WO2022070128A1

WO2022070128A1 - System and method for automatic mortality detection and management

Info

Publication number: WO2022070128A1
Application number: PCT/IB2021/058992
Authority: WO
Inventors: Mahender PAL SINGH
Original assignee: Pal Singh Mahender
Priority date: 2020-09-30
Filing date: 2021-09-30
Publication date: 2022-04-07

Abstract

According to the invention there is, therefore, provided a system, method, process and apparatus to enable the estimation of the bird mortality in a broiler farm using the 2D Computer Vision technique wherein: Bird Video Images Capturing Devices are 2D IP Camera, Central Processing Unit is a high-performance desktop computer and IP cameras remotely connect via Wi-Fi to the desktop computer. After installation, the cameras take videos of the house and send it to the desktop computer. The software installed on the computer processes the video streams into a predetermined number of frames at a fixed interval to detect individual birds and estimates its stationarity to signal a dead bird that are in camera's field of view. The alert on detection of a dead bird along with its coordinates inside the house are sent to the end user as needed.

Description

SYSTEM AND METHOD FOR AUTOMATIC MORTALITY DETECTION AND MANAGEMENT

FIELD OF THE INVENTION

This invention generally relates to the field of livestock operation and specifically to mortality detection of Chicken in Poultry Farming.

BACKGROUND

It is a known fact, that Mortality and Feed Conversion Ratio (FCR) are two critical performance indicators of any livestock operation. The feed conversion ratio (FCR) is the amount of feed ingested by an animal which can be converted into one kilo of live weight. In poultry farming, feed efficiency is a major variable to determine the cost of a kilogram of poultry meat. Along with healthy growth, a major issue is the percentage mortality which directly affects the overall yield and health of the flock.

Farm operators constantly strive to minimize mortality and improve FCR by making adjustments to farming practice and feed as livestock growth is highly sensitive to a number of controllable and uncontrollable factors. Mortality and FCR are two of the key observable measure to understand and monitor the performance of livestock. These measures can shed light on any current or future unexpected performance issues in the farm. Poultry industry, which can be further categorized as a Breeder Farm, Hatchery, Growth Farm, and Processing Plant, mortality and body weight assume importance for different reasons and resolutions, but eventually driven by the same goal of efficiency and as such a critical input to ensure that the farm performance is as per the expectations.

The present invention uses two-dimensional images visual information of the chickens to detect a dead bird using advanced image processing methods that extract features (predictors) from these images. Multiple video streams are processed at high speed to extract images at regular intervals that are fed to the Al engine. The central processing unit provides detailed information on detecting a dead bird which is further processed to send alerts for manual removal of the dead bird from the house and undertake advanced analytics reports to evaluate farm performance. The invention is developed to provide real time visibility for livestock mortality through artificial intelligence and advanced image processing technologies. A two-dimensional sensor device records the video images as an input to the Al engine. The Al engine detects a dead bird based on stationarity score after analyzing the images. The summary of the system output (whether or not dead bird(s) were detected, when and where) is sent as alerts through computers/phones/tablets for further action.

POSSIBLE PRIOR ARTS TO THE PRESENT INVENTION

The following inventions have been identified as possible prior art to the present invention:

CN207022914 relates to a flat full automatic monitoring system of birds colony health status that supports family, including base, infrared emission end and infrared count end. Infrared emitting end and infrared count end are installed on two sides respectively relatively, an infrared receiver, count end processing module and wireless output module are drawn together to infrared count hand ladle, count end processing module is used for count and data storage through the poultry, the wireless output module is used for uploading the poultry activity enumeration data that processing module was held to the count to receiving terminal or server. Will the utility model discloses be fixed in the space for activities of the flat birds of supporting the family, poultry quantity that each time quantum of count end processing module automatic recording passes through, the change through poultry activity quantity determines colony's health status. The utility model discloses relying on sensing technology monitoring poultry activity, having reduced the human input, the result is in time just accurate, has realized that this process is changed to automatic by the hand labor, the monitoring and the record that can be used to the flat birds activity of supporting the family of each growth period.

