JP6818283B1 - Counting system, counting device, counting method and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To measure an accurate number of objects. SOLUTION: An image pickup unit 110 installed at a position where an individual moving in a flow path can be imaged, and a detection unit 120 for detecting an individual from an image captured by the image pickup unit 110 by using a learning model 130. It has a counting unit 140 that counts the number of individuals detected by the unit 120, and an output unit 150 that outputs the number counted by the counting unit 140. [Selection diagram] Fig. 2

Description

The present invention relates to counting systems, counting devices, counting methods and programs.

It goes without saying that it is essential to count the number of fish traded in fish commerce. For example, an apparatus is considered in which an object and a liquid are made to flow in an inclined flow path, an image thereof is taken from a predetermined angle, and the object is measured by analyzing the taken image (see, for example, Patent Document 1). .).

Japanese Patent No. 5561838

The apparatus described in Patent Document 1 detects an object by performing a binarization process on a captured image. Therefore, there is a problem that it is difficult to measure an accurate number of objects.

An object of the present invention is to provide a counting system, a counting device, a counting method and a program capable of measuring an accurate number of objects.

The counting system of the present invention
The flow path in which the individual moves and
An imaging unit installed at a position where the individual moving in the flow path can be imaged,
A detection unit that detects the individual from the image captured by the imaging unit using a deep learning model,
A counting unit that counts the number of the individuals detected by the detection unit,
It has an output unit that outputs the number counted by the counting unit.

Further, the counting device of the present invention is
An imaging unit installed at a position where an individual moving in the flow path can be imaged,
A detection unit that detects the individual from the image captured by the imaging unit using a deep learning model,
A counting unit that counts the number of the individuals detected by the detection unit,
It has an output unit that outputs the number counted by the counting unit.

Moreover, the counting method of the present invention
The process of imaging the flow path in which an individual moves, and
A process of detecting the individual from the captured image using a deep learning model, and
The counting process for counting the number of the detected individuals and
The process of outputting the counted number is performed.

In addition, the program of the present invention
A program that lets a computer run
The procedure for imaging the flow path in which an individual moves, and
A detection procedure for detecting the individual from the captured image using a deep learning model, and
A counting procedure for counting the number of the detected individuals, and
The output procedure for outputting the counted number is executed.

In the present invention, an accurate number of objects can be measured.

It is a figure which shows the 1st Embodiment of the counting system of this invention. It is a figure which shows an example of the internal structure of the counting apparatus shown in FIG. It is a figure which shows an example of the appearance of the counting system shown in FIG. It is a flowchart for demonstrating an example of the counting method in the counting system shown in FIG. It is a figure which shows the 2nd Embodiment of the counting system of this invention. It is a figure which shows an example of the internal structure of the counting apparatus shown in FIG. It is a figure which shows an example of the appearance of the flow path shown in FIG. It is a figure which shows an example of the appearance of the counting system shown in FIG. It is a figure which shows an example of the frame which surrounds the fish detected by the detection part shown in FIG. It is a figure for demonstrating the counting process performed by the counting part shown in FIG. It is a flowchart for demonstrating an example of the counting method in the counting apparatus shown in FIG.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
(First Embodiment)

FIG. 1 is a diagram showing a first embodiment of the counting system of the present invention. As shown in FIG. 1, the counting system in the present embodiment has a counting device 100 and a flow path 200. The flow path 200 is a flow path through which the target individual moves. The counting device 100 counts the number of individuals moving in the flow path 200.

FIG. 2 is a diagram showing an example of the internal configuration of the counting device 100 shown in FIG. As shown in FIG. 2, the counting device 100 shown in FIG. 1 includes an imaging unit 110, a detecting unit 120, a learning model 130, a counting unit 140, and an output unit 150. Note that FIG. 2 shows only the main components related to the present embodiment among the components included in the counting device 100 shown in FIG.

