Fish length measuring device and fish length measuring method

Abstract

To provide a fish body length measurement device that measures an accurate fork length of a single fish.SOLUTION: A fish body length measurement device comprises: a camera system 105 that photographs a plurality of images in a time sequence; a fish detection unit 110 that detects fishes on the plurality of images; a portion detection unit 120 that detects a portion of the detected fish from the plurality of images; a tracking unit 130 that tracks identical fishes of the fishes detected on the plurality of images; and a body length calculation unit 160 that calculates a body length of the tracked identical fish.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to the technical field of measuring fish length, and more particularly to an apparatus for calculating fish length by selecting an image for calculating fish length from a plurality of images taken with a stereo camera. [Background technology]

[0002] The Background Description contains information useful in understanding the present invention. No admission is made that any of the information provided herein is prior art or relevant to the claimed invention, or that any publication specifically or implicitly referenced is prior art.

[0003] Aquaculture is concerned with the production of aquatic organisms, including fish, mollusks, crustaceans, aquatic plants, and freshwater and marine plants and animals under controlled conditions for all or part of their life cycle. For successful aquaculture, information on feeding costs, labor efforts, fish numbers, and fish length must be checked.

[0004] Therefore, there is a need for technology that can measure fish length without directly touching the fish.

[0005] Measuring fish length helps regulate the growth of fish in the pens. Fish abundance is determined by measuring biomass. There are established techniques that use the number and weight of fish to calculate biomass.

[0006] The fish weight is an important factor here, and depends on the accuracy of the "fork length" (fish length) and "body height" (fish height) measurements. Fish weight is calculated using formula (1) and is said to require an error of less than 5%. Fish weight = α × fork length 3 ... (1)

[0007] It has become clear that the error in calculating "fork length" should be within 2%. Therefore, the object of the present invention is to provide a device that can measure "fork length" with an error of 2% or less.

[0008] In conventional methods, fish length is measured using a stereo camera system.

[0009] The conventional approach consists of the following: -Measure the fork length of several fish and calculate the average The average value obtained is considered to be the true value of the fork length. Summary of the Invention [Problem to be solved by the invention]

[0010] However, the prior art has some problems. Referring to Figure 9(a) (prior art), the measured fork length 92 is always less than the true fork length 91, so the average of the measured values ​​is not the true value, and the fork length measurement error is generally more than 2%. Also, due to tail beat motion, the fork length measurement error is more than 2%.

[0011] The measurement error of the conventional technology is given by equation (2). Measurement error = random error + systematic error (2) Referring to Fig. 9(b) (prior art), the conventional method can calculate the random error by calculating the average value of the fork length measurements, but cannot calculate the systematic error, so the true value of the fork length cannot be calculated accurately.

[0012] The problem of calculating the exact fork length of a single fish is solved by the present invention.

[0013] In the conventional technology, oversampling is performed by capturing images at a frame rate such that the frame rate is greater than twice the tailbeat frequency.

[0014] In the known art, oversampling generates a heavy process load on the system.

[0015] This problem is solved in the present invention by a process of sampling single fish, either undersampling or oversampling, and a technique for changing the frame rate operation mode upon fish detection.

[0016] The present invention is characterized by using a stereo camera to continuously capture multiple images at specific time intervals. The present invention proposes a technique for extracting images in which the fish is not bent. This is achieved by measuring the distance between the camera and tail in each frame and extracting frames in which the distance between the camera and tail is at the intermediate distance between the maximum and minimum values.

[0017] The present invention provides frame rate calculation for the frame rate operation mode when fish are detected from tail beat frequency estimation. Such an approach reduces the load on the system. [Means for solving the problem]

[0018] Solutions to one or more shortcomings of the existing invention, as well as additional advantages, are provided in the present disclosure. Further features and advantages are realized through the technical features of the present disclosure. Other embodiments and aspects of the present disclosure are described in detail herein and are considered a part of the claimed disclosure.

[0019] The present invention relates to a fish length measuring device including a camera system, a fish detection unit, a part detection unit, a tracking unit, and a body length calculation unit. The camera system captures a plurality of images in time series. The fish detection unit detects fish in the plurality of images, and the part detection unit detects parts of the fish detected in the plurality of images. The tracking unit tracks the same fish among the fish detected in the plurality of images. The body length calculation unit calculates the body length of the same tracked fish.