CN111198549 discloses a poultry farming monitoring and management system based on big data. The poultry farming monitoring and management system comprises an image acquisition module, an image pre-processing module, a basic feature database, a poultry feature recognition module, a farming progress overall planning module, a management server, a farming regulation and control analysis module and a parameter dynamic adjustment module. The collected image is identified and processed, an image collection distance grade and an attachment area ratio are obtained, and the actual weight level of the poultry is screened out through the image collection distance grade and the attachment area grade; the speculated weight levels corresponding to the standard farming time periods are screened out according to the Huqiu poultry farming progress time; the detected poultry farming growth error degree coefficient is counted according to the speculated weight level and the actual weight level, and thus, the error between the actual poultry weight and the expected poultry weight is analyzed, nonstandard operation in the poultry farming process can be reflected, rapid growth of poultry is promoted or the phenomena that the poultry growth speed is too low and the poultry quality is reduced are avoided.

CN109241941 discloses a method for monitoring poultry quantity in a breeding farm based on depth learning analysis. The method comprises the following steps: (1) according to a fixed sampling frequency, converting the monitoring video into a static chart to obtain a training picture data set; (2), marking that picture data set; (3), performing Gaussian convolution on each labeled picture to convert it into a density map, and the actual quantity of poultry in each picture is calculated; (4), dividing the picture data set at a ratio of 8:2 into training set and test set as the input of training model and test model respectively, training depth learning model offline, selecting the best model for poultry quantity monitoring by comparing the MAE of different parameter models on the test set; (5), real-time decoding the poultry quantity monitoring video obtained in step (4), inputting into the trained model, integrating the density map output by the model, and obtaining the poultry quantity within the monitoring range. The invention realizes the real-time monitoring of the poultry quantity more accurately.

CN109360650 relates to a software design livestock and poultry disease rapid diagnosis and treatment system. The existing theoretical knowledge of a livestock disease only aims at a single disease and the attack mid-term symptom of the single disease, but the actual clinical practice of a veterinarian is mixed infection, which is difficult to diagnose. There is still a big difficulty in applying a theory to actual diagnosis and treatment. The invention relates to the software design livestock and poultry disease rapid diagnosis and treatment system. A user side is a choice question according to a livestock and poultry disease broad heading design characteristic. The choice question carries out classification from the three aspects of a diseased livestock and poultry situation, general symptom examination and autopsy examination. The system is designed according to the routine clinical diagnosis and treatment method of the livestock and poultry, the veterinarian can provide disease information according to diagnosis and treatment habits, the disease can be diagnosed through software, and the similar diseases and the mixed infections which are difficult to identify through diagnosis and treatment can be diagnosed and a result and a treatment method are acquired; and the accuracy and the cure rate of disease diagnosis and treatment can be greatly increased, and the mortality of the livestock and poultry is reduced.

IN201841046061 helps in explaining how effectively technology like IOT, Imagery analytics helps the farmer to control & reduce the bird’s mortality rate, and also recommends medicine dosage for birds by allowing remote monitoring. The doctors or bird specialist who take advantages of advanced analytics (i.e. Imagery analytics) to decide the bird’s health and probable recommendation for medicines to control the mortality and disease spreading across poultry. Images taken from the poultry will be stored in an unstructured store with its supporting metadata and data. The algorithmic imagery models will help to identify the deficiencies and recommend the dosages according to the color and stains in different parts of the body.

PROBLEMS WITH THE PRIOR ART

MORTALITY

Detecting mortality in a poultry farm is a critical task that faces the following practical challenges:

1. It is a highly manual process, requiring multiple visits daily to the farms, which are quite often located in remote areas.

2. Constant interaction between chicken and humans is not advisable on account of animal welfare and bio-security.

3. Detection timing is of the essence since the dead bird should be removed at the earliest possible time to avoid spreading infection but that is not possible when this task is manual and performed at a scheduled time.

4. It is important to conduct postmortem asap to know the cause of death so that preventative actions can be taken to avoid flock wide spread of possible diseases if death occurred due to some ailment. This requirement is extremely difficult to manage even if multiple visits to the farm are undertaken on a fixed schedule. 5. There is a constant fear of theft since the farm hand can disguise stolen chicken as mortality numbers resulting in revenue leakage. This not only affects the mortality index, but also the FCR.

MORTALITY DETECTION

At present, the mortality is detected by regular visits to the farm and visual inspection. Although intuitive, this method involves a mandatory physical visit to the house regularly and visual inspection of the whole house to find a dead bird, if any. Being a highly manual process, it depends heavily on the available manpower as well as the expertise, efficiency and strict adherence to the visit protocol of the human resource and has limitations associated in the way it is carried out. Any excess or inefficient handling by farm hands can potentially stress the birds adversely, impacting their well-being and retarding their growth rate. As such, this method should only be undertaken on a periodic basis to minimize human handling. Since it is done periodically, certain rules need to be followed strictly to increase the accuracy of this method which is problematic.

