JP2023129574A - Body weight estimation device, method for estimating body weight, and program - Google Patents

Abstract

To estimate body weight highly precisely.SOLUTION: A body estimation device includes: acquisition means for acquiring a first image of a target of which weight is an estimation target; classification means for classifying a posture of the estimation target in the first image acquired by the acquisition means into any one of a plurality of postures defined in advance; and estimation means for estimating the target weight from the first image by using an estimation model for the posture classified by the classification means. The postures are defined by a combination of presence or absence of a vertical bending and a sequence of heights of the neck part, the body part, and the hip part. The classification means classifies a posture into one of the postures on the basis of the first image according to the presence or absence of the vertical bending in a lateral direction of the estimation target and the sequence of the heights of the neck part, the body part, and the hip part.SELECTED DRAWING: Figure 5

Description

The present invention relates to a body weight estimation device, a body weight estimation method, and a program.

Livestock farmers have traditionally kept track of the weight of their livestock. This is because, for example, when livestock such as pigs exceed a certain weight, their value as meat may decrease.

By the way, the number of livestock raised by a livestock farm may exceed several hundred heads (or several hundred animals, several hundred birds, etc.). Therefore, when using a scale or the like, it takes time and effort to determine the weight of livestock. On the other hand, there is a known technique for photographing livestock with a video camera and estimating the weight from the area of the livestock in the photographed image (for example, see Patent Document 1).

Japanese Patent Application Publication No. 6-3181

However, for example, as the livestock moves, the posture of the livestock at the time of photographing varies, so with the above conventional technology, the error between the estimated weight and the true weight may become large and the estimation accuracy may decrease. Ta.

The embodiments of the present invention have been made in view of the above points, and an object thereof is to estimate body weight with high accuracy.

In order to achieve the above object, a body weight estimating device according to an embodiment of the present invention includes an acquisition means for acquiring a first image of a target for weight estimation; The first method uses a classification means for classifying which of a plurality of predefined orientations the estimation target pose is, and an estimation model corresponding to the orientation classified by the classification means. estimating means for estimating the weight of the estimation target from an image, and the plurality of postures include the presence or absence of bending in the left and right direction, and the ranking of the height of the neck, the height of the body, and the height of the buttocks. Based on the first image, the classification means determines whether or not the estimation target is bent in the left-right direction, the height of the neck, the height of the body, and the height of the buttocks. The method is characterized in that it classifies which of the plurality of postures it is based on the ranking of the postures.

Weight can be estimated with high accuracy.

1 is a diagram showing an example of the overall configuration of a body weight estimation system according to the present embodiment. FIG. 3 is a diagram for explaining an example of an aspect ratio. FIG. 3 is a diagram for explaining an example of symmetry. FIG. 3 is a diagram for explaining an example of an approximate curve. FIG. 3 is a diagram showing an example of a functional configuration of a weight estimation processing section according to the present embodiment. 7 is a flowchart illustrating an example of learning processing according to the present embodiment. FIG. 2 is a diagram (part 1) for explaining an example of preprocessing. FIG. 2 is a diagram (part 2) for explaining an example of preprocessing. FIG. 3 is a diagram for explaining an example of classification results. FIG. 3 is a diagram for explaining an example of a regression equation. It is a flow chart which shows an example of body weight estimation processing concerning this embodiment.

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention (hereinafter also referred to as "this embodiment") will be described below with reference to the drawings. In this embodiment, the pig's posture is classified into several positions, a regression equation is created in advance for estimating the weight for each of these postures, and then the pig's weight is estimated using these regression equations. explain. However, the object whose weight is to be estimated (hereinafter also referred to as "estimation object") is not limited to pigs, and may be various livestock such as cows, horses, sheep, and chickens, for example. Furthermore, the object of weight estimation may be not only livestock but also wild animals, for example.

<Overall configuration>
First, the overall configuration of a body weight estimation system 1 according to the present embodiment will be described with reference to FIG. 1. FIG. 1 is a diagram showing an example of the overall configuration of a body weight estimation system 1 according to the present embodiment.

As shown in FIG. 1, the body weight estimation system 1 according to the present embodiment includes a body weight estimation device 10 that estimates body weight, and a camera device 20 that photographs a pig P whose body weight is to be estimated. The weight estimating device 10 and the camera device 20 are communicably connected, for example, wirelessly, wired, or both.