The imaging unit 110 is an imaging means such as a camera that images the flow path 200. Specifically, the imaging unit 110 images the surface on the flow path 200 where the target individual moves. The imaging unit 110 images an individual moving in the flow path 200 at a frame rate capable of imaging a sufficient number of times, for example, at a frame rate capable of imaging the individual five times or more while the individual passes through the imaging range. The detection unit 120 detects an individual by using the image captured by the imaging unit 110 and the learning model 130 in which the characteristics of a large number of individuals are deep-learned in advance. That is, the detection unit 120 recognizes an individual by using the image captured by the imaging unit 110 and the deep learning model that has learned the characteristics of a large number of individuals. The counting unit 140 counts the number of individuals detected by the detection unit 120. The output unit 150 outputs the number counted by the counting unit 140.

FIG. 3 is a diagram showing an example of the appearance of the counting system shown in FIG. In the counting system shown in FIG. 1, as shown in FIG. 3, the counting device 100 is installed so that the imaging unit 110 is located at a position where a part of the flow path 200 can be imaged. The output unit 150 may be provided on the outside of the housing of the counting device 100, and the counting unit 140 and the output unit 150 may be connected wirelessly or by wire. Further, the imaging unit 110 and other components may be housed in separate housings. In this case, the imaging unit 110 and the detection unit 120 are connected wirelessly or by wire. The housing is dustproof and waterproof. Further, the counting device 100 and the flow path 200 are provided on the ship, for example, so that the fish can be counted on the ship. The ship used for this counting work is not particularly limited as long as it can perform counting work on board, but for example, it is preferable to use a work ship.

The counting method in the counting system shown in FIG. 1 will be described below. FIG. 4 is a flowchart for explaining an example of a counting method in the counting system shown in FIG.

First, in order to start measurement, when the counting device 100 receives an input for starting measurement based on a predetermined operation from the outside, the imaging unit 110 starts imaging the flow path 200 (step S1). Subsequently, the detection unit 120 detects an individual by using the image captured by the imaging unit 110 and the learning model 130 in which the characteristics of a large number of individuals are deep-learned in advance (step S2). When the detection unit 120 detects an individual, the counting unit 140 counts up the counter value (step S3). After that, the counting device 100 determines whether or not the measurement is completed (step S4). Here, the determination of step S4 is performed by determining whether or not the counting device 100 has received the input of the end of measurement based on a predetermined operation from the outside. When the counting device 100 determines that the measurement is completed, the output unit 150 outputs the number of individuals measured by the counting unit 140 (step S5). On the other hand, if the counting device 100 determines that the measurement is not completed, the process of step S2 is performed.

In this way, the counting device 100 detects a target individual from the captured image of the flow path 200 using the deep learning model, counts the number of the detected individuals, and outputs the number. Therefore, the exact number of objects can be measured.
(Second Embodiment)

FIG. 5 is a diagram showing a second embodiment of the counting system of the present invention. As shown in FIG. 5, the counting system in the present embodiment has a counting device 101 and a flow path 201. In this embodiment, a case where the target individual is a fish (hereinafter referred to as a fish) will be described as an example. Further, the fish are farmed fish or natural fish, and the fish species belong to the genus Amberjack, for example. The size of the fish is not specified.

FIG. 6 is a diagram showing an example of the internal configuration of the counting device 101 shown in FIG. As shown in FIG. 6, the counting device 101 shown in FIG. 5 includes an imaging unit 110, a detection unit 121, a learning model 131, a counting unit 141, an output unit 150, a tracking unit 161 and a distance measuring unit 171. And have. Note that FIG. 6 shows only the main components related to the present embodiment among the components included in the counting device 101 shown in FIG. The counting device 101 is dustproof and waterproof. The imaging unit 110 and the output unit 150 are the same as those in the first embodiment.