Some inherent inadequacies of the process include:

• Failure to detect all dead birds at every visit;

• Time required to detect all dead birds (on many days there may not be any dead bird so this task results in complete wastage of a resource, but it is unavoidable if we are using a manual process);

• Negligence resulting in inaccurate recording of the numbers, time and location of the dead bird;

• Establishing the visit schedule as it may depend on the overall health, season, flock density etc. In other words, it is important to allow schedule flexibility to maximize the effectiveness of the task, which is a significant challenge.

OBJECTS OF THE INVENTION

• The main object of the invention is aimed at automatic detection of bird mortality using the two-dimensional (2D) Computer Vision technique.

• A further object of the invention is to provide the data in real time and on-demand. • A further object of the invention is to send an alert on detection of a dead bird to ensure timely action can be taken to remove the dead bird.

• A further object of the invention is to eliminate unnecessary visits to the house that are undertaken only to detect birds in a house by way of human visual inspection.

• A further object of the invention is to analyze the dead bird data over time to detect patterns in mortality that could shed light on structural deficiencies of the house or system operations which could lead to actions that will reduce mortality.

STATEMENT AND SUMMARY OF THE INVENTION

According to the invention there is, therefore, provided a system, method, process and apparatus to enable the estimation of the bird mortality in a broiler farm using the 2D Computer Vision technique wherein:

• Bird Video Images Capturing Devices are 2D IP Cameras.

• Central Processing Unit is a high-performance desktop computer.

• IP cameras remotely connect via Wi-Fi to the desktop computer.

Along with these hardware components another important part of this invention is the Artificial Intelligence (Al) engine which detects the dead bird from the 2D photos extracted from multiple 2D cameras video streams. After installation, the cameras take videos of the house and send it to the desktop computer. The software installed on the computer processes the video streams into a predetermined number of frames at a fixed interval to detect individual birds and estimates its stationarity score to signal a dead bird that are in camera’s field of view. The alert on detection of a dead bird along with its coordinates inside the house are sent to the end user as the case may be.

Our invention differs from the prior art devices by incorporating the following concepts:

• eliminates the need for human intervention.

• uses only 2D images for detecting dead birds.

• images are sent via a Local Area Network to a computer located in each house on which the mortality detection engine is deployed. The output of this engine is in the form of text data and small size gifs on the local computer that are sent to a central database on the Cloud. This allows the system to keep the data sent to the Cloud to a minimum.

• data sent to the Cloud is further processed for reporting and analytics.

• it enables real time mortality detection visibility in the house.

• it enables remote tracking of the mortality in the house.

• an array of 2D IP cameras is deployed to cover the whole house to ensure that a dead bird anywhere in the house is detected.

• the system operates 24X7 for sending timely alerts.

• it can be either installed at fixed locations or mounted on a mobile mechanism to include conveyor or a drone to ensure full coverage without investing in too many cameras.

• if the cameras are fixed, we increase its coverage using the Pan, Tilt and Zoom capabilities to scan a wider area to reduce the number of cameras. In order to ensure optimal coverage with least number of cameras, we use a custom developed optimization engine that recommends the camera installation points based on the house dimensions and layout.

• if the system is mounted on a moving setup, it can take images from any section of the farm with pinpoint information of the location.

• it has reasonable installation & setup cost to deliver highly accurate results.

• it is an easy to deploy solution that can be functional very quickly.

DETAILED DESCRIPTION

The description of the preferred embodiment is meant to demonstrate the broad working principles of the invention without limitation as to possible adaptations, extensions, applications etc., which would be obvious to a person skilled in the art. In the interest of brevity and for the purposes of exemplary explanation, references have been made to a system, described herein without limitation, to describe the invention which is essentially directed toward catering to the problem of real-time mortality detection of the livestock.