The camera device 20 is a photographing device installed right above the photographing area A for photographing the pig P. The camera device 20 generates a photographed image by photographing the pig P in the photographing area A from directly above. Moreover, the camera device 20 is equipped with a depth sensor, and can measure the depth of each point of the pig P and obtain the height of each of these points. Note that the photographed image is represented by a point group of pixel values and depth values.

Note that the camera device 20 does not necessarily need to be installed directly above the photographing area A, but may be installed at a position where the pig P can be photographed from above. Further, the camera device 20 may generate a plurality of photographed images by, for example, continuously photographing one pig P a predetermined number of times (or continuously photographing for a predetermined time).

The weight estimating device 10 is a computer or computer system that estimates the weight of the pig P from the captured image generated by the camera device 20. When estimating the weight of the pig P, the weight estimating device 10 estimates the weight of the pig P using a regression equation corresponding to the posture of the pig P from among regression equations created in advance for each posture of the pig. . Note that the regression equation may be referred to as a regression model or the like, and is an example of an estimation model for estimating the weight of the estimation target.

Here, the weight estimation device 10 includes a weight estimation processing section 100 and a storage section 110. The weight estimation processing unit 100 is realized by a process that causes a CPU (Central Processing Unit) to execute one or more programs installed in the weight estimation device 10. Further, the storage unit 110 can be realized using, for example, an HDD (Hard Disk Drive) or an SSD (Solid State Drive).

The weight estimation device 10 performs a "learning process" in which the weight estimation processing unit 100 calculates a regression equation for estimating the weight of the pig P using learning data prepared in advance, and a regression process corresponding to the posture of the pig P. A "body weight estimation process" is executed to estimate the weight of the pig P using the formula. Furthermore, the storage unit 110 stores information and various processing results (for example, learning data, regression equations, weight estimation results, etc.) necessary for executing the learning process and the weight estimation process.

<Pig posture>
Here, the posture of the pig will be explained. In this embodiment, the pig's posture is determined by a combination of the presence or absence of bending in the lateral direction (that is, the horizontal direction of the pig) and the height order of each predetermined point in the pig's neck, body, and rump. shall be defined as In this embodiment, for example, the presence or absence of bending in the lateral direction is expressed as "present" (bent) or "absent" (not bent), and the height of each predetermined point in each of the neck, body, and buttocks is determined. The height (that is, the z coordinate value) is set as z _a , z _b , and z _c , and the height is ranked as "1,""2," and "3" in descending order of height (that is, in descending order of z coordinate value). shall be taken as a thing. Note that the height of each of these points can be obtained, for example, by measuring the depth with a depth sensor included in the camera device 20.

In this embodiment, the pig's posture is determined by the following 12 patterns of postures Q ₁ to 1, using the presence or absence of lateral bending and the ranking of the heights of predetermined points in each of the neck, body, and buttocks. Q: It is classified as ₁₂ . Note that hereinafter, for convenience, the presence or absence of bending in the lateral direction will also be simply referred to as "presence or absence of bending." In addition, hereinafter, the ranking of the height z _a of a predetermined point in the neck area will be referred to as the "neck ranking", the ranking of the height z _b of the predetermined point in the body part will be referred to as the "body ranking", and the ranking of the height z b of the predetermined point in the buttocks area will be referred to as the "neck ranking". The rank of height _zc is also expressed as "hip rank".

・Posture Q ₁ : Curvedness = None, and (head position, body position, buttocks position) = (1, 2, 3)
・Posture Q ₂ : Curvedness = None, and (head position, body position, buttocks position) = (1, 3, 2)
・Posture Q ₃ : Curvedness = None, and (head position, body position, buttocks position) = (2, 1, 3)
・Posture Q ₄ : Curvedness = None, and (head position, body position, buttocks position) = (2, 3, 1)
・Posture Q ₅ : Curvedness = None, and (head position, body position, buttocks position) = (3, 1, 2)
・Posture Q ₆ : Curvedness = None, and (head position, body position, buttocks position) = (3, 2, 1)
・Posture _Q7 : Curvedness = Yes, and (head position, body position, buttocks position) = (1, 2, 3)
・Posture Q ₈ : Curvedness = Yes, and (head position, body position, buttocks position) = (1, 3, 2)
・Posture Q ₉ : Curvedness = Yes, and (head position, body position, buttocks position) = (2, 1, 3)
・Posture Q ₁₀ : Curvedness = Yes, and (head position, body position, buttocks position) = (2, 3, 1)
・Posture Q ₁₁ : Curvedness = Yes, and (head position, body position, buttocks position) = (3, 1, 2)
・Posture Q ₁₂ : Curvedness = Yes, and (head position, body position, buttocks position) = (3, 2, 1)
Here, the above-mentioned presence or absence of bending is determined using a value representing how much the pig is bent to the left or right (this is referred to as the "degree of bending"). If the degree of curvature is small (for example, less than a predetermined threshold value), it is "absent". Such degree of curvature may be defined, for example, as in any one of (1) to (4) below.

