CN-116311008-B - Abnormal chicken coop identification method, device and system and robot

CN116311008BCN 116311008 BCN116311008 BCN 116311008BCN-116311008-B

Abstract

The invention provides a method, a device, a system and a robot for identifying an abnormal chicken coop, belonging to the technical field of livestock and poultry breeding; the sampling image is identified to obtain a first identification result and a second identification result respectively, the first identification result comprises the number of chicken heads, the number of eggs and the cockscomb erection state of each chicken in the target chicken coop, the second identification result comprises cockscomb color information of each chicken, the first identification result and the second identification result are input into a decision tree model, and the category of the target chicken coop is determined. According to the abnormal chicken cage identification method, device and system and robot, based on multi-sensor coordination and multi-source data fusion, the chicken individuals in the chicken cage are comprehensively extracted in the internal and external characteristics and then sent to the decision tree judgment model for classification, so that the data is more reliable, and the classification result is more accurate.

Inventors

LI BIN
LIANG XUEWEN
ZHAO YULIANG
JIANG LIN
JIA NAN
ZHU JUN
ZHAO WENWEN
WANG HAIFENG
LIU LIRONG

Assignees

北京市农林科学院智能装备技术研究中心

Dates

Publication Date: 20260508
Application Date: 20220909

Claims (11)