The detection unit 121 has the following functions added to the functions of the detection unit 120 in the first embodiment. The detection unit 121 detects a fish by using the image captured by the imaging unit 110 and the learning model 131 in which the characteristics of a large number of individuals excluding the caudal fin of the fish are deep-learned in advance. At this time, the detection unit 121 may use a deep learning compression technique to which quantization / pruning of deep learning is applied. The detection unit 121 may detect a fish by using an image of a portion excluding a fin portion (for example, a pectoral fin, etc.) other than the caudal fin. In this way, by creating a deep learning model using the fins with large changes excluded from the annotation targets as teacher data, it is possible to avoid complication of the learning model and perform more accurate detection. .. The fin portion having a large change is a portion having a large range of motion and the imaged shape changes according to the timing of imaging. Further, when the detection unit 121 detects a fish, the detected fish may be surrounded by a frame called a Bounding Box, and the surrounded frame may be used as a detection result. By displaying the detection result using the frame in this way, it is possible for the operator to easily recognize the result of detecting the fish.

The tracking unit 161 tracks the individual detected by the detection unit 121. The tracking unit 161 performs a so-called tracking process, and the tracking process may be any tracking process as long as it can track the detected individual, and is not particularly limited. The distance measuring unit 171 measures the moving distance of the individual tracked by the tracking unit 161. Among the individuals tracked by the tracking unit 161, the counting unit 141 is an individual whose moving distance measured by the distance measuring unit 171 exceeds a predetermined distance threshold value and the time during which the individual is in the imaging range exceeds a predetermined time. Is counted as the number of fish. For example, the counting unit 141 is imaged in a certain number of frames or more of the frames imaged by the imaging unit 110 among the individuals tracked by the tracking unit 161 and is 3 minutes of the imaging range (inside the frame) of the imaging unit 110. Individuals that have moved a distance of 1 or more are counted. Among the individuals tracked by the tracking unit 161, the counting unit 141 captures images of individuals whose movement distance measured by the distance measurement unit 171 does not exceed a predetermined distance threshold value or even if the movement distance exceeds a predetermined distance threshold value. If the time existing within the range is within a predetermined time, the individual may be excluded from the counting target. The distance threshold value to be compared with this moving distance and a predetermined time can be set from the outside. Further, the counting unit 141 may be capable of resetting the count value for each measurement.

FIG. 7 is a diagram showing an example of the appearance of the flow path 201 shown in FIG. As shown in FIG. 7, the flow path 201 shown in FIG. 5 is a flow path composed of a bottom surface 211 and two side walls 221,231. The surface of the bottom surface 211 is preferably, for example, embossed so that fish can easily move, but the surface thereof is not limited to this. The material inside the flow path 201 is preferably stainless steel or fiber reinforced plastic (FRP: Fiber-Reinforced Plastics) that does not easily rust. The width of the flow path 201, that is, the distance between the side wall 221 and the side wall 231 is a size that allows the fish to move while touching the bottom surface 211, and is preferably about 1 to 100 cm. The distance between the side wall 221 and the side wall 231 is preferably equal to or greater than, for example, the body width or height of the target fish. The height of the side wall 221 and the side wall 231 is preferably about 1 to 100 cm. The heights of the side wall 221 and the side wall 231 are preferably equal to or greater than the body width of the target fish, for example, but the height at which the fish moving in the flow path 201 does not deviate from the flow path 201 (does not jump out). If it is good.

FIG. 8 is a diagram showing an example of the appearance of the counting system shown in FIG. In the counting system shown in FIG. 5, as shown in FIG. 8, the counting device 101 is installed so that the imaging unit 110 is located at a position where an individual moving in the flow path 201 can be imaged. The flow path 201 is provided on the ship, not in water. The output unit 150 may be provided on the outside of the housing of the counting device 101, and the counting unit 141 and the output unit 150 may be connected wirelessly or by wire. Further, the imaging unit 110 and other components may be housed in separate housings. In this case, the imaging unit 110 and the detection unit 121 are connected wirelessly or by wire. The housing is dustproof and waterproof. Further, the counting device 101 and the flow path 201 are provided on the ship, for example, so that the fish can be counted on the ship. Further, the counting device 101 is provided with a battery (for example, a rechargeable storage battery or the like) inside, and each component provided in the counting device 101 is supplied with power from the battery to operate. The fish are released from the cage upstream of the flow path 201 using a large net. The flow path 201 may be divided into a plurality of small flow paths. In that case, each of the plurality of small channels is formed to have different sizes (widths) from each other so that the fish released on the channel 201 can be selected as being counted or not being counted according to the size. You may.