[0020] In one aspect of the present invention, the distance calculation unit calculates camera-part distances between the camera system and parts of the same tracked fish on the multiple images. The image selection unit selects images of the same tracked fish from the multiple images in which the same fish is tracked based on the camera-part distances of the same tracked fish. The body length calculation unit calculates the body length of the same tracked fish from the selected images.

[0021] The image selection unit of the present invention searches for an intermediate image among multiple images of tracking the same fish, which has an intermediate distance between the camera parts, is smaller than other images having the largest inter-camera distance, and is larger than other images having the smallest inter-camera distance, and selects the intermediate image as the selected image.

[0022] Preferably, when a plurality of intermediate images are found, the image selection unit selects, as the selected image, an image having a camera-to-part distance at a median distance from among the plurality of intermediate images.

[0023] Preferably, when a plurality of intermediate images are found, the image selection unit selects, as a selected image, an image in which the camera-to-part distance is closest to the average distance from among the plurality of intermediate images.

[0024] Preferably, when a plurality of intermediate images are found, the image selection unit selects, from among the plurality of intermediate images, an image whose time is closest to the average time as the selected image.

[0025] In one aspect of the present invention, the part detection unit detects a part of the fish closer to the tail than to the mouth as the part.

[0026] In another aspect of the present invention, the part detection unit detects the tail of the fish as the part.

[0027] In one aspect of the present invention, the shooting frequency control unit is configured to control a time interval between multiple images taken by the camera system, the time interval decreasing when the fish detection unit detects a fish and increasing when the fish detection unit does not detect a fish.

[0028] Preferably, when the fish detection section detects a fish, the photographing frequency control section calculates a frequency component from the distance between the camera parts, and controls the time interval based on the frequency component.

[0029] In still another aspect of the present invention, the photographing frequency control section controls the time interval so that the sampling frequency corresponding to the time interval is at least twice the frequency component.

[0030] In one aspect of the present invention, the part position calculation unit calculates the positions of the parts of the same tracked fish on the multiple images and calculates a line that best fits the positions of the parts of the same tracked fish. The image selection unit selects images of the tracked same fish from the multiple images in which the same fish is tracked based on the relative positions of the parts and the line. The body length calculation unit calculates the body length of the tracked same fish from the selected images.

[0031] In one aspect of the present invention, the image selection unit selects images of the same tracked fish from among a plurality of images in which the position of the part is less than a predetermined distance from the line.

[0032] In one aspect of the present invention, the image selection unit selects an image of the same tracked fish from among a plurality of images in which the location of the part is closest to the line.

[0033] In one aspect of the present invention, the part detection unit detects a part of the fish closer to the tail than to the mouth as the part.

[0034] In one aspect of the present invention, the part detection unit detects the tail of the fish as the part.

[0035] In one aspect of the present invention, the shooting frequency control unit controls the time interval between multiple images taken by the camera system, which time interval decreases when the fish detection unit detects a fish and increases when the fish detection unit does not detect a fish.

[0036] According to another aspect of the present invention, a method includes capturing a plurality of images in a time sequence, detecting fish on the plurality of images, detecting parts of the detected fish on the plurality of images, and tracking a same fish among the detected fish on the plurality of images, and then calculating a body length of the tracked same fish.

[0037] The above summary is illustrative only and is not intended to be in any way limiting. In addition to the exemplary aspects, embodiments, and features described above, further aspects, embodiments, and features will become apparent by reference to the drawings and the following detailed description. Effect of the Invention

[0038] The present invention provides a device for accurately calculating the size of a fish using a stereo camera system by taking a plurality of photographs, selecting one of the photographs in which the fish is not bent, and calculating the size of the fish from the selected photograph. [Brief description of the drawings]

