Method for scoring the physical condition score of cows by image processing

Abstract

The purpose of designing this system is to introduce a way to change the manual and high-error BCS measurement with an automatic and reliable technique. Increasing the accuracy of measuring this parameter is very important for increasing the health and reproduction of livestock. Therefore, we intend to measure the BCS by using digital images taken from the back of the cow and then applying machine learning estimating algorithms. Therefore, after passing through a certain gate, each cow's information is stored by RFID reader and then image is taken from the back. This data is sent to the main server with an unlimited power supply and its main features are extracted.

Description

method for scoring the physical condition score of cows by Image processingTechnical Field

[0001] This invention relates to monitoring the health and improving pregnancy and nutrition of cows.Background Art

[0002] ARRANGEMENT AND METHOD FOR DETERMINING A BODY CONDITION SCORE OF AN ANIMAL is published by WO / 2010 / 063527 Number. According to one aspect of the invention an arrangement for determining a body condition score of an animal is provided, the arrangement comprising a three-dimensional camera system directed towards the animal and provided for instantaneously recording at least one three-dimensional image of the animal; and an image processing device connected to the three-dimensional camera system and provided for forming a three-dimensional surface representation of a portion of the animal from the three-dimensional image recorded by the three-dimensional camera system; for statistically analyzing the surface of the three-dimensional surface representation, in particular analyzing the unevenness, irregularity, or texture of the surface of the three-dimensional surface representation; and for determining the body condition score of the animal based on the statistically analyzed surface of the three-dimensional surface representation.

[0003] The claimed invention does not use the three-dimensional camera to take the image from the animal.Summary of Invention

[0004] The purpose of designing this system is to introduce a way to change the manual and high-error BCS measurement with an automatic and reliable technique. Increasing the accuracy of measuring this parameter is very important for increasing the health and reproduction of livestock.

[0005] Therefore, we intend to measure the BCS by using digital images taken from the back of the cow and then applying machine learning estimating algorithms.Therefore, after passing through a certain gate, each cow’s information are stored by RFID reader and then image is taken from the back. This data is sent to the main server with an unlimited power supply and its main features are extracted.

[0006] The BCS value is then calculated from them. For analysis and also the availability of BCS value, the obtained results are uploaded on the mobile application or tablet of an expert or farmer.

[0007] Implementing an automatic BCS estimator system will increase the accuracy of the BCS parameter as well as reduce the staff. However, there are always challenges in implementing such systems. The proposed system should be able to record, process and regularly estimate images of cows.

[0008] Also, the database and related hardware must have the capacity to store a large volume of images. On the other hand, information about the time of birth, number of offspring and physical condition of each cow, in addition to the BCS parameter, should be stored in a separate database and the values obtained from the BCS should be checked regularly by an expert.Technical Problem

[0009] Body condition score (BCS) is a method to measure the amount of subcutaneous body fat or energy reserve in cows, regardless of body weight and frame size and it evaluates with numbers from 1 to 5. BCS is an important indicator for appropriate feeding management as well as improve herd nutrition, health, production, and pregnancy rate. However, BCS measuring is extremely difficult and time-consuming process for a large-scale farmer that measured manually by skilled person.

[0010] Because BCS are measured by humans, they may be erroneous. For instance, the numbers obtained by human observers after each observation are the same in 58.1% of cases. In 32.6% of cases, there are 0.25% units and in 6.8% of cases 0.5 difference respectively. Also, changes of close to 0.25 are not achieved by semi-skilled evaluators. With the advancement of technology in the agricultural and livestock industry, many researchers have shifted to the use of automated evaluation systems. Systems that measure and analyze the numberof BCS in real time and with higher accuracy than humans. The system must continuously monitor the process of tissue cover changes in cows and periodically announce the amount of BCS in the real-time and with high accuracy.

[0011] Automatic BCS as an intelligent system can help farmers and improve the performance of animal health, production, and pregnancy rate. It is very essential that the proposed intelligent assistant uses cheap hardware features to accurately estimate the amount of BCS. Therefore, the three main parameters that should be considered in the proposed system are efficiency, accuracy and being real-time process.Solution to Problem

[0012] The device includes a main camera and a boot computer system. The main camera communicates with the system via a network protocol. Also, to measure the livestock identification number and the time of taking the desired photo, other hardware has been designed that is installed as a gate in front of the livestock.

[0013] In fact, the animals are guided in an enclosed direction, and as soon as they reach the RFID gate, the cow identification number is read from its RFID neck tag, and the camera is also instructed to take a photo of the animal from behind. After fast processing of the image and extraction of its BCS, this score for this trap is stored in the relevant database.

