gear.
  • the motor is fixedly connected with a motor bracket, and each moving wheel 2 is responsible for controlling the steering by a servo steering gear connected to it, and each moving wheel 2 is powered by a servo motor connected to it; four servo steering gears are connected with the control device , the steering of the servo steering gear is controlled by the control device, so as to realize the independent steering of the four wheels.
  • an anti-collision radar 5 is respectively installed in four directions of front, rear, left and right of the fuselage 1, which can comprehensively detect obstacles.
  • the power supply device of this embodiment includes a battery and a power manager, the battery is connected to the power manager through a wire harness, the power manager is connected to the control device, the driving device, the data processing device 17, the communication device and the information collection device through a wire harness, and the battery responsible for the power supply of the inspection robot.
  • the information acquisition device of the present embodiment comprises RGB camera 6, infrared thermal imaging camera 7 and binocular stereo camera 8, and the top of inspection robot is provided with pan-tilt 9, and pan-tilt 9 is connected with inspection robot by telescopic support rod 10 , the pan-tilt 9 is rotationally connected with the support rod 10, the binocular stereo camera 8 is connected above the pan-tilt 9, and the RGB camera 6 and the infrared thermal imaging camera 7 are respectively connected to both sides of the pan-tilt 9.
  • RGB camera 6 is used to take color images of meat ducks; infrared thermal imaging camera 7 is used to take infrared thermal imaging images of meat ducks, and binocular stereo camera 8 is used to take RGB-D images of meat ducks.
  • the RGB camera 6, the infrared thermal imaging camera 7 and the binocular stereo camera 8 are connected to the data processing device 17 through a wire harness.
  • Two servo steering gears for controlling the four-way rotation of the cloud platform 9 are installed inside the platform 9 .
  • the support rod 10 is a hydraulic telescopic rod.
  • the inspection robot is also provided with a duck foot ring base station 11, an RFID receiver 12 and an environmental sensor 13.
  • the duck foot ring base station 11 is used to receive the information of the duck foot ring installed on the meat duck
  • the RFID receiver 12 is used to Receive the information of the monitoring device of the feed barrel and drinking water barrel and the label information of the operation point
  • the environmental sensor 13 is used to detect the environment
  • the duck foot ring base station 11, the RFID receiver 12 and the environmental sensor 13 are respectively connected to the data processing device 17 by communication.
  • the duck foot ring base station 11, the RFID receiver 12 and the environment sensor 13 of this embodiment are connected with the data processing device 17 through a wire harness.
  • the RFID receiver 12 comprises a feed bucket RFID receiver, a drinking bucket RFID receiver and a limit bar RFID receiver, the feed bucket RFID receiver is used to communicate with the monitoring device on the feed bucket, and the drinking water bucket RFID receiver is used to communicate with the drinking bucket.
  • the monitoring device communicates, and the RFID receiver of the limit bar is used to read the RFID tag information set on the limit bar.
  • the environmental sensor 13 is integrated with an integrated temperature sensor, humidity sensor, carbon dioxide sensor, ammonia sensor and light intensity sensor.
  • the inspection robot of this embodiment is equipped with an RGB camera 6, an infrared thermal imaging camera 7, a binocular stereo camera 8, a duck foot ring base station 11, an RFID receiver 12, and an environmental sensor 13, which can monitor the physiological behavior and body temperature of meat ducks. , weight, exercise, average feed intake and real-time monitoring of environmental parameters, which can comprehensively monitor the growth of meat ducks.
  • the inspection robot is provided with a data processing device 17 and a communication device, so that the inspection robot of the present embodiment is an integrated design of collection of meat duck physiological growth parameters, computer communication, data processing and management, analysis and decision-making, and network transmission, Realize the real-time, automatic, continuous and efficient data collection and analysis of the entire meat duck monitoring process, which is helpful for managers to grasp the growth status of meat ducks in real time.
  • a status display screen 14 is also provided directly in front of the fuselage 1 of this embodiment, and the status display screen 14 is used to display the current task status of the inspection robot, such as the completion of duck foot ring data reception, the completion of environmental sensor detection, the RGB image Shooting is complete, etc.
  • the rear of the fuselage 1 is provided with a data display screen 15, and the data display screen 15 is used for the display of data collection and analysis results and the health status of the meat duck, such as the temperature value of the head of the meat duck, the number of steps taken by the meat duck, etc.
  • the status display screen 14 is connected to the control device, RGB camera 6, infrared thermal imaging camera 7, binocular stereo camera 8, duck foot ring base station 11, RFID receiver 12, environmental sensor 13 and data processing device 17 through the wiring harness, and the data display
  • the screen 15 is connected with the data processing device 17 through a wire harness.
  • the front side of the fuselage 1 of this embodiment is also provided with an illuminating lamp 16 .
  • the embodiment of the present invention provides a meat duck physiological growth information inspection method, which monitors the physiological information, behavior information, exercise information, feed intake and water consumption information of meat ducks, and can detect the meat ducks in time.
  • a comprehensive monitoring of the growth of meat ducks to ensure the health of meat duck breeding and improve the slaughter rate.
  • the physiological indicators that healthy meat ducks should have can be obtained through the growth cycle and gender information. If the collected physiological information is not within the range of indicators, it is abnormal; at the same time, if the meat ducks are injured, it will directly affect The mental state, appetite and behavioral performance of meat ducks.
  • the health evaluation model has various indicators of healthy meat ducks, which can be compared with the collected information and given a score. Simple and intuitive grasp of the growth of meat ducks.
  • the embodiment of the present invention also provides an inspection robot that can collect and process the above information, and can collect color images, infrared thermal imaging images and other data in real time, and use computer vision technology to analyze various physiological conditions of meat ducks in real time. Behavior, body temperature, weight, etc. of meat ducks are combined with other physiological parameters of meat ducks to provide abnormal warning and growth status score for the current physiological state of meat ducks.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Environmental Sciences (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • Animal Husbandry (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Business, Economics & Management (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Animal Behavior & Ethology (AREA)
  • Evolutionary Biology (AREA)
  • Mining & Mineral Resources (AREA)
  • Marketing (AREA)
  • Primary Health Care (AREA)
  • Marine Sciences & Fisheries (AREA)
  • Tourism & Hospitality (AREA)
  • Economics (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • General Business, Economics & Management (AREA)
  • Zoology (AREA)
  • Biodiversity & Conservation Biology (AREA)
  • Agronomy & Crop Science (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Geometry (AREA)

Abstract

一种肉鸭生理生长信息巡检方法，通过实时采集的彩色图像、红外热成像图像等数据，利用计算机视觉技术，可实时分析肉鸭的多种生理行为、肉鸭的体温、体重等，融合其它肉鸭生理参数，对肉鸭当前的生理状态进行异常预警和生长状况评分，是集合肉鸭生理生长参数采集、计算机通信、数据处理与管理、分析与决策、网络传输的一体化设计，实现整个肉鸭监测流程数据采集以及分析的实时化、自动化、连续化、高效化，有利于管理人员实时掌握肉鸭的生长状况。一种用于采集上述数据的巡检机器人，巡检机器人搭载了RGB摄像机（6）、红外热成像摄像机（7）、双目立体摄像机（8）、RFID接收器（12）、鸭脚环基站（11）、环境传感器（13）。