As shown in FIG. 8, the flow path 201 is arranged at an angle with respect to the horizontal plane 301 so that the fish slide and move from the upstream to the downstream along the bottom surface of the flow path 201. When there are a plurality of fish moving in the flow path 201, it is possible that a plurality of fish overlap each other on the upstream side of the flow path 201. When a plurality of fish overlap each other, it becomes difficult for the detection unit 121 to detect the fish existing under the overlap from the image captured by the imaging unit 110. Therefore, the flow path 201 is inclined with respect to the horizontal plane to prevent overlap. The inclination angle is preferably 4 to 6 degrees, but the amount of fish, the material, processing, and structure of the bottom surface, the amount of water flowing from upstream to downstream together with the fish to make the fish slippery on the flow path 201, etc. Depending on the situation, it may be appropriately set (for example, set within the range of about 1 to 30 degrees).

Further, since the flow path 201 is inclined with respect to the horizontal plane as described above, the possibility that a plurality of fish overlap each other is lower in the downstream side than in the upstream side of the flow path 201 where the fish start to move. Therefore, the imaging unit 110 is installed so as to image the flow path 201 on the downstream side as much as possible. For example, the imaging unit 110 is installed so as to image the flow path 201 on the downstream side of the midpoint between the upstream and downstream of the flow path 201.

Further, a hydrophilic sheet (hydrophilic coating) 401 is provided between the imaging unit 110 and the flow path 201. The hydrophilic sheet 401 may be a member (hydrophilic member) such as a transparent plate or sheet that has been subjected to a hydrophilic treatment, and may be stretched so as to cover the lens of the imaging unit 110. By providing the hydrophilic sheet 401, it is possible to prevent water droplets and water droplets from interfering with image imaging. Further, at least one of a polarizing filter and a dimming filter may be provided as a filter 402 between the imaging unit 110 and the hydrophilic sheet 401. By providing the filter 402, it is possible to avoid blurring of the image captured by the imaging unit 110 due to the influence of strong light.

FIG. 9 is a diagram showing an example of a frame surrounding the fish detected by the detection unit 121 shown in FIG. As shown in FIG. 9, the detection unit 121 performs a process of surrounding the fish 501 excluding the caudal fin 511 with the detection frame 521. The detection frame 521 is the Bounding Box described above. The detection frame 521 is displayed when the output unit 150 displays the fish 501 imaged by the imaging unit 110. By surrounding the fish 501 with the detection frame 521, when the output unit 150 displays the image captured by the imaging unit 110, the user who sees the image displayed by the output unit 150 identifies the fish 501 from the image. It will be easier to do.

FIG. 10 is a diagram for explaining the counting process performed by the counting unit 141 shown in FIG. As shown in FIG. 10, when the object detected by the detection unit 121 as a fish moves on the flow path 201 on the flow path 201 beyond a predetermined distance threshold value, the counting unit 141 determines that the individual is a fish to be counted. Count. On the other hand, when the movement of the individual detected by the detection unit 121 as a fish is equal to or less than the distance threshold value, the counting unit 141 does not count the individual as a non-counting target.

The counting method in the counting device 101 shown in FIG. 6 will be described below. FIG. 11 is a flowchart for explaining an example of a counting method in the counting device 101 shown in FIG.