(1) Definition using aspect ratio As shown in Fig. 2, in a photographed image of a pig taken with camera device 20, the aspect ratio is determined as follows using the vertical length and horizontal length of the circumscribed rectangle of the pig part. This aspect ratio is defined as the degree of curvature.

Aspect ratio = horizontal length / vertical length At this time, the closer the aspect ratio is to 1, the more the pig is bent to the left or right. Therefore, for example, using a certain threshold th ₁ (0<th ₁ <1), if the aspect ratio (degree of bending) is less than th ₁ , then the presence or absence of bending = none, and if the aspect ratio is th ₁ or more, then It is sufficient if the presence or absence of bending = presence.

(2) Definition using symmetry As shown in FIG. 3(a), an xyz coordinate system is set for a photographed image of a pig with the camera device 20. Further, in the photographed image, a circumscribed rectangle of the pig portion is set, and then a dividing line is set to horizontally divide this circumscribed rectangle into two equal parts. Here, for convenience, it is assumed that the dividing line is parallel to y=0. Note that if the circumscribed rectangle is tilted with respect to the photographed image, the circumscribed rectangle is rotated so that the tilt is eliminated.

At this time, for each x-coordinate value, the difference between the number of points whose y-coordinate value is larger than the dividing line and the number of points whose y-coordinate value is smaller than the dividing line is calculated. In other words, for each x-coordinate value, the absolute value of the difference between the y-coordinate value (upper end value) of the upper end of the pig part and the y-coordinate value (lower end value) of the lower end of the pig part is calculated. The results of this calculation are shown in FIG. 3(b). FIG. 3B shows the difference (absolute value of the difference) between the upper end value and the lower end value for each x-coordinate value. Then, the sum of these differences (total difference) is defined as the degree of curvature.

At this time, the larger the total difference, the smaller the symmetry of the pig, indicating that it is curved to the left or right. Therefore, using a certain threshold value th ₂ , if the total difference (degree of bending) is less than th ₂ , the presence or absence of bending may be determined to be absent, and if the total difference is greater than or equal to th ₂ , the presence or absence of bending may be determined to be present.

(3) Definition using approximate curves As shown in FIG. 4, an xyz coordinate system is set for a photographed image of a pig with the camera device 20. At this time, for each x-coordinate value, identify the midpoint between the y-coordinate value (upper end value) of the upper end of the pig part and the y-coordinate value (lower end value) of the lower end of the pig part, and set these midpoints. The curve passing through is the center curve. For example, a quadratic curve (approximate curve) that approximates the central curve is found, and the quadratic coefficient of the approximate curve (that is, coefficient a when the approximate curve is expressed as y=ax ² +bx+c) is determined as the degree of curvature. do.

At this time, the larger the absolute value of the second-order coefficient, the more leftward or rightward the curve is. Therefore, using a certain threshold th ₃ , if the absolute value of the second-order coefficient is less than th ₃ , the presence or absence of bending = no, and if the absolute value of the second-order coefficient is th ₃ or more, the presence or absence of bending = yes. It's fine if you do this.

(4) Definition based on a combination of two or more of the above (1) to (3). For example, the aspect ratio of (1) above, the total difference of (2) above, and the quadratic of (3) above. The sum or weighted sum of any two or more of the absolute values of the coefficients is defined as the degree of curvature.

<Functional configuration>
Next, the functional configuration of the weight estimation processing section 100 according to this embodiment will be described with reference to FIG. 5. FIG. 5 is a diagram showing an example of the functional configuration of the weight estimation processing section 100 according to the present embodiment.

As shown in FIG. 5, the weight estimation processing unit 100 according to the present embodiment includes an acquisition unit 101, a preprocessing unit 102, a posture class classification unit 103, a regression equation calculation unit 104, and a weight estimation unit 105. is included.