Description

一种肉鸭生理生长信息巡检方法及系统 技术领域
本发明涉及肉鸭养殖技术领域，特别是涉及一种肉鸭生理生长信息巡检方法及系统。
背景技术
随着我国家禽养殖产业技术和装备不断完善，肉鸭的养殖规模不断地扩大，相当多肉鸭养殖场的养殖量达十万只以上。国内规模化的养鸭舍的单位范围内养殖密度大、肉鸭活动范围有限，一旦因疾病突发死亡后，存在无法及时清理病鸭、极易引起个体间交叉感染，导致大规模的个体死亡的问题。因此，定时地巡检、观察肉鸭的生长生理状态有着十分重要的意义。传统的方法是工作人员定时地对鸭舍进行巡查逐一判别，但人工巡查存在实时性差、耗费人力、鸭舍内刺激气味有害于人体健康等缺点。同时鸭是敏感的家禽，工作人员巡检时容易引起肉鸭的应激反应，不满足动物福利的需求。因此，目前，出现了通过机器采集信息对肉鸭进行监控的技术，但是目前的监测方法只能根据鸭舍的环境状况来监控肉鸭的养殖，无法对肉鸭实际的生长生理状况进行监控。
中国专利申请CN201610246433.4(公开号为CN105867488A)公开了一种白番鸭舍在线智能监测系统，包括环境指标监测模块、环境活动视频监测模块、一号无线传输模块、中央处理器、二号无线传输模块、智能管理模块和智能控制模块，智能管理模块包括服务器、台式电脑、笔记本和手机APP，智能控制模块包括除湿机、风机、取暖设备、光照调节装置、水泵控制装置和声光报警器，环境指标监测模块、环境活动视频监测模块通过一号无线传输模块与中央处理器连接，中央处理器通过二号无线传输模块与服务器、智能控制模块连接，服务器与台式电脑、笔记本和手机APP连接。该专利只能监控鸭舌的环境状况，无法监测肉鸭的实际生长生理状况。
发明内容
本发明的目的是提供一种可全面监测且健康评价情况准确性高的肉鸭生理生长信息巡检方法及系统。
为了实现上述目的，本发明提供了一种肉鸭生理生长信息巡检方法，包括如下步骤：
S1、获取对应肉鸭的生长周期和性别；
S2、通过巡检机器人采集肉鸭的生理信息、行为信息、运动量信息、采食量及饮水量信息；肉鸭的生理信息包括体温和体重；其中，肉鸭的体温信息采集及处理如下：获取肉鸭的红外热成像图像，将得到的红外热成像图像输入肉鸭测温模型，肉鸭测温模型为卷积神经网络模型，肉鸭测温模型先确定热成像图像中肉鸭头部的位置，再将红外热成像图像中肉鸭头部位置与该红外热成像图像的温度矩阵联立，取出当前鸭头区域的温度矩阵数值，将当前区域中温度最高值作为肉鸭头部的温度值；肉鸭测温模型包括热红外肉鸭头部检测模型和肉鸭头部温度确定模型，采集到的红外热成像图像输入热红外肉鸭头部检测模型中，热红外肉鸭头部检测模型输出该图像中的肉鸭头部位置信息，确定肉鸭头部在该图像中的位置，热红外肉鸭头部检测模型具体输出肉鸭头部ROI区域，肉鸭头部温度确定模型提取该图像中肉鸭头部温度矩阵数据，并输出其中的最高温度，作为肉鸭的头部温度；其中，肉鸭头部温度确定模型的温度矩阵信息的具体转化步骤包括：1、对红外热图像及其温宽条进行灰度化处理，获取红外热图像及其温宽条的灰度图，其中温宽条包括各像素的R分量、G分量和B分量；2、由于温宽条的灰度图中各像素对应的温度值是已知的，从而可根据该对应的温度值，经过对红外热图像上每一个像素的插值处理，进一步得到红外热图像上各像素的温度值。然后，肉鸭头部温度确定模型将肉鸭头部ROI区域与该红外热成像图像的温度矩阵联立，提取肉鸭头部ROI区域中的温度矩阵中最高温度数值作为该肉鸭的温度；
S3、将步骤S1获得的生长周期和性别信息以及步骤S2采集的信息输入健康评价模型，对肉鸭当前的生理状态进行异常预警和生长状况评分。
作为优选方案，肉鸭的体重信息采集及处理如下：获取肉鸭的RGB-D图像，将得到的RGB-D图像输入肉鸭估重模型，肉鸭估重模型为卷积神经网络模型，肉鸭估重模型先确定肉鸭在RGB-D图像中的投影面积，根据神经网络模型内的肉鸭投影面积与肉鸭重量的映射关系，输出肉鸭的重量估计值。
作为优选方案，肉鸭的行为信息采集及处理如下：获取肉鸭彩色图像，将得到的彩色图像分别输入肉鸭检测模型和肉鸭行为模型，肉鸭检测模型和肉鸭行为模型均为神经网络模型，通过肉鸭检测模型输出肉鸭在彩色图像中的目标位置信 息，将肉鸭在彩色图像中的目标位置信息输入至肉鸭行为模型中，通过肉鸭行为模型输出对应目标位置的肉鸭行为信息。
本发明还提供一种基于上述肉鸭生理生长信息巡检方法的巡检系统，包括大数据平台、巡检机器人、数据处理装置、通信装置和多个信息采集装置，所述信息采集装置、所述通信装置和所述数据处理装置搭载于所述巡检机器人上，所述巡检机器人用于在养殖区域移动，所述信息采集装置与所述数据处理装置通讯连接，所述数据处理装置用于对所述信息采集装置采集到的信息进行分析并决策，所述信息采集装置、所述数据处理装置和所述大数据平台与所述通信装置通讯连接，所述通信装置用于将所述信息采集装置、所述数据处理装置的数据传输至所述大数据平台；所述数据处理装置包括肉鸭定位模块、肉鸭头部位置确定模块、肉鸭头部温度确定模块、掩膜分割模块、估重模块、肉鸭行为模块、运动量信息采集模块、采食量及饮水量信息模块，肉鸭定位模块用于接收获取的肉鸭图像并通过肉鸭检测模型得到图像中的肉鸭位置；肉鸭头部位置确定模块用于接收肉鸭定位模块输出的肉鸭位置信息并通过肉鸭头部检测模型得到图像中肉鸭头部位置；肉鸭头部温度确定模块用于接收肉鸭头部位置确定模块输出的肉鸭头部位置并通过肉鸭头部温度确定模型得到肉鸭头部的最高温；掩膜分割模块用于接收获取的图像并通过掩膜分割子模型确定图像中的肉鸭轮廓信息；估重模块用于接收掩膜分割模块输出的肉鸭轮廓信息并通过估重子模型得到对应的肉鸭重量信息；肉鸭行为模块用于接收肉鸭定位模块输出的肉鸭位置并通过肉鸭行为模型标注该肉鸭的行为；运动量信息采集模块用于统计肉鸭每小时的步数；采食量及饮水量信息模块用于统计肉鸭的平均采食量和平均饮水量。
作为优选方案，所述巡检机器人包括机身、移动轮和悬架，所述移动轮通过所述悬架与所述机身连接，所述机身内设有供电装置、驱动装置和控制装置，所述驱动装置和所述控制装置与所述供电装置连接，所述驱动装置与所述移动轮连接以带动所述移动轮，所述机身的上方连接有激光雷达，所述机身的周向均匀地设置有防撞雷达，所述激光雷达和所述防撞雷达与所述控制装置通讯连接，所述控制装置与所述驱动装置通讯连接。