First, in order to start the measurement, when the counting device 101 receives the input for starting the measurement based on a predetermined operation from the outside, the imaging unit 110 starts the imaging of the flow path 201 (step S11). Subsequently, the detection unit 121 detects an individual by using the image captured by the imaging unit 110 and the learning model 131 in which the characteristics of a large number of individuals are deep-learned in advance (step S12). At this time, as described above, the detection unit 121 may perform detection except for the fin portion having a large change. When the detection unit 121 detects an individual, the tracking unit 161 tracks the individual detected by the detection unit 121 (step S13). Subsequently, the distance measuring unit 171 measures the moving distance of the individual tracked by the tracking unit 161 (step S14).

Then, the counting unit 141 determines whether or not the movement distance measured by the distance measuring unit 171 exceeds the preset distance threshold value for the individual tracked by the tracking unit 161 (step S15). When the moving distance measured by the distance measuring unit 171 exceeds a preset distance threshold value for the individual tracked by the tracking unit 161, the counting unit 141 measures the time during which the individual is within the imaging range (step). S16). The counting unit 141 determines whether or not the time that the individual exists in the imaging range exceeds a preset time (step S17). When the time that the individual exists in the imaging range exceeds a preset time, the counting unit 141 counts up the counter value (step S18). On the other hand, the counting unit 141 determines that the movement distance measured by the distance measurement unit 171 does not exceed a preset distance threshold value or the movement distance exceeds a predetermined distance threshold value for the individual tracked by the tracking unit 161. If the time the individual is in the imaging range is within a predetermined time, the counter value is not counted up.

Here, the detection unit 121 erroneously detects that the noise that is not the target individual, such as the mere reflection of light in the flow path 201, the dirt on the flow path 201, the pattern, and a part of the worker, is an individual (fish). In this case, in order to prevent it from being counted as a fish, the counting unit 141 determines whether or not the moving distance of the individual detected by the detecting unit 121 and the time that the individual exists in the imaging range are fish. Used for (counting). If the individual detected by the detection unit 121 is merely a reflection of light in the flow path 201, the individual does not move. Alternatively, even if such light reflection or the like is erroneously detected, its existence time is often short. On the other hand, when the individual detected by the detection unit 121 is a fish, the individual (fish) moves on the flow path 201 over a certain period of time. The counting unit 141 uses the presence or absence of this movement and the time during which the individual is within the imaging range to determine whether or not the individual is a fish. Further, the distance threshold value used for determining the movement is a distance capable of determining whether the individual detected by the detection unit 121 is a moving fish or a reflection of light that does not move, and is a flow path. It is set in advance according to the shape and inclination angle of 201, the installation position, the speed and quantity of fish being put into the flow path 201, the time zone for measurement, and the like. Further, the preset time is a time for detecting that the fish has moved in the flow path 201, and is the shape and inclination angle of the flow path 201, the installation position, and the speed at which the fish is put into the flow path 201. It is set in advance according to the quantity and quantity.

After that, the detection unit 121 determines whether or not there are a plurality of individuals detected by the detection unit 121 (step S19). When the number of individuals detected by the detection unit 121 is not a plurality, the counting device 101 determines whether or not the measurement is completed (step S20). Here, the counting device 101 determines in step S20 by determining whether or not the input for the end of measurement has been received based on a predetermined operation from the outside. When the counting device 101 determines that the measurement is completed, the output unit 150 outputs the number of individuals measured by the counting unit 141 (step S21). The output unit 150 may display the number of individuals measured by the counting unit 141 on the display, or transmit information indicating the number of individuals measured by the counting unit 141 to another device. Is also good. The display for displaying the number of individuals may be a display provided in the counting device 101, a dedicated display connected to the counting device 101, or a display provided in the communication terminal. On the other hand, if the counting device 101 determines that the measurement is not completed, the process of step S12 is performed.