[0014] To solve the problem of checking the current status of the estimated BCS in the proposed system, the output results are regularly sent to the expert or livestock mobile application.

[0015] Advantageous Effects of Invention

[0016] BCS estimation systems are desired to reduce the time and costs of the traditional BCS estimation techniques. Objective evaluation of BCS in a way that is easy and safe is the basic role of these systems.Brief Description of Drawings

[0017] Anatomical points of the cow

[0018] Fig 2: Block diagram of proposed method.

[0019] Fig 3: Capturing the image from the top view.

[0020] Fig 4: Illustrates the Jarvis algorithm.

[0021] Fig 5: The central point and extremities of the cow's body contourDescription of Embodiments

[0022] The use of automatic BCS estimation systems reduces time and cost compared to BCS estimation techniques. These systems provide objective evaluation of BCS in a simpler way for farmers. The proposed system increases the accuracy of the calculation due to its low cost, ease of implementation and use of advanced algorithms. In this method, first quality images are taken by the camera and then stored by a database. The pre-processed images and BCS estimation operations are then applied by machine learning algorithms. Recently, machine learning algorithms in the field of estimation and detail recognition in images have shown high performance, and also using estimating algorithms in this field can achieve high accuracy (Fig 1).

[0023] Mathematical morphology offers a form-based approach for digital images processing. Especially, the mathematical morphology operator abstracts image- related information. Computational morphology utilizes nonlinear algebra tools and acts with a series of points, their vicinity, and their forms. Morphological operations simplify and restrict the images, maintain the major features of the subject and discard others. Morphological operations are utilized for the following objectives (Fig 2):

[0024] • Image pre-processing (noise filtering, image simplification with intermittent filters)

[0025] • Uniform structure alteration (framing, thinning and thickening)

[0026] • Zonation and separation of the subject from the context

[0027] • A quantitative description of a subject

[0028] According to the assessment, capturing the image from the top view and at an appropriate angle (Fig 3), compared to the vertical view of cow's body, displays the body features, in particular, the protrusions around the hook and hip bones in a desirable manner and the imaging is carried out more precisely.

[0029] Some features including hull convex, average circle diameter, area, etc. are inferred from the cow image contour using various mathematical morphology operations. For instance, a hull convex is the smallest convex polygon that involves all the contour points. Hull convex is obtained by various methods, the most common of which are Jarvis and Graham algorithms. There is a simple idea behind such algorithms, such that the points are first sorted based on their y- coordinates, and in case y is equal, they are sorted by the x-coordinates and arranged in the P array. In this condition, the point p [0] is absolutely one of the hull convex points. So, this point is entered in an array so- called C. Now among the other points, a point is found that has the lowest polar angle to the point p [0] and is added to the array C. The same process is repeated for point p [1] to attain the last point of the array P (Fig 4).

[0030] Hull convex represents the convexity and concavity extent of the contour and can be utilized as a key feature. In the present research, hull convex and contour convexity and concavity points of contour were obtained using the aforementioned algorithms and the sum of the distances of these points to the central point was considered as an index. The sum of the distances from these points to the hull convex edge was also considered as a feature.

[0031] Such features can indicate the filling rate around the hook and pin bones and the tail area efficiently. Indeed, the lower the subcutaneous fat of the cow, the less filling rate these areas will be. Hence, there are more hull defects and concavity. The areas around the pelvis, pin, and pelvic joint that play a key role in bovine obesity and leanness, as well as the areas around the ligaments, are influential in the filling of defects, and the amount of this feature.

[0032] Other features, including extreme points, are extracted from the contours. The Fig 5 illustrates some of these points.

[0033] BCS estimation in the conventional approach is done by a veterinarian via examining only one side of the cow's body. But, according to this method, both sides of the cow's body are evaluated and the resultant score is average body scores on both sides of the cow's body and improves the measurement accuracy.

Claims

1) According to claim 1, using RFID tags and a side camera around the imaging system to capture side views and identify each animal separately.2) According to claim 1, incorporating an alert notification option in the application to notify the user of BCS (abnormal) alerts outside the normal range.3) to claim 1, adding an option to record long-term data in the application for drawing health trend charts of animals in the long term and analyzing the health or diseases that occurred for each animal separately.4) According to claim 1, including a prediction and estimation option in the application to assess the health trend of animals and predict the time of disease occurrence for each animal separately.5) to claim 1, adding a performance prediction option in the application to examine the level of livestock productivity and estimate the time of childbirth.

Images