作为优选方案，所述信息采集装置包括RGB摄像机、红外热成像摄像机和双目立体摄像机，所述巡检机器人的上方设有云台，所述云台通过可伸缩的支撑杆与所述巡检机器人连接，所述云台与所述支撑杆转动连接，所述双目立体摄像机 连接在所述云台的上方，所述RGB摄像机和所述红外热成像摄像机分别连接于所述云台的两侧。
作为优选方案，所述巡检机器人上还设有鸭脚环基站、RFID接收器和环境传感器，所述鸭脚环基站用于接收安装在肉鸭上的鸭脚环的信息，所述RFID接收器用于接收饲料桶和饮水桶的监控装置的信息以及作业点的标签信息，所述环境传感器用于检测环境，所述鸭脚环基站、所述RFID接收器和所述环境传感器分别与所述数据处理装置通讯连接。
与现有技术相比，本发明的有益效果在于：
本发明通过选取肉鸭的生理信息、行为信息、运动量信息、采食量及饮水量信息进行监测，可及时察觉肉鸭的各种异常情况，对肉鸭的生长有全面的监控，保证肉鸭养殖的健康，提高出栏率。在监测时，通过生长周期和性别信息，可获得健康肉鸭应该具备的生理指标，若采集到的生理信息不在指标范围内，则为异常；同时，若肉鸭有伤病，会直接影响到肉鸭的精神状态、食欲和行为表现，因此，可通过行为信息、运动量信息、采食量及饮水量信息来判断肉鸭是否有异常。本发明根据肉鸭的生长周期和性别，健康评价模型具有健康肉鸭的各项指标，可与采集到的信息进行对比，给予评分，同时在分数过低时进行预警，可让工作人员更简单直观地掌握肉鸭的生长情况。
附图说明
图1是本发明实施例的肉鸭生理生长信息巡检方法的流程图。
图2是本发明实施例的步骤S2信息获取的流程图。
图3是本发明实施例的步骤S2中信息处理的流程图。
图4是本发明实施例的步骤S3中健康评价的流程图。
图5是本发明实施例的巡检机器人的作业流程图。
图6是本发明实施例的巡检机器人的第一视角结构示意图。
图7是本发明实施例的巡检机器人的第二视角结构示意图。
图中，1-机身；2-移动轮；3-悬架；4-激光雷达；5-防撞雷达；6-RGB摄像机；7-红外热成像摄像机；8-双目立体摄像机；9-云台；10-支撑杆；11-鸭脚环基站；12-RFID接收器；13-环境传感器；14-状态显示屏；15-数据显示屏；16-照明灯；17-数据处理装置；18-天线。
具体实施方式
下面结合附图和实施例，对本发明的具体实施方式作进一步详细描述。以下实施例用于说明本发明，但不用来限制本发明的范围。
在本发明的描述中，需要说明的是，术语“中心”、“纵向”、“横向”、“上”、“下”、“前”、“后”、“左”、“右”、“竖直”、“水平”、“顶”、“底”、“内”、“外”等指示的方位或位置关系为基于附图所示的方位或位置关系，仅是为了便于描述本发明和简化描述，而不是指示或暗示所指的装置或元件必须具有特定的方位、以特定的方位构造和操作，因此不能理解为对本发明的限制。
在本发明的描述中，需要说明的是，除非另有明确的规定和限定，术语“安装”、“相连”、“连接”应做广义理解，例如，可以是固定连接，也可以是可拆卸连接，或一体地连接；可以是机械连接，也可以是电连接；可以是直接相连，也可以通过中间媒介间接相连，可以是两个元件内部的连通。对于本领域的普通技术人员而言，可以具体情况理解上述术语在本发明中的具体含义。
此外，在本发明的描述中，除非另有说明，“多个”的含义是两个或两个以上。
实施例一
如图1至图7所示，本发明优选实施例的一种肉鸭生理生长信息巡检方法，包括如下步骤：
S1、获取对应肉鸭的生长周期和性别；
S2、通过巡检机器人采集肉鸭的生理信息、行为信息、运动量信息、采食量及饮水量信息；
S3、将步骤S1获得的生长周期和性别信息以及步骤S2采集的信息输入健康评价模型，对肉鸭当前的生理状态进行异常预警和生长状况评分。
本实施例通过选取肉鸭的生理信息、行为信息、运动量信息、采食量及饮水量信息进行监测，可及时察觉肉鸭的各种异常情况，对肉鸭的生长有全面的监控，保证肉鸭养殖的健康，提高出栏率。在监测时，通过生长周期和性别信息，可获得健康肉鸭应该具备的生理指标，若采集到的生理信息不在指标范围内，则为异常；同时，若肉鸭有伤病，会直接影响到肉鸭的精神状态、食欲和行为表现，因此，可通过行为信息、运动量信息、采食量及饮水量信息来判断肉鸭是否有异常。本实施例根据肉鸭的生长周期和性别，健康评价模型具有健康肉鸭的各项指标， 可与采集到的信息进行对比，给予评分，同时在分数过低时进行预警，可让工作人员更简单直观地掌握肉鸭的生长情况。
具体地，在步骤S2中，肉鸭的生理信息包括体温和体重。
在本实施例中，肉鸭的体温信息采集及处理如下：获取肉鸭的红外热成像图像，将得到的红外热成像图像输入肉鸭测温模型，肉鸭测温模型为卷积神经网络模型，肉鸭测温模型先确定热成像图像中肉鸭头部的位置，再将红外热成像图像中肉鸭头部位置与该红外热成像图像的温度矩阵联立，取出当前鸭头区域的温度矩阵数值，将当前区域中温度最高值作为肉鸭头部的温度值。本实施例通过红外热成像摄像机获取肉鸭的红外热成像图像。肉鸭测温模型包括热红外肉鸭头部检测模型和肉鸭头部温度确定模型，采集到的红外热成像图像输入热红外肉鸭头部检测模型中，热红外肉鸭头部检测模型输出该图像中的肉鸭头部位置信息，确定肉鸭头部在该图像中的位置，本实施例的热红外肉鸭头部检测模型具体输出肉鸭头部ROI区域。肉鸭头部温度确定模型提取该图像中肉鸭头部温度矩阵数据，并输出其中的最高温度，作为肉鸭的头部温度。具体地，肉鸭头部温度确定模型的温度矩阵信息的具体转化步骤包括：1、对红外热图像及其温宽条进行灰度化处理，获取红外热图像及其温宽条的灰度图，其中温宽条包括各像素的R分量、G分量和B分量；2、由于温宽条的灰度图中各像素对应的温度值是已知的，从而可根据该对应的温度值，经过对红外热图像上每一个像素的插值处理，进一步得到红外热图像上各像素的温度值。然后，肉鸭头部温度确定模型将肉鸭头部ROI区域与该红外热成像图像的温度矩阵联立，提取肉鸭头部ROI区域中的温度矩阵中最高温度数值作为该肉鸭的温度。