On the other hand, when it is determined in step S19 that there are a plurality of individuals detected by the detection unit 121, the tracking unit 161 determines whether or not there are unprocessed individuals (step S22). The tracking unit 161 distinguishes between individuals who have completed the processes of steps S13 to S18 and individuals who have not yet performed the processes of steps S13 to S18. For example, the tracking unit 161 may attach a marker that can be distinguished from each other to each individual detected by the detection unit 121, and change the marker so that it can be identified whether or not the marker has been processed. Further, the distance measuring unit 171 may perform the processing of the tracking unit 161. If there are untreated individuals, the treatments of steps S13 to S18 are performed on the untreated individuals. If there are no untreated individuals, the process of step S20 is performed. When the detection unit 121 detects a plurality of individuals in step S12, the processes of steps S13 to S178 may be performed in parallel for each of the detected individuals.

The output unit 150 may output the number of individuals measured by the counting unit 141 in real time, or the counting unit 141 outputs the number of individuals at a predetermined timing during measurement. Is also good. The predetermined timing in this case is, for example, the timing of each time the counting unit 141 counts an individual, or the timing of counting a predetermined number (for example, 10), or a unit time preset from the start of measurement (for example, 10). It may be the timing for each passage (1 minute).

The learning model 131 shown in FIG. 6 may be stored in a transferable recording medium such as a memory card and mounted on the counting device 101 for use. In that case, if different learning models are stored in each of the plurality of memory cards, the counting device 101 can immediately execute the learning model suitable for the individual to be measured by simply replacing the memory card to be attached. it can. Alternatively, a plurality of learning models different from each other may be stored in one memory card, and the learning model to be used may be switched and executed according to the measurement target.

In this way, the counting device 101 detects the target individual from the captured image of the flow path 201 installed on the ship by using the deep learning model, and the moving distance of the detected individual exceeds the distance threshold value. Count as a fish. Therefore, it is possible to measure the exact number of fish that are the objects on board.

In the above, each component is assigned to each function (process), but this allocation is not limited to the above. Further, the above-described form is merely an example of the configuration of the component elements, and the present invention is not limited to this.

The processing performed by each of the above-described components may be performed by logic circuits manufactured according to the purpose. Further, a recording medium in which a computer program (hereinafter, referred to as a program) in which the processing contents are described as a procedure can be read by a device (for example, counting devices 100, 101; hereinafter, referred to as an information processing device) including each component. The program recorded on the recording medium may be read by the information processing apparatus and executed. Recording media that can be read by an information processing device include floppy (registered trademark) disks, magneto-optical disks, DVDs (Digital Versailles Disc), CDs (Compact Discs), Blu-ray (registered trademarks) Discs, and USB (Universal Serials). Bus) In addition to a transferable recording medium such as a memory, it refers to a memory such as a ROM (Read Only Memory) and a RAM (Random Access Memory) built in an information processing device, an HDD (Hard Disk Drive), and the like. The program recorded on the recording medium is read by a CPU provided in the information processing device, and the same processing as described above is performed under the control of the CPU. Here, the CPU operates as a computer that executes a program read from a recording medium in which the program is recorded. It should be noted that FPGA (Field Programmable Gate Array) may be used to realize the above-mentioned logic circuit.

100, 101 Counting device 110 Imaging unit 120, 121 Detection unit 130, 131 Learning model 140, 141 Counting unit 150 Output unit 161 Tracking unit 171 Distance measuring unit 200, 201 Flow path 211 Bottom surface 221,231 Side wall 301 Horizontal sheet 401 Hydrophilic sheet 402 Filter 501 Fish 511 Caudal fin 521 Detection frame

Claims (16)