Figure 1 describes the base system architecture. COMPONENTS OF THE MORTALITY DETECTION DEVICE

The mortality detection system/method comprise the following units:

• Image Capturing Device (a standard 2D IP camera)

• Central Processing Unit (Desktop/Laptop Computer)

• Connectivity Device (LAN Router)

IMAGE CAPTURING DEVICE

The mortality detecting system/method consist of an array of IP Cameras which records 2D video images of the area of interest which could be the whole house. Images are main input data to the system’s Al engines. Images are collected continuously throughout the day. These are standard IP cameras with specific features that allow the system to extend its field of view by using Pan, Tilt and Zoom, for instance HIKVISION 4 Megapixel IP Camera DS-2DE3A404IW- DE / W (2.8-12 mm) 4 x IR Wi-Fi Network PTZ Camera that can be replaced with any similarly capable cameras easily.

CENTRAL PROCESSING UNIT (DESKTOP/LAPTOP COMPUTER)

The system uses any high-performance computer on which the Al engines are installed.

• Video processing algorithm extracts 2D frames from all camera streams.

• Using these extracted images, the Al engines determines the stationarity index of every image processed and detect mortality at any given time.

• Based on this stationarity index it establishes the mortality score of the bird.

Any computational device capable of processing a large number of images in real time, for instance, Dell G5 15 5590 Gaming Laptop, Dell G5 15 559020Q25 but any computer with similar features will serve the needs of the system.

CONNECTIVITY DEVICE

A local Wi-Fi signal receiver is installed on the system which connects multiple IP Cameras to the Computer inside the house. The video streams are sent via the local Wi-Fi setup to the computer for further processing. On detection of a dead bird, the information may be sent to cloud servers using an external Wi-Fi connection for access through web or mobile applications.

Any standard router that allows the house to be fully WiFi enabled can be used to deploy the system, for instance, TP-Link Wi-Fi Wireless LAN Router, 1 lac 1733 + 800 Mbps, MU-MIMO Gigabit Archer A 10, Archer A 10 but the same can be easily substituted by any similarly capable router.

SYSTEM AND WORKING

In one embodiment of the present invention, the necessary hardware and the underlying methods can be trained and used for the breeder and layer farms as well as for the aqua, swine, cattle and other livestock for enhancing farm performance using effective and timely mortality detection.

In another embodiment of the invention, any type of 2D image capturing device can be used to acquire images of the birds as input into the mortality detection engine.

In another embodiment of the invention, by installing various additional sensors, the same device can be used to capture multiple parameters in real time that can generate insights for preventative measures to ensure good health and minimize mortality of the birds in the farm at a lower cost.

Claims

We claim:

1. A livestock mortality estimation system comprising: an image capturing element, a local processing unit, a connectivity element and a protective component.

2. The system as claimed in claim 1 wherein the image capturing element is a standard two- dimensional internet protocol camera with specific features that allow the system to extend its field of view by using pan, tilt and zoom etc.

3. The system as claimed in claim 1 wherein the local processing unit is a microcomputer.

4. The system as claimed in claim 1 wherein the connectivity element is a Wi-Fi module installed on the microcomputer in the system which connects to a Wi-Fi network.

5. The system as claimed in claim 1 wherein the protective component comprises a surge protector that guards against voltage spikes, a casing to house all the other components 10 and a cable to connect to the power source.

6. The system as claimed in claim 3 wherein the microcomputer is installed with an Artificial Intelligence (Al) engine.

7. The system as claimed in claim 2 wherein the IP camera is remotely connected via Wi-Fi network to the local microcomputer.

8. The system as claimed in claim 2 wherein the IP camera can be either installed at fixed locations or mounted on a mobile mechanism to include conveyor or a drone.

9. A livestock mortality estimation method comprising;

- multiple two-dimensional video streams from multiple image capturing element,

- transferring these video streams via Wi-Fi connection on local area network to the local microcomputer,

- processing of the video streams on the local microcomputer into a predetermined number of frames at fixed intervals,

- extraction of predictive features from the images, - analysis of extracted features to estimate the stationarity index for every image processed using the artificial intelligence engine to detect mortality at any given time and

- transferring the mortality score in a text file to application to a central database on the Cloud servers using the Wi-Fi connection.

10. The method as claimed in claim 9 wherein there is a provision of generation of alert signal through the mortality score which can be sent to the end user for further action.

11. The method as claimed in claim 9 wherein the mortality estimation is executed on a real time basis on the local microcomputer continuously or as per a predetermined schedule or remotely triggered on demand as required.

12. The method as claimed in claim 9 wherein text file generated as output by the Al engine on the local microcomputer doesn’t require too much bandwidth and can be sent to the Cloud based system for further access by users through computers/mobile phones/tablets.