其中，热红外肉鸭头部检测模型的建立是通过肉鸭检测模型再训练得到的。肉鸭检测模型用于在图像中确定肉鸭的位置。肉鸭检测模型的建立方法包括：1、采集训练数据，所述训练数据包括限位栏的肉鸭群RGB彩色图像；2、人工标注肉鸭位置信息；3、将训练数据作为模型输入，将对应目标肉鸭位置作为模型输出，基于深度学习的方法对初始的模型进行训练，得到所述的肉鸭检测模型。在肉鸭检测模型的基础上，热红外肉鸭头部检测模型对已知位置的肉鸭进行其头部的确定。热红外肉鸭头部检测模型的建立方法如下：1、采集热红外肉鸭图像训练集；2、人工标注热红外肉鸭图像数据的肉鸭头部位置；3、标注好的热红外肉鸭训练数据集作为热红外肉鸭头部检测模型的输入，对热红外肉鸭头部检测模型进行训 练，进而得出训练好的热红外肉鸭头部检测模型。
本实施例的肉鸭的体重信息采集及处理如下：肉鸭的体重信息采集及处理如下：获取肉鸭的RGB-D图像，将得到的RGB-D图像输入肉鸭估重模型，肉鸭估重模型为卷积神经网络模型，肉鸭估重模型先确定肉鸭在RGB-D图像中的投影面积，根据神经网络模型内的肉鸭投影面积与肉鸭重量的映射关系，输出肉鸭的重量估计值。本实施例的肉鸭的RGB-D图像是通过双目立体摄像机采集的。肉鸭估重模型包括掩膜分割子模型和估重子模型，掩膜分割子模型用于确定图像中的肉鸭轮廓信息，估重子模型用于根据肉鸭轮廓信息输出对应的肉鸭重量信息。采集到的RGB-D图像输入至掩膜分割子模型得出被分割的肉鸭RGB-D掩模图像，再将掩模图像其输入估重子模型，得出肉鸭重量的估计值。掩膜分割子模型的建立方法如下：1、采集RGB-D图像以及该图像下对应肉鸭的重量值共同构成数据集；2、人工标记RGB-D图像中肉鸭轮廓信息；3、将标注的数据输入卷积神经网络模型进行训练，得到掩膜分割子模型。估重子模型的建立方法如下：将标注肉鸭轮廓信息与对应的肉鸭重量输信息入估重子模型进行训练，最后得出训练好的肉鸭估重模型。
另外，本实施例的肉鸭的行为信息采集及处理如下：获取肉鸭彩色图像，将得到的彩色图像分别输入肉鸭检测模型和肉鸭行为模型，肉鸭检测模型和肉鸭行为模型均为神经网络模型，通过肉鸭检测模型输出肉鸭在彩色图像中的目标位置信息，将肉鸭在彩色图像中的目标位置信息输入至肉鸭行为模型中，通过肉鸭行为模型输出对应目标位置的肉鸭行为信息。肉鸭行为模型的建立是基于对肉鸭检测模型的二次训练，包括：1、采集训练数据，训练数据包括限位栏的肉鸭行为彩色图像；2、标注肉鸭行为信息，包括采食、饮水、站立、躺下、啄羽等。3、基于准备的训练样本数据，基于深度卷积循环神经网络的方法对预训练的肉鸭行为模型进行训练，得出肉鸭行为模型。需要说明的是，在模型训练阶段，采集肉鸭行为图像样本信息需涵括肉鸭采食、饮水、站立、躺下的行为，进而将这些较为完备的行为信息作为训练样本，从而可以得到预测效果较好的模型，进而在利用训练好的模型和实时采集的肉鸭行为图像进行肉鸭行为预测时，可以得到较为准确的预测效果。
此外，本实施例的运动量信息采集主要是采集肉鸭的行走步数。获取肉鸭运动量数据后，统计肉鸭每小时步数，计算并分析单只肉鸭的运动量数据，主要是 分析每个小时的步数是否过少或过多，和分析在白天时的步数和晚上时的步数，以及分析在白天各个时间段的步数，最后输出肉鸭该天的运动量数据。肉鸭的行走步数是通过穿戴在肉鸭上的鸭脚环进行采集的，其中鸭脚环不间断记录肉鸭的步数信息，步数信息以小时为单位，即以“每小时多少步”传输到健康评价模型。鸭脚环与鸡脚环相同。
进一步地，鸭脚环上设置该肉鸭的身份标识，在获取肉鸭的生长周期和性别信息时，根据个体肉鸭携带的身份标识对该肉鸭进行定位，以获取单只肉鸭相应的生长周期和性别。在生理信息、行为信息和运动量信息采集时，可与肉鸭相对应，该肉鸭的身份标识包括其定位信息，使RGB彩色图像、热红外肉鸭图像和RGB-D图像获取时，对同一位置进行获取，进而可获得同一肉鸭的相应图像，从而将该肉鸭的生理信息和行为信息联系起来，并且身份标识包括肉鸭的生长周期和性别，获取图像可与肉鸭的生长周期和性别相对应，在健康评价时，能对单只肉鸭的生长周期、性别、生理信息、行为信息和运动量信息进行全面的分析，实现精准评价。在本实施例在鸭脚环设置有蓝牙模块，可在信息采集时进行数据通信，并且，鸭脚环上还可设置GPS定位模块并在图像获取时发送位置信息，实现单只肉鸭的定位。
本实施例的采食量及饮水量信息采集是通过肉鸭所处的鸭笼中的饲料桶和饮水桶的重量进行确定的，是用于确定该鸭笼中的肉鸭的平均采食量和平均饮水量。饲料桶和饮水桶的重量由设置在饲料桶和饮水桶上传感器进行测量，并通过RFID进行通信，在信息获取时，通过RFID接收器读取传感器的数据。在采食量及饮水量确定时，具体地，获取此时的饲料桶和饮水桶的重量，得到肉鸭采食量和饮水量数据；接着统计该鸭笼的鸭群的采食量和饮水量，主要是将本次饲料桶和饮水桶的重量与上次获取的饲料桶和饮水桶的重量相减；之后计算并分析鸭群采食量和饮水量数据，主要是计算平均采食量和平均饮水量，具体如下：平均采食量＝(本次饲料桶重量–上次饲料桶重量)÷肉鸭数量；平均饮水量＝(本次饮水桶重量–上次饮水桶重量)÷肉鸭数量，并分析采食量和饮水量与上次相比，是否增加或减少以及变化幅度；最后输出肉鸭平均采食量和饮水量数据到健康评价模型进行评价。
综合地，步骤S2信息采集流程如下：S2.1、实时拍摄肉鸭彩色图像，图像传输至肉鸭检测模型中，该模型输出对应的肉鸭目标位置信息；S2.2、根据得出的肉鸭目标位置信息，输入至肉鸭行为模型中，该模型输出对应目标位置的肉鸭行为 信息；S2.3、实时采集红外热成像图像数据，红外热成像图像传输至热红外肉鸭头部检测模型，该模型输出对应的肉鸭头部ROI区域；S2.4、将肉鸭头部ROI区域与该红外热成像图像的温度矩阵联立，提取肉鸭头部ROI区域中的温度矩阵中最高温度数值作为该肉鸭的温度；S2.5、实时采集肉鸭的RGB-D图像，输入至掩膜分割子模块得出被分割的肉鸭RGB-D掩模图像，再将掩模图像其输入估重子模型，得出肉鸭重量的估计值；S2.6、获取鸭脚环步数信息并进行均值计算，得到该鸭群的肉鸭的运动量情况，对群体肉鸭的运动进行打分，打分准则为：日均步数8000步及以上划为S级，日均步数6000-7999步划为A级，日均步数4000-5999步划为B级，日均步数3999步及以下划为C级；S2.7、获饲料桶和饮水桶重量信息，得出与初始重量的差值信息，根据肉鸭的个数，计算出该鸭群中肉鸭的平均采食量和平均饮水量。其中，平均采食量＝(本次饲料桶重量–上次饲料桶重量)÷肉鸭数量；平均饮水量＝(本次饮水桶重量–上次饮水桶重量)÷肉鸭数量。