In the learning process, the acquisition unit 101 acquires (reads) a plurality of captured images (a group of captured images) stored in the storage unit 110 as learning data. Here, the photographed image acquired as learning data is an image of a pig whose weight is at least known. Note that in the photographed image acquired as learning data, it may be further known whether the posture of the pig is among postures Q ₁ to Q ₁₂ , for example.

Further, in the weight estimation process, the acquisition unit 101 acquires a captured image of the pig P to be estimated from the storage unit 110 or the camera device 20.

The preprocessing unit 102 performs predetermined preprocessing on the captured image acquired by the acquisition unit 101 in the learning process and the weight estimation process. Note that the preprocessing is a process for improving the accuracy of estimating the pig's weight, and does not necessarily have to be performed.

In the learning process, the posture class classification section 103 classifies a group of photographed images (a group of photographed images acquired by the acquisition section 101 or a group of photographed images after preprocessing by the preprocessing section 102) into a plurality of classes (hereinafter referred to as "posture classes"). ) (in this embodiment, classified into 12 posture classes). Hereinafter, these 12 posture classes are also expressed as C ₁ to C ₁₂ , i=1, . . . , 12, and posture Q _i corresponds to posture class C _i .

In addition, in the body weight estimation process, the posture class classification section 103 classifies the photographed image (the photographed image acquired by the acquisition section 101 or the photographed image after preprocessing by the preprocessing section 102) into one of posture classes C ₁ to C ₁₂ . Classify into.

In the learning process, the regression equation calculation unit 104 calculates a regression equation for each of posture classes C ₁ to C ₁₂ (that is, each of postures Q ₁ to Q ₁₂ ). These regression equations are calculated by performing regression analysis using captured images classified into posture classes C ₁ to C ₁₂ , respectively.

In the weight estimation process, the weight estimation unit 105 uses a regression equation corresponding to the posture class into which the captured image (the captured image acquired by the acquisition unit 101 or the captured image after preprocessing by the preprocessing unit 102) is classified. , calculate the weight of the pig P to be estimated. Thereby, the weight of the pig P to be estimated is estimated.

<Learning process>
Next, a learning process for calculating a regression equation for each posture using learning data will be described with reference to FIG. FIG. 6 is a flowchart showing an example of the learning process according to this embodiment. In the following, it is assumed that a plurality of photographed images (a group of photographed images) of pigs whose weights are at least known are stored in the storage unit 110. Note that for these plurality of captured images (captured image group), for example, it may be known which of the postures Q ₁ to Q ₁₂ the pig is in.

Step S101: The acquisition unit 101 acquires a group of captured images from the storage unit 110 as learning data.

Note that the storage unit 110 may store a group of photographed images that have been previously subjected to preprocessing described in step S102, which will be described later, and the acquisition unit 101 may acquire the group of photographed images after the preprocessing. In this case, it is not necessary to perform the process of step S102, which will be described later.

Step S102: Next, the preprocessing unit 102 performs predetermined preprocessing on each of the captured images included in the captured image group. Here, the preprocessing includes, for example, "pig extraction processing" and "head and tail exclusion processing." Note that as preprocessing, both "pig extraction processing" and "head and tail exclusion processing" may be performed, or only one of them may be performed. Further, the pretreatment itself may not be performed. If the preprocessing itself is not performed, the process of step S102 is not performed.

(Pig extraction process)
Pig extraction processing involves identifying the boundaries of the pig, for example, in cases where the pig is in close contact with a wall or the pig is in close contact with another pig in the photographed image. This is a process that extracts only the image area of the pig part.

As shown in FIG. 7(a), an xyz coordinate system is set for the photographed image. Then, at each x-coordinate value of the photographed image, a z-coordinate value at each y-coordinate value is obtained. The example shown in FIG. 7A shows a case where the z coordinate value at each y coordinate value where y≧0 is acquired at x=940. A graph (cross-sectional graph) of the z-coordinate values obtained in this way is shown in FIG. 7(b).

At this time, a point where the z-coordinate value is locally small is defined as a point on the boundary. In the example shown in FIG. 7(b), y=680 is a point on the boundary. By performing this for each x-coordinate value, the boundary is identified. Then, the image area within the boundary is extracted as the image area of the pig part.

Note that the reason why a point with a locally small z-coordinate value is defined as a point on the boundary is that there are generally two points on the boundary for one x-coordinate value. For this reason, instead of a point where the z coordinate value is locally small, a point with the smallest z coordinate value and a point with the next smallest z coordinate value may be used as points on the boundary.