The flow path in which the individual moves and
An imaging unit installed at a position where the individual moving in the flow path can be imaged,
A detection unit that detects the individual from the image captured by the imaging unit using a deep learning model,
A counting unit that counts the number of the individuals detected by the detection unit,
An output unit for outputting the number of said counting unit has counted,
A tracking unit that tracks the individual detected by the detection unit,
It has a distance measuring unit that measures the moving distance of the individual tracked by the tracking unit.
The counting unit is a counting system that counts the number of individuals detected by the detection unit whose moving distance measured by the distance measuring unit exceeds a predetermined distance threshold value . In the counting system according to claim 1 ,
In the counting unit, among the individuals detected by the detection unit, the moving distance measured by the distance measuring unit exceeds the distance threshold value, and the individual exists within the image range captured by the imaging unit. A counting system that counts the number of individuals whose time exceeds a predetermined time. In the counting system according to claim 1 or 2 .
A counting system in which a transparent hydrophilic member subjected to hydrophilic treatment is provided between the imaging unit and the flow path. In the counting system according to claim 3 ,
A counting system in which at least one of a polarizing filter and a dimming filter is provided between the imaging unit and the hydrophilic member. In the counting system according to any one of claims 1 to 4 ,
A counting system in which the flow path is arranged so as to be inclined with respect to a horizontal plane so that the individual moves in one direction along the bottom surface of the flow path. In the counting system according to claim 5 ,
The flow path is a counting system having a size that allows the individual to move while touching the bottom surface. In the counting system according to claim 5 or 6 .
The imaging unit is a counting system installed at a position capable of imaging a flow path on the downstream side of an intermediate point between the upstream and downstream of the flow path. In the counting system according to any one of claims 1 to 7 .
A counting system in which the individual is a fish. In the counting system according to claim 8 ,
The detection unit is a counting system that detects the fish by using the deep learning model for an image of a portion of the fish excluding the caudal fin. In the counting system according to claim 8 or 9 .
The target fish species of the individual is a counting system belonging to the genus Amberjack. An imaging unit installed at a position where an individual moving in the flow path can be imaged,
A detection unit that detects the individual from the image captured by the imaging unit using a deep learning model,
A counting unit that counts the number of the individuals detected by the detection unit,
An output unit that outputs the number counted by the counting unit and
A tracking unit that tracks the individual detected by the detection unit ,
It has a distance measuring unit that measures the moving distance of the individual tracked by the tracking unit.
The counting unit is a counting device that counts the number of the individuals detected by the detecting unit whose moving distance measured by the distance measuring unit exceeds a predetermined distance threshold value . In the counting device according to claim 11 ,
Among the individuals detected by the detection unit, the counting unit is such that the moving distance measured by the distance measuring unit exceeds the distance threshold value and the individual is within the image range captured by the imaging unit. A counting device that counts the number of individuals whose time exceeds a predetermined time. The process of imaging the flow path in which an individual moves, and
A process of detecting the individual from the captured image using a deep learning model, and
The counting process for counting the number of the detected individuals and
And outputting the number obtained by the counting,
The process of tracking the detected individual and
The process of measuring the movement distance of the tracked individual is performed.
The counting process is a counting method for counting the number of the detected individuals whose travel distance exceeds a predetermined distance threshold value . In the counting method according to claim 13 ,
In the counting process, among the detected individuals, the measured movement distance exceeds the distance threshold value, and the time during which the individual exists in the imaged image range exceeds a predetermined time. A counting method that counts the number of. On the computer
The procedure for imaging the flow path in which an individual moves, and
A detection procedure for detecting the individual from the captured image using a deep learning model, and
A counting procedure for counting the number of the detected individuals, and
An output procedure for outputting the number obtained by the counting,
A tracking procedure for tracking the individual detected by the detection procedure, and
A measurement procedure for measuring the movement distance of the individual tracked by the tracking procedure, and a measurement procedure.
As the counting procedure, a program for counting the number of the individuals whose movement distance measured by the measurement procedure exceeds a predetermined distance threshold value among the individuals detected by the detection procedure . In the program of claim 15 .
In the counting procedure, among the individuals detected by the detection procedure, the time when the moving distance measured by the measuring procedure exceeds the distance threshold value and the individual exists within the captured image range is predetermined. A program that counts the number of individuals over time.

Images