根据采集和处理完毕的数据，健康评价模型进行多数据融合分析，对肉鸭的生理生长状况做预警及评估处理。
本实施例的健康评价模型包括生理状态预警模型和生长状况评分模型。生理状态预警模型和生长状况评分模型的建立方法包括但不限于决策树、随机森林、支持向量机和BP神经网络。
生理状态预警模型对生理状态异常的肉鸭进行预警。将采集到的肉鸭头部温度数据、肉鸭行为信息、运动量信息、采食量及饮水量数据输入到生理状态预警模型，生理状态预警模型先检测肉鸭头部的温度是否处于健康的范围值，当肉鸭头部的温度为异常值时，生理状态预警模型再根据肉鸭头部温度的偏差程度，以及当前肉鸭的行为、运动量、采食量、饮水量等数据，对健康不达标如体温异常、运动量等异常的肉鸭做出“健康异常”预警。
生长状况评分模型对肉鸭生长状况进行评分，分为A、B、C、D四个等级。将肉鸭的生长周期、性别、体重数据、采食量及饮水量数据输入至生长状况评分模型，生长状况评分模型结合肉鸭当前的体重、生长周期、性别、采食量、饮水量等，对肉鸭当前生长状况进行评分。以同样生长周期和性别的肉鸭为模板，对采集到数据进行对比，超过一定的体重范围、采食量和饮水量范围，在变化幅度进行等差加减分，本实施例在体重健康范围值内可以采取增分，超过健康范围值采取减分，比如体重超过0.1kg，加上5分，体重超过0.2kg，加上10分；体重轻 于0.1kg则扣除5分，体重轻于0.2kg，则扣除10分。根据肉鸭的分数进行评级。本实施例在鸭舍巡检中实时显示肉鸭生理生长状态，以便工作人员现场处理，同时将预警和评分结果上传至物联网平台，方便远程监控，有利于工作人员进一步监督排查，对数据进一步的处理，以供相关人员对肉鸭的生长生理状况做出进一步分析与决策。
可选地，本实施例的肉鸭检测模型、肉鸭行为模型和热红外肉鸭头部检测模型基于Faster R-CNN、SSD和YOLO等卷积神经网络模型训练实现；肉鸭估重模型为Mask R-CNN卷积神经网络模型。
此外，本实施例还对鸭舍的环境信息进行采集，包括当前环境温度、湿度、二氧化碳浓度、氨气浓度、光照强度，以为肉鸭的生长提供舒适的环境。
本实施在步骤S2中，通过巡检机器人采集肉鸭的生理信息、行为信息、运动量信息、采食量及饮水量信息。
巡检机器人上搭载RGB摄像机、红外热成像摄像机、双目立体摄像机、鸭脚环基座、RFID接收器和环境传感模块。RGB摄像机用于采集肉鸭彩色图像；红外热成像摄像机用于采集肉鸭红外热成像图像；双目立体摄像机用于采集RGB-D图像；鸭脚环基站用于接收鸭脚环的信息，采集肉鸭的运动量数据；RFID接收器用于接收饲料桶和饮水桶的传感器数据，用于采集肉鸭采食量与饮水量数据；环境传感模块用于采集环境温湿度、氨气浓度、CO 2浓度等数据。
本实施例的巡检步骤如下：在鸭舍中设置巡检机器人移动的通道，通道的两侧设置限位栏，限位栏上安装RFID标签，用于标记各个停留检测作业点。巡检机器人依靠其上设置的激光雷达摄像机对周围环境进行地图建模，形成平面地图后，用户设定机器人的巡检路径，在巡检路径上规划机器人需要停留的作业点以及巡检机器人的充电点，作业点通常为通道两侧的某限位栏的中心点。在限位栏处设置有该限位栏的RFID标签，机器人在采集数据前首先读取该限位栏的标签信息，对应该限位栏信息后开始采集位于该限位栏内的肉鸭数据。
具体地，巡检机器人上的激光雷达可发射激光束，当激光光束向前传播遇到障碍物后会折返，激光雷达接收到折返的光束，根据折返的时间计算目标与激光雷达的相对距离，用以准确测量视场中物体轮廓边沿与设备间的相对距离，并根据轮廓信息组成点云图像并绘制成环境地图，形成建模地图，进一步建立巡检路径。建模地图利用激光雷达在特定客户端与巡检机器人的工控机进行地图建模， 后根据建好的地图规划路径，设置路径停留点，巡检机器人根据设定的路径与停留点进行采集数据。巡检机器人根据预设路径可以规划充电点，可在巡检机器人电量不足时自动返回充电点进行充电。在巡检中，将充电点设为起始点和终点，巡检机器人每两个小时执行一次巡检作业，每经过一个栏位停留一段时间，进行数据采集和分析，当完成最后一个限位栏巡检后，巡检机器人自动返回充电点，等待下一次巡检。
本实施例的巡检机器人搭载了RGB摄像机、红外热成像摄像机、双目立体摄像机、RFID接收器、鸭脚环基站、环境传感模块能够对肉鸭的生理行为、体温、体重、运动量、平均采食量以及环境参数实时监测，是监测最为全面的肉鸭巡检机器人。
实施例二
本实施例提供一种基于实施例一的肉鸭生理生长信息巡检方法的系统，包括大数据平台、巡检机器人、数据处理装置17、通信装置和多个信息采集装置，信息采集装置、通信装置和数据处理装置17搭载于巡检机器人上，巡检机器人用于在养殖区域移动，信息采集装置与数据处理装置17通讯连接，数据处理装置17用于对信息采集装置采集到的信息进行分析并决策，信息采集装置、数据处理装置17和大数据平台与通信装置通讯连接，通信装置用于将信息采集装置、数据处理装置17的数据传输至大数据平台。数据处理装置17可将采集的所有数据打包传回至大数据平台，可对数据进一步的处理，以供相关人员对肉鸭的生长生理状况做出进一步分析与决策。
本实施例的数据处理装置17包括肉鸭定位模块、肉鸭头部位置确定模块、肉鸭头部温度确定模块、掩膜分割模块、估重模块、肉鸭行为模块、运动量信息采集模块、采食量及饮水量信息模块，肉鸭定位模块用于接收获取的肉鸭图像并通过肉鸭检测模型得到图像中的肉鸭位置；肉鸭头部位置确定模块用于接收肉鸭定位模块输出的肉鸭位置信息并通过肉鸭头部检测模型得到图像中肉鸭头部位置；肉鸭头部温度确定模块用于接收肉鸭头部位置确定模块输出的肉鸭头部位置并通过肉鸭头部温度确定模型得到肉鸭头部的最高温；掩膜分割模块用于接收获取的图像并通过掩膜分割子模型确定图像中的肉鸭轮廓信息；估重模块用于接收掩膜分割模块输出的肉鸭轮廓信息并通过估重子模型得到对应的肉鸭重量信息；肉鸭行为模块用于接收肉鸭定位模块输出的肉鸭位置并通过肉鸭行为模型标注该肉鸭 的行为；运动量信息采集模块用于统计肉鸭每小时的步数；采食量及饮水量信息模块用于统计肉鸭的平均采食量和平均饮水量。因此，本实施例在巡检机器人上搭载数据处理装置17，可实时实地输出数据分析结果，便于工作人员及时收悉相应情况并方便排查。