1. An abnormal chicken coop identification method, comprising the steps of: Acquiring a sampling image of a target chicken coop; the sampling image is identified to respectively obtain a first identification result and a second identification result, wherein the first identification result comprises the number of chicken heads, the number of eggs and the cockscomb erection state of each chicken in the target chicken coop; Inputting the first recognition result and the second recognition result into a decision tree model, and determining the category of the target chicken coop; the categories include low-laying chicken cages, normal chicken cages, and dead chicken cages; acquiring sound information of each chicken in the target chicken coop; correspondingly, inputting the sound information, the first recognition result and the second recognition result into a decision tree model, and determining the category of the target chicken coop; the obtaining the sound information of each chicken in the target chicken coop comprises the following steps: Collecting sound signals of each chicken in the target chicken coop within a preset time length; acquiring characteristic parameters related to the sound signal, wherein the characteristic parameters are linear prediction cepstrum coefficients extracted after a starting point and an ending point of the sound signal are determined through analysis of the zero crossing rate and the energy of the sound signal; Inputting the characteristic parameters into a pre-trained sound classification model to acquire the sound information output by the sound classification model; the sound classification model is obtained by training by using a hidden Markov model as an initial model.
2. Method for identifying an abnormal chicken coop according to claim 1, wherein identifying the sampled image, obtaining the first identification result comprises: Inputting the sampling images into a pre-trained chicken head detection network model to obtain a plurality of chicken head sub-images in the sampling images, and obtaining the number of the chicken head sub-images as the number of chicken heads in the target chicken coops; Inputting the sampling images into a pre-trained egg detection network model to obtain a plurality of egg sub-images in the sampling images, and obtaining the number of the egg sub-images as the number of eggs in the target chicken coops; Inputting the sampling images into a pre-trained cockscomb detection network model to obtain a plurality of cockscomb sub-images in the sampling images, and determining the cockscomb standing state in each cockscomb sub-image.
3. The method for identifying an abnormal chicken coop according to claim 1, wherein identifying the sampled image and obtaining the second identification result specifically comprises: Inputting the sampling image into a pre-trained cockscomb segmentation network model to obtain at least one cockscomb sub-image segmented in the sampling image output by the cockscomb segmentation network model; and determining the cockscomb color information of each chicken according to the distribution of RGB values of each cockscomb sub-image.
4. The method for identifying abnormal chicken cages according to claim 1, further comprising obtaining body temperature information of each chicken in the target chicken cage; correspondingly, the sound information, the body temperature information, the first recognition result and the second recognition result are input into a decision tree model, and the category of the target chicken coop is determined.
5. The method for identifying abnormal chicken cages according to claim 4, wherein the acquiring the body temperature information of each chicken in the target chicken cage comprises: collecting the body surface temperature of each chicken at a plurality of sampling moments; and (3) based on a least square method, performing binary linear regression on the body surface temperature of each chicken at the sampling moments so as to acquire the body temperature information of each chicken.
6. The method of identifying an abnormal chicken coop of claim 4, wherein the inputting the sound information, the body temperature information, the first identification result, and the second identification result into a decision tree model, determining the category of the target chicken coop comprises: based on a pre-constructed characteristic data table for determining the category to which the hencoop belongs, the characteristic data table records the mapping relation of sound information, body temperature information, chicken head number matching degree, egg number matching degree, cockscomb standing state and cockscomb color information and the category to which the hencoop belongs; Generating a confusion matrix table according to the sound information, the body temperature information, the chicken head number matching degree, the egg number matching degree, the cockscomb standing state and the cockscomb color information related to the target cockscomb by using the characteristic data table, so as to calculate the probability that the target cockscomb belongs to any category according to the confusion matrix table; According to the probability that the target coop belongs to any category, respectively determining information entropy corresponding to sound information, body temperature information, chicken head number matching degree, egg number matching degree, cockscomb standing state and cockscomb color information; determining the information gain corresponding to the sound information, the body temperature information, the chicken head number matching degree, the chicken comb standing state and the chicken comb color information according to the information entropy corresponding to the sound information, the body temperature information, the chicken head number matching degree, the chicken comb standing state and the chicken comb color information respectively; Respectively carrying out normalization processing on the sound information, the body temperature information, the chicken head number matching degree, the egg number matching degree, the cockscomb standing state and the information gain corresponding to the cockscomb color information, and determining weights corresponding to the sound information, the body temperature information, the chicken head number matching degree, the egg number matching degree, the cockscomb standing state and the cockscomb color information; Calculating the confidence coefficient of the target chicken coop with the type of dead chicken coop according to the chicken head number matching degree, the corresponding weight, the body temperature information and the corresponding weight; Calculating the confidence coefficient of the category of the target chicken coop as a low-laying chicken coop according to the egg number matching degree and the corresponding weight thereof, the cockscomb standing state and the corresponding weight thereof, the cockscomb color information and the corresponding weight thereof, the body temperature information and the corresponding weight thereof, and the sound information and the corresponding weight thereof; Determining the category of the target chicken coop according to the confidence that the category of the target chicken coop is a dead chicken coop and the confidence that the category of the target chicken coop is a low-laying chicken coop; The chicken head number matching degree is determined by comparing the number of identified chicken heads in the target chicken coop with the standard number of chicken heads in the target chicken coop, and the egg number matching degree is determined by comparing the number of identified eggs in the target chicken coop with the standard number of eggs in the target chicken coop.
7. The method for identifying abnormal chicken cages according to claim 1, wherein the upper left corner of each chicken cage in the chicken house is provided with two-dimensional codes with the same shape, and each two-dimensional code is used for storing the position information of the chicken cage in which the two-dimensional code is located; The obtaining the sampling image of the target chicken coop comprises the following steps: Any henhouse image acquired by a sampling camera is acquired; acquiring coordinate information of two adjacent two-dimensional codes in the henhouse image to determine hencoop coordinate information of the target hencoop in the henhouse image, wherein the target hencoop is a hencoop contained in a region defined by the same side edge of the two adjacent two-dimensional codes in the henhouse image; cutting out a sampling image of the target chicken coop from the chicken coop image according to the chicken coop coordinate information.
8. The method for identifying an abnormal chicken coop according to claim 7, wherein acquiring coordinate information of two adjacent two-dimensional codes in the chicken coop image to determine chicken coop coordinate information of the target chicken coop in the chicken coop image comprises: according to the coordinate information of each two-dimensional code in the henhouse image, the world coordinate information of each two-dimensional code is determined by combining the optical center coordinates of the target henhouse distance sampling camera, the internal parameters of the sampling camera and the focusing distance of the target henhouse; And determining the coop coordinate information of the target coop according to the world coordinate information of the two-dimensional codes.
9. The method of identifying abnormal chicken cages according to claim 7, further comprising, after determining the category of the target chicken cage: And positioning a virtual area of the target chicken coop in the virtual chicken coop diagram, and marking the category of the target chicken coop in the virtual area.
10. An apparatus for identifying abnormal chicken coops, comprising: the image acquisition unit is used for acquiring a sampling image of the target chicken coop; The image detection unit is used for identifying the sampling image to respectively acquire a first identification result and a second identification result, wherein the first identification result comprises the number of chicken heads, the number of eggs and the cockscomb standing state of each chicken in the target chicken coop; The chicken coop classification unit is used for inputting the first identification result and the second identification result into a decision tree model and determining the category of the target chicken coop; the categories include low-laying chicken cages, normal chicken cages, and dead chicken cages; acquiring sound information of each chicken in the target chicken coop; correspondingly, inputting the sound information, the first recognition result and the second recognition result into a decision tree model, and determining the category of the target chicken coop; the obtaining the sound information of each chicken in the target chicken coop comprises the following steps: Collecting sound signals of each chicken in the target chicken coop within a preset time length; acquiring characteristic parameters related to the sound signal, wherein the characteristic parameters are linear prediction cepstrum coefficients extracted after a starting point and an ending point of the sound signal are determined through analysis of the zero crossing rate and the energy of the sound signal; Inputting the characteristic parameters into a pre-trained sound classification model to acquire the sound information output by the sound classification model; the sound classification model is obtained by training by using a hidden Markov model as an initial model.
11. An abnormal chicken coop identification system, characterized in that a method of identifying an abnormal chicken coop as claimed in any one of claims 1-9 is run; the system also comprises a user terminal, and an identity mark, a two-dimensional code, a sound sensor and a temperature sensor which are arranged on each chicken coop; The voice sensor is used for collecting voice information of chickens in each chicken coop, the temperature sensor is used for collecting temperature information of chickens in each chicken coop, and each identity mark stores serial number information of the chicken coop where the identity mark is located, and voice information and temperature information collected at each sampling moment uploaded by the voice sensor and the temperature sensor.