(Exclusion processing of head and tail parts)
The head and tail parts may move during shooting, which often affects the accuracy of weight estimation. Therefore, by excluding (deleting) the image regions of the head and tail, it is possible to improve the accuracy of weight estimation. Although it is possible to improve the accuracy of body weight estimation by excluding only the image area of either the head or the tail, it is better to exclude both the head and tail image areas. It is possible to further improve the accuracy of weight estimation.

First, the boundaries of the pig in the captured image are identified using a method similar to the pig extraction process described above. Then, the head portion and tail portion are excluded (deleted) as follows.

That is, for example, by using line segments connecting adjacent points on the boundary, we can identify points where the angle between the adjacent line segments is less than or equal to a predetermined threshold (in other words, the recesses of the recesses on the boundary that are less than or equal to the predetermined angle). Identify it as the base of the. The head is then excluded by connecting the base of the neck with a straight line to form a boundary.

Further, for example, an image area having a width of 1/3 to 1/5 or less of the width (ie, the width in the y-axis direction) of the pig part (the image area of the whole pig) is excluded as the tail part. Note that since the width of the pig part varies depending on the x-coordinate value, it may be set to the width of the pig part at a predetermined x-coordinate value (for example, x=0, etc.), for example.

In the following, the preprocessing unit 102 performs the above preprocessing (“pig extraction processing” and “head and tail exclusion processing”) on each captured image included in the captured image group. I will explain as follows.

Step S103: The posture class classification unit 103 classifies each captured image included in the group of captured images after preprocessing by the preprocessing unit 102 into a posture class. As a method for classifying into posture classes, for example, the k-means method may be used. Alternatively, for example, SVM (Support Vector Machine) or the like may be used. Note that at this time, the number of posture classes is 12 for classification.

Here, the posture class classification unit 103 extracts an arbitrary feature amount (hereinafter referred to as "first feature amount") from each captured image, and uses this first feature amount to Classify the captured image into one of the posture classes.

As an example, the classification results are shown in FIG. 9 where X is the degree of curvature, Y is z _b −z _c , Z is z _b −z _a , and the first feature amount is (X, Y, Z). In this case, each photographed image is classified within the three-dimensional space of XYZ. Note that the data points in FIG. 9 represent each photographed image included in the photographed image group.

As a result, each photographed image included in the photographed image group is classified into one of the 12 posture classes. Here, the user may determine which of the postures Q ₁ to Q ₁₂ each of the 12 posture classes corresponds to, for example, from the posture of the pig in each photographed image belonging to the posture class. , if a label indicating one of postures Q ₁ to Q ₁₂ is assigned to each photographed image included in the photographed image group, the determination is made from the distribution of labels assigned to each photographed image belonging to the posture class. It's okay. Hereinafter, as described above, it is assumed that i=1, . . . , 12, and posture Q _i corresponds to posture class C _i .

Step S104: Next, the regression formula calculation unit 104 calculates a regression formula for each of the posture classes C ₁ to C ₁₂ (that is, each of postures Q ₁ to Q ₁₂ ). Here, the regression equation calculation unit 104 extracts an arbitrary feature amount (hereinafter referred to as "second feature amount") from each photographed image, and calculates the second feature amount using a regression analysis method. A regression equation expressing the relationship between volume and body weight is calculated. Note that, as described above, the weight corresponding to each photographed image is known.

The second feature amount may be, for example, the projected area of the pig part in the captured image, or the surface area of the pig part in the captured image. Note that if the head portion and/or tail portion is excluded in the preprocessing, the pig portion refers to the image area after the exclusion.

As a result, as shown in FIG. 10, i=1, . 12, a regression equation f _i corresponding to the posture Q _i (that is, the posture class C _i ) is calculated. Then, the regression f _i (i=1, . . . 12) calculated by the regression formula calculation unit 104 is stored in the storage unit 110. Note that FIG. 10 illustrates a regression curve represented by the regression equation _fi .

<Weight estimation process>
Next, a weight estimation process for estimating the weight of the pig P to be estimated will be described with reference to FIG. 11. FIG. 11 is a flowchart illustrating an example of the weight estimation process according to this embodiment.

Step S201: The acquisition unit 101 acquires a captured image of the pig P to be estimated from the storage unit 110 or the camera device 20.