可选地，本实施例的数据处理装置17为计算机，数据处理装置17安装在巡检机器人上。通信装置包括天线18，天线18连接在巡检机器人上。
具体地，本实施例的巡检机器人包括机身1、移动轮2和悬架3，移动轮2通过悬架3与机身1连接，机身1内设有供电装置、驱动装置和控制装置，驱动装置和控制装置与供电装置连接，驱动装置与移动轮2连接以带动移动轮2，机身1的上方连接有激光雷达4，机身1的周向均匀地设置有防撞雷达5，激光雷达4和防撞雷达5与控制装置通讯连接，控制装置与驱动装置通讯连接。悬架3包括减震器，可为巡检机器人起到减震作用。激光雷达4发射激光束，当激光光束向前传播遇到障碍物后会折返，激光雷达4接收到折返的光束，根据折返的时间计算目标与激光雷达4的相对距离，用以准确测量视场中物体轮廓边沿与巡检机器人的相对距离，并根据轮廓信息组成点云图像并绘制成环境地图，形成建模地图，进一步建立巡检路径。建模地图利用激光雷达4在特定客户端与控制装置进行地图建模，后根据建好的地图规划路径，设置路径停留点，巡检机器人根据设定的路径与停留点进行采集数据。巡检机器人根据预设路径可以规划充电点，可在巡检机器人电量不足时自动返回充电点进行充电。防撞雷达5用于机器人在巡航时探测四周是否存在障碍物，以及时调整巡检机器人的行走方向和路线。
进一步地，机身1包括外壳和底盘，外壳罩设在底盘上，使机身1成四棱台状，供电装置、驱动装置和控制均设置在外壳内。本实施例的控制装置为工控机。驱动装置包括伺服舵机、伺服电机和差速器，本实施例的机身1连接有四个移动轮2，每一个移动轮2都和与其相邻的一个伺服舵机用舵机支架和伺服电机用电机支架固定连接，每一个移动轮2都由与其相连一个伺服舵机负责控制转向，每一个移动轮2都由与其相连的一个伺服电机提供动力；四个伺服舵机与控制装置相连接，由控制装置控制伺服舵机的转向，从而实现四个车轮的独立转向。另外，本实施例在机身1的前、后、左、右四个方向上各安装了一个防撞雷达5，可全面探测障碍物。此外，本实施例的供电装置包括蓄电池和电源管理器，蓄电池与电源管理器通过线束连接，电源管理器与控制装置、驱动装置、数据处理装置17、 通信装置和信息采集装置通过线束连接，蓄电池负责巡检机器人的整机供电。
本实施例的信息采集装置包括RGB摄像机6、红外热成像摄像机7和双目立体摄像机8，巡检机器人的上方设有云台9，云台9通过可伸缩的支撑杆10与巡检机器人连接，云台9与支撑杆10转动连接，双目立体摄像机8连接在云台9的上方，RGB摄像机6和红外热成像摄像机7分别连接于云台9的两侧。RGB摄像机6用于拍摄肉鸭的彩色图像；红外热成像摄像机7用于拍摄肉鸭的红外热成像图像，双目立体摄像机8用于拍摄肉鸭的RGB-D图像。RGB摄像机6、红外热成像摄像机7和双目立体摄像机8通过线束与数据处理装置17连接。云台9的内部安装有两个用于控制云台9四向转动的伺服舵机。支撑杆10为液压伸缩杆。
另外，巡检机器人上还设有鸭脚环基站11、RFID接收器12和环境传感器13，鸭脚环基站11用于接收安装在肉鸭上的鸭脚环的信息，RFID接收器12用于接收饲料桶和饮水桶的监控装置的信息以及作业点的标签信息，环境传感器13用于检测环境，鸭脚环基站11、RFID接收器12和环境传感器13分别与数据处理装置17通讯连接。本实施例的鸭脚环基站11、RFID接收器12和环境传感器13通过线束与数据处理装置17连接。RFID接收器12包括饲料桶RFID接收器、饮水桶RFID接收器和限位栏RFID接收器，饲料桶RFID接收器用于与饲料桶上的监控装置进行通信，饮水桶RFID接收器用于与饮水桶上的监控装置进行通信，限位栏RFID接收器用于读取设置在限位栏上的RFID标签信息。环境传感器13上集成了集成温度传感器、湿度传感器、二氧化碳传感器、氨气传感器和光强传感器。
本实施例的巡检机器人搭载了RGB摄像机6、红外热成像摄像机7、双目立体摄像机8、鸭脚环基站11、RFID接收器12、环境传感13，能够对肉鸭的生理行为、体温、体重、运动量、平均采食量以及环境参数实时监测，可对肉鸭的生长进行全面的监测。并且巡检机器人上设置数据处理装置17和通信装置，使本实施例的巡检机器人为集合肉鸭生理生长参数采集、计算机通信、数据处理与管理、分析与决策、网络传输的一体化设计，实现整个肉鸭监测流程数据采集以及分析的实时化、自动化、连续化、高效化，有利于管理人员实时掌握肉鸭的生长状况。
此外，本实施例的机身1的正前方还设有状态显示屏14，状态显示屏14用于显示巡检机器人当前的任务状态，比如鸭脚环数据接收完毕、环境传感器检测完毕、RGB图像拍摄完成等。机身1的正后方设有数据显示屏15，数据显示屏15用于数据收集分析结果展示以及肉鸭健康状态，比如肉鸭头部的温度数值，肉鸭 的行走步数等。状态显示屏14通过线束与控制装置、RGB摄像机6、红外热成像摄像机7、双目立体摄像机8、鸭脚环基站11、RFID接收器12、环境传感器13连接和数据处理装置17连接，数据显示屏15通过线束与数据处理装置17连接。本实施例的机身1的前侧还设有照明灯16。
综上，本发明实施例提供一种肉鸭生理生长信息巡检方法，其通过选取肉鸭的生理信息、行为信息、运动量信息、采食量及饮水量信息进行监测，可及时察觉肉鸭的各种异常情况，对肉鸭的生长有全面的监控，保证肉鸭养殖的健康，提高出栏率。在监测时，通过生长周期和性别信息，可获得健康肉鸭应该具备的生理指标，若采集到的生理信息不在指标范围内，则为异常；同时，若肉鸭有伤病，会直接影响到肉鸭的精神状态、食欲和行为表现，因此，可通过行为信息、运动量信息、采食量及饮水量信息来判断肉鸭是否有异常。本实施例根据肉鸭的生长周期和性别，健康评价模型具有健康肉鸭的各项指标，可与采集到的信息进行对比，给予评分，同时在分数过低时进行预警，可让工作人员更简单直观地掌握肉鸭的生长情况。并且，本发明实施例还提供一种可采集上述信息并进行处理的巡检机器人，可实时采集的彩色图像、红外热成像图像等数据，利用计算机视觉技术，可实时分析肉鸭的多种生理行为、肉鸭的体温、体重等，融合其它肉鸭生理参数，对肉鸭当前的生理状态进行异常预警和生长状况评分。
以上所述仅是本发明的优选实施方式，应当指出，对于本技术领域的普通技术人员来说，在不脱离本发明技术原理的前提下，还可以做出若干改进和替换，这些改进和替换也应视为本发明的保护范围。