Description

Abnormal chicken coop identification method, device and system and robot Technical Field The invention relates to the technical field of livestock and poultry breeding, in particular to a method, a device and a system for identifying abnormal chicken cages and a robot. Background Large-scale cultivation becomes a main cultivation mode in recent years. In the breeding process, the breeding benefits of enterprises can be seriously affected by the dead chickens and the low-yield laying hens, and the dead chickens are timely identified and cleaned, so that the method has important significance. The manual inspection of dead chickens and low-yield laying hens has the problems of high labor intensity, low accuracy, untimely discovery and the like, and the adoption of intelligent inspection equipment to replace manual work is particularly important. Detection and identification equipment for dead chickens and low-yield laying hens appears in the market at present, but most of the detection and identification equipment has the defects of single physical sign index classification, low identification rate and no popularization. The development of efficient and accurate identification and positioning methods by means of related novel means is an urgent need. Disclosure of Invention The invention provides a method, a device, a system and a robot for identifying abnormal chicken cages, which are used for solving the defects of high labor intensity and low identification precision in the prior art that dead chicken and low-yield chicken are required to be identified manually and realizing accurate identification and positioning of the chicken cages with dead chicken and low-yield chicken. In a first aspect, the invention provides a method for identifying an abnormal chicken coop, comprising the steps of obtaining a sampling image of a target chicken coop; the sampling image is identified to respectively obtain a first identification result and a second identification result, wherein the first identification result comprises the number of chicken heads, the number of eggs and the cockscomb erection state of each chicken in the target chicken coop; Inputting the first recognition result and the second recognition result into a decision tree model, and determining the category of the target chicken coop; the categories include low-laid chicken cages, normal chicken cages, and dead chicken cages. According to the method for identifying the abnormal chicken coops provided by the invention, the step of identifying the sampling image and the step of obtaining the first identification result comprise the following steps: Inputting the sampling images into a pre-trained chicken head detection network model to obtain a plurality of chicken head sub-images in the sampling images, and obtaining the number of the chicken head sub-images as the number of chicken heads in the target chicken coops; Inputting the sampling images into a pre-trained egg detection network model to obtain a plurality of egg sub-images in the sampling images, and obtaining the number of the egg sub-images as the number of eggs in the target chicken coops; Inputting the sampling images into a pre-trained cockscomb detection network model to obtain a plurality of cockscomb sub-images in the sampling images, and determining the cockscomb standing state in each cockscomb sub-image. According to the identification method of the abnormal chicken coop provided by the invention, the sampling image is identified, and the second identification result is obtained, which comprises the following steps: Inputting the sampling image into a pre-trained cockscomb segmentation network model to obtain at least one cockscomb sub-image segmented in the sampling image output by the cockscomb segmentation network model; and determining the cockscomb color information of each chicken according to the distribution of RGB values of each cockscomb sub-image. The method for identifying the abnormal chicken coop further comprises the steps of obtaining sound information of each chicken in the target chicken coop; Correspondingly, inputting the sound information, the first recognition result and the second recognition result into a decision tree model, and determining the category of the target chicken coop. According to the method for identifying the abnormal chicken coops provided by the invention, the method for acquiring the sound information of each chicken in the target chicken coops comprises the following steps: Collecting sound signals of each chicken in the target chicken coop within a preset time length; acquiring characteristic parameters related to the sound signal, wherein the characteristic parameters are linear prediction cepstrum coefficients extracted after a starting point and an ending point of the sound signal are determined through analysis of the zero crossing rate and the energy of the sound signal; Inputting the characteristic parameters into a pre-trained sound c