Step S202: Next, the preprocessing unit 102 performs preprocessing on the captured image, similar to step S102 in FIG. Note that if preprocessing is performed during learning processing (including when a captured image after preprocessing is acquired as learning data), preprocessing is also performed on the captured image, but preprocessing is not performed during learning processing. If the preprocessing is not performed, no preprocessing is performed on the photographed image.

Step S203: Next, the posture class classification unit 103 classifies the captured image into posture classes using the same method as in step S103 of FIG. At this time, the posture class classification unit 103 extracts a first feature amount from the captured image and classifies the captured image into one of the posture classes using the first feature amount. As a result, the captured image is classified into one of the posture classes C ₁ to C ₁₂ .

Step S204: Next, the weight estimating unit 105 calculates the weight of the pig P using the regression equation corresponding to the posture class classified in step S203 above. That is, the weight estimating unit 105 extracts the second feature amount from the captured image, and calculates the weight of the pig P using the second feature amount using the regression equation. Thereby, the weight of pig P is estimated.

Note that the weight estimated by the weight estimating unit 105 may be stored in the storage unit 110, for example, or may be output to a display device such as a display. In addition to this, the information may be output to, for example, another device connected to the weight estimating device 10 via a network.

<Summary>
As described above, the body weight estimation system 1 according to the present embodiment creates a regression equation (estimation model) for each posture of the pig using learning data prepared in advance. When estimating the weight of a pig, the body weight estimation system 1 according to the present embodiment estimates the weight using an estimation model corresponding to the posture of the pig to be estimated. Thereby, even if the posture of the pig to be estimated varies, the weight of the pig can be estimated with high accuracy according to the posture.

Note that in this embodiment, the postures of the pigs are classified into 12 postures, but this is just an example, and the classification method and number of classifications may be different. In particular, the classification method and number of classifications may differ depending on the type of pig (e.g. Yorkshire, Landrace, etc.), or the type of estimation target (e.g. pig, cow, horse, sheep, chicken, etc.) ) The method of classification and the number of classifications may be different depending on the type of classification.

The present invention is not limited to the above-described specifically disclosed embodiments, and various modifications and changes can be made without departing from the scope of the claims.

1 Weight Estimation System 10 Weight Estimation Device 20 Camera Device 100 Weight Estimation Processing Unit 101 Acquisition Unit 102 Preprocessing Unit 103 Posture Class Classification Unit 104 Regression Equation Calculation Unit 105 Weight Estimation Unit 110 Storage Unit

Claims (4)

Acquisition means for acquiring a first image of a subject for weight estimation;
a classification means for classifying which of a plurality of predefined postures the posture of the estimation target in the first image acquired by the acquisition means is;
Estimating means for estimating the weight of the estimation target from the first image using an estimation model corresponding to the posture classified by the classifying means;
has
The plurality of postures are postures defined for each combination of the presence or absence of bending in the left and right direction, and the ranking of the height of the neck part, the height of the body part, and the height of the buttocks part,
The classification means is
Based on the first image, which of the plurality of postures is determined based on the presence or absence of bending of the estimation target in the left and right direction, and the ranking of the height of the neck, the height of the body, and the height of the buttocks. A weight estimation device characterized by classifying whether there is a weight or not. comprising a model creation means for creating an estimation model corresponding to the orientation for each orientation;
The acquisition means is
Obtain a second set of images of objects whose weight is known as learning data,
The classification means is
classifying which of the plurality of postures the target posture is in each second image included in the learning data acquired by the acquisition means;
The model creation means includes:
A claim characterized in that, for each posture, an estimation model corresponding to the posture is created using the second image classified as the posture and the weight of the subject in the second image. The weight estimation device according to item 1. an acquisition procedure of acquiring a first image of a subject for weight estimation;
a classification procedure for classifying which of a plurality of predefined postures the estimation target posture in the first image acquired in the acquisition procedure is;
an estimation step of estimating the weight of the estimation target from the first image using an estimation model corresponding to the posture classified in the classification step;
The computer executes
The plurality of postures are postures defined for each combination of the presence or absence of bending in the left and right direction, and the ranking of the height of the neck part, the height of the body part, and the height of the buttocks part,
The classification procedure is
Based on the first image, which of the plurality of postures is determined based on the presence or absence of bending of the estimation target in the left and right direction, and the ranking of the height of the neck, the height of the body, and the height of the buttocks. A method for estimating body weight, which is characterized by classifying whether it is present or not. A program for causing a computer to function as the weight estimating device according to claim 1 or 2.

Images