Claims (7)

  1. 一种肉鸭生理生长信息巡检方法，其特征在于，包括如下步骤：
    S1、获取对应肉鸭的生长周期和性别；
    S2、通过巡检机器人采集肉鸭的生理信息、行为信息、运动量信息、采食量及饮水量信息；肉鸭的生理信息包括体温和体重；其中，肉鸭的体温信息采集及处理如下：获取肉鸭的红外热成像图像，将得到的红外热成像图像输入肉鸭测温模型，肉鸭测温模型为卷积神经网络模型，肉鸭测温模型先确定热成像图像中肉鸭头部的位置，再将红外热成像图像中肉鸭头部位置与该红外热成像图像的温度矩阵联立，取出当前鸭头区域的温度矩阵数值，将当前区域中温度最高值作为肉鸭头部的温度值；肉鸭测温模型包括热红外肉鸭头部检测模型和肉鸭头部温度确定模型，采集到的红外热成像图像输入热红外肉鸭头部检测模型中，热红外肉鸭头部检测模型输出该图像中的肉鸭头部位置信息，确定肉鸭头部在该图像中的位置，热红外肉鸭头部检测模型具体输出肉鸭头部ROI区域，肉鸭头部温度确定模型提取该图像中肉鸭头部温度矩阵数据，并输出其中的最高温度，作为肉鸭的头部温度；其中，肉鸭头部温度确定模型的温度矩阵信息的具体转化步骤包括：1、对红外热图像及其温宽条进行灰度化处理，获取红外热图像及其温宽条的灰度图，其中温宽条包括各像素的R分量、G分量和B分量；2、由于温宽条的灰度图中各像素对应的温度值是已知的，从而可根据该对应的温度值，经过对红外热图像上每一个像素的插值处理，进一步得到红外热图像上各像素的温度值；然后，肉鸭头部温度确定模型将肉鸭头部ROI区域与该红外热成像图像的温度矩阵联立，提取肉鸭头部ROI区域中的温度矩阵中最高温度数值作为该肉鸭的温度；
    S3、将步骤S1获得的生长周期和性别信息以及步骤S2采集的信息输入健康评价模型，对肉鸭当前的生理状态进行异常预警和生长状况评分。
  2. 根据权利要求1所述的肉鸭生理生长信息巡检方法，其特征在于，肉鸭的体重信息采集及处理如下：获取肉鸭的RGB-D图像，将得到的RGB-D图像输入肉鸭估重模型，肉鸭估重模型为卷积神经网络模型，肉鸭估重模型先确定肉鸭在 RGB-D图像中的投影面积，根据神经网络模型内的肉鸭投影面积与肉鸭重量的映射关系，输出肉鸭的重量估计值。
  3. 根据权利要求1所述的肉鸭生理生长信息巡检方法，其特征在于，肉鸭的行为信息采集及处理如下：获取肉鸭彩色图像，将得到的彩色图像分别输入肉鸭检测模型和肉鸭行为模型，肉鸭检测模型和肉鸭行为模型均为神经网络模型，通过肉鸭检测模型输出肉鸭在彩色图像中的目标位置信息，将肉鸭在彩色图像中的目标位置信息输入至肉鸭行为模型中，通过肉鸭行为模型输出对应目标位置的肉鸭行为信息。
  4. 一种基于权利要求1-3任一项所述的肉鸭生理生长信息巡检方法的巡检系统，其特征在于，包括大数据平台、巡检机器人、数据处理装置、通信装置和多个信息采集装置，所述信息采集装置、所述通信装置和所述数据处理装置搭载于所述巡检机器人上，所述巡检机器人用于在养殖区域移动，所述信息采集装置与所述数据处理装置通讯连接，所述数据处理装置用于对所述信息采集装置采集到的信息进行分析并决策，所述信息采集装置、所述数据处理装置和所述大数据平台与所述通信装置通讯连接，所述通信装置用于将所述信息采集装置、所述数据处理装置的数据传输至所述大数据平台；
    所述数据处理装置包括肉鸭定位模块、肉鸭头部位置确定模块、肉鸭头部温度确定模块、掩膜分割模块、估重模块、肉鸭行为模块、运动量信息采集模块、采食量及饮水量信息模块，肉鸭定位模块用于接收获取的肉鸭图像并通过肉鸭检测模型得到图像中的肉鸭位置；肉鸭头部位置确定模块用于接收肉鸭定位模块输出的肉鸭位置信息并通过肉鸭头部检测模型得到图像中肉鸭头部位置；肉鸭头部温度确定模块用于接收肉鸭头部位置确定模块输出的肉鸭头部位置并通过肉鸭头部温度确定模型得到肉鸭头部的最高温；掩膜分割模块用于接收获取的图像并通过掩膜分割子模型确定图像中的肉鸭轮廓信息；估重模块用于接收掩膜分割模块输出的肉鸭轮廓信息并通过估重子模型得到对应的肉鸭重量信息；肉鸭行为模块用于接收肉鸭定位模块输出的肉鸭位置并通过肉鸭行为模型标注该肉鸭的行为；运动量信息采集模块用于统计肉鸭每小时的步数；采食量及饮水量信息模块用于 统计肉鸭的平均采食量和平均饮水量。
  5. 根据权利要求4所述的肉鸭生理生长信息巡检系统，其特征在于，所述巡检机器人包括机身(1)、移动轮(2)和悬架(3)，所述移动轮(2)通过所述悬架(3)与所述机身(1)连接，所述机身(1)内设有供电装置、驱动装置和控制装置，所述驱动装置和所述控制装置与所述供电装置连接，所述驱动装置与所述移动轮(2)连接以带动所述移动轮(2)，所述机身(1)的上方连接有激光雷达(4)，所述机身(1)的周向均匀地设置有防撞雷达(5)，所述激光雷达(4)和所述防撞雷达(5)与所述控制装置通讯连接，所述控制装置与所述驱动装置通讯连接。
  6. 根据权利要求4所述的肉鸭生理生长信息巡检系统，其特征在于，所述信息采集装置包括RGB摄像机(6)、红外热成像摄像机(7)和双目立体摄像机(8)，所述巡检机器人的上方设有云台(9)，所述云台(9)通过可伸缩的支撑杆(10)与所述巡检机器人连接，所述云台(9)与所述支撑杆(10)转动连接，所述双目立体摄像机(8)连接在所述云台(9)的上方，所述RGB摄像机(6)和所述红外热成像摄像机(7)分别连接于所述云台(9)的两侧。
  7. 根据权利要求4所述的肉鸭生理生长信息巡检系统，其特征在于，所述巡检机器人上还设有鸭脚环基站(11)、RFID接收器(12)和环境传感器(13)，所述鸭脚环基站(11)用于接收安装在肉鸭上的鸭脚环的信息，所述RFID接收器(12)用于接收饲料桶和饮水桶的监控装置的信息以及作业点的标签信息，所述环境传感器(13)用于检测环境，所述鸭脚环基站(11)、所述RFID接收器(12)和所述环境传感器(13)分别与所述数据处理装置通讯连接。
PCT/CN2022/126503 2022-01-06 2022-10-20 一种肉鸭生理生长信息巡检方法及系统 WO2023130804A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210011141.8A CN114037552B (zh) 2022-01-06 2022-01-06 一种肉鸭生理生长信息巡检方法及系统
CN202210011141.8 2022-01-06

Publications (1)

Publication Number Publication Date
WO2023130804A1 true WO2023130804A1 (zh) 2023-07-13

Family

ID=80147282

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/126503 WO2023130804A1 (zh) 2022-01-06 2022-10-20 一种肉鸭生理生长信息巡检方法及系统

Country Status (2)

Country Link
CN (1) CN114037552B (zh)
WO (1) WO2023130804A1 (zh)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114037552B (zh) * 2022-01-06 2022-05-06 华南农业大学 一种肉鸭生理生长信息巡检方法及系统
CN115119766B (zh) * 2022-06-16 2023-08-18 天津农学院 一种基于深度学习和红外热成像的母猪发情检测方法
CN115985502B (zh) * 2023-03-16 2023-06-13 华南农业大学 一种基于多元感知的狮头鹅健康决策方法及装置
CN116110586B (zh) * 2023-04-13 2023-11-21 南京市红山森林动物园管理处 基于YOLOv5以及SlowFast的大象健康管理系统

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070093965A1 (en) * 2003-05-30 2007-04-26 Harrison Harry J Use of infrared thermography in live animals to predict growth efficiency
CN106289532A (zh) * 2016-08-05 2017-01-04 西安西拓电气股份有限公司 一种红外热图像的温度提取方法及装置
CN109711507A (zh) * 2019-02-21 2019-05-03 广西走山农牧开发有限公司 一种家禽养殖管理系统
CN210909999U (zh) * 2019-09-09 2020-07-03 上海朗驰佰特智能技术有限公司 室内巡检轮式机器人
CN111626985A (zh) * 2020-04-20 2020-09-04 北京农业信息技术研究中心 基于图像融合的家禽体温检测方法及禽舍巡检系统
CN113223035A (zh) * 2021-06-07 2021-08-06 南京农业大学 笼养鸡智能巡检系统
CN113662530A (zh) * 2021-07-02 2021-11-19 华南农业大学 一种猪只生理生长状态监测预警方法
CN114037552A (zh) * 2022-01-06 2022-02-11 华南农业大学 一种肉鸭生理生长信息巡检方法及系统

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111767794A (zh) * 2020-05-28 2020-10-13 华南农业大学 基于机器视觉的笼养家禽异常行为检测方法及检测系统

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070093965A1 (en) * 2003-05-30 2007-04-26 Harrison Harry J Use of infrared thermography in live animals to predict growth efficiency
CN106289532A (zh) * 2016-08-05 2017-01-04 西安西拓电气股份有限公司 一种红外热图像的温度提取方法及装置
CN109711507A (zh) * 2019-02-21 2019-05-03 广西走山农牧开发有限公司 一种家禽养殖管理系统
CN210909999U (zh) * 2019-09-09 2020-07-03 上海朗驰佰特智能技术有限公司 室内巡检轮式机器人
CN111626985A (zh) * 2020-04-20 2020-09-04 北京农业信息技术研究中心 基于图像融合的家禽体温检测方法及禽舍巡检系统
CN113223035A (zh) * 2021-06-07 2021-08-06 南京农业大学 笼养鸡智能巡检系统
CN113662530A (zh) * 2021-07-02 2021-11-19 华南农业大学 一种猪只生理生长状态监测预警方法
CN114037552A (zh) * 2022-01-06 2022-02-11 华南农业大学 一种肉鸭生理生长信息巡检方法及系统

Also Published As

Publication number Publication date
CN114037552A (zh) 2022-02-11
CN114037552B (zh) 2022-05-06

Similar Documents

Publication Publication Date Title
WO2023130804A1 (zh) 一种肉鸭生理生长信息巡检方法及系统
Fang et al. Comparative study on poultry target tracking algorithms based on a deep regression network
CN111709333B (zh) 基于笼养鸡异常粪便的溯源预警系统及健康监测方法
CN110612921A (zh) 一种定位栏母猪的监测系统及方法
CN108876822B (zh) 一种行为危险度评估方法及家庭安防看护系统
US20170124727A1 (en) Method and device for automated parameters calculation of an object
CN104936439A (zh) 用于预测对象状态的结果的系统和方法
CN103489006A (zh) 一种基于计算机视觉的水稻病虫草害诊断方法
JP2019205425A (ja) 死亡鶏検知システム、鶏生死判定プログラム、および鶏生死判定装置
JP7376078B2 (ja) 平飼い鶏舎管理システム
KR20160092538A (ko) 센서와 무인비행체를 이용한 방목 가축 질병 예측 관리시스템 및 관리방법
TWI661770B (zh) 智慧深度學習農漁培養系統
Bhoj et al. Image processing strategies for pig liveweight measurement: Updates and challenges
US20240040995A1 (en) Method, apparatus and system for detecting carbon emission-involved gas from ruminant
CN114898405B (zh) 基于边缘计算的便携式肉鸡异常监测系统
CN113762113A (zh) 一种牲畜分娩行为监控方法及装置
CN108670256A (zh) 一种奶牛呼吸频率监测系统及方法
CN113989538A (zh) 基于深度图像的鸡群均匀度估测方法、装置、系统及介质
CN210247950U (zh) 智能检测装置和智能检测系统
CN211532340U (zh) 一种定位栏母猪的监测系统
El Moutaouakil et al. A design of a smart farm system for cattle monitoring
CN111652084B (zh) 一种异常蛋鸡识别方法及装置
CN209236152U (zh) 一种奶牛呼吸频率监测系统
CN113408334A (zh) 一种小龙虾全链条数据采集与智能检测方法及装置
KR20200080450A (ko) 딥러닝 기반 과수 작물 수확량 예측 시스템 및 방법

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: 22918256

Country of ref document: EP

Kind code of ref document: A1