[0039] It should be noted that the drawings depict only typical embodiments of the present invention and should not be considered to limit its scope. A detailed description is provided with reference to the drawings. In the drawings, a reference number identifies the figure in which the reference number first appears. The same numbers are used throughout the drawings to reference features, components, etc. Some embodiments of systems, methods, or structures according to embodiments of the present invention are described with reference to the drawings, by way of example. [Figure 1] FIG. 2 is a block diagram showing structural features of the present invention. [Diagram 2] 13 is a flowchart illustrating an example of a procedure for calculating a camera-to-tail distance. [Diagram 3] 1 shows a graphical representation of camera-to-tail distance measurements over time. [Figure 4] 1 shows a graphical representation of the strobe fork length calculation taken as output. [Diagram 5]10 is a flowchart showing an example of a procedure for measuring a fork length. [Figure 6] 13 shows a graphical representation of frame extraction in relation to camera-to-tail distance. [Figure 7] FIG. 13 illustrates a graphical representation of frame rate control in two different states. [Figure 8] 10 is a flowchart showing an example of controlling a frame rate. [Figure 9] The problems with the prior art are shown below. [Figure 10] FIG. 2 is a block diagram showing structural features of another embodiment of the present invention. [Figure 11] 13 shows an example where the tail position over time is plotted.

[0040] The drawings depict embodiments of the present invention for purposes of example only. Those skilled in the art will readily recognize from the following description that alternative embodiments of the structures and methods illustrated herein may be used without departing from the principles of the disclosure described herein. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0041] A fish length measuring device is shown in Figures 1, 2 and 5. In this embodiment, the device includes stereo cameras 105, 205a and 205b, 505a and 505b, fish detection units 110, 210 and 510, part detection units 120, 220a and 520a, tracking units 130, 230 and 530, distance calculation units 140, 240 and 540, image selection unit 150 and body length calculation units 160 and 560.

[0042] In one embodiment of the present invention, the stereo camera 105, 205a, 205b, 505a, 505b, which may be called a camera system, captures a pair of images 205c, 505c in time series. The first camera 205a, 505a and the second camera 205b, 505b form a stereo camera facing the fish, and at time t1, pair 1 is captured, at time t2, pair 2 is captured, and so on. nNow take a pair of images of pair n in series, synchronized, for example, by general purpose input / output (GPIO) synchronization, bus timestamp synchronization, audio signal synchronization, or a similar conventional method.

[0043] The fish detector 110, 210, 510 is configured to perform the task of detecting fish on a plurality of images 205c, 505c. The fish are detected on each image and represented, for example but not limited to, in one of the following ways: A rectangular area surrounding a single fish indicates the presence of detected fish. Instead of rectangular regions, filled regions are still used. Fish detection is traditionally done by processing circuits or cameras. Traditional point detection methods are used to perform the task of detecting fish, such as machine learning, feature point detection, and edge detection techniques.

[0044] The part detection unit 120, 220a, 520a, also called tail detection unit, detects parts such as the tail of each of the paired fish on the multiple paired images 205c, and determines the two-dimensional position of each part on the paired images. The part (preferably the tail) is indicated by a bold dot in step 220a. The present invention is not limited to this, and a part close to the tail of the fish detected on the paired images (preferably closer to the tail than the mouth) can also be considered for the purpose of the present invention. There are several conventional point detection methods that can be used, such as feature point detection, edge detection, with or without machine learning capabilities.

[0045] The pairing unit 220b, 520b pairs fish on each pair of images 205c. It pairs the tails (i.e., parts) corresponding to each of the pair of images. It is clear that there are several conventional pairing methods that can be used, such as block matching, with or without machine learning capabilities.

[0046] In the present invention, there is no restriction on the execution order of the fish detection section and the pairing section, and the execution order of these sections can be interchanged.

[0047] The tracking unit 130, 230, 530 functions as a single fish tracking unit, and tracks a pair of fish in a pair of images captured in time series. The tracking unit 130, 230, 530 inputs the output of the fish detection unit 110, 210, 510 or the part detection unit 120, 220a, and tracks the detection target detected by these. The tracking unit 130, 230, 530 tracks the detection target at time t n A pair of fish photographed at time t n+1 The purpose of the present invention is to track a fish over time in order to know whether a paired fish photographed in the same place is the same fish or not. The tracking units 130, 230, and 530 regard the fish as a target and represent the target in various ways, such as a part of a rectangle, a filled outline, or a dot.

[0048] The tracker 130, 230, 530 compares the fish detected in other frames in a variety of ways, including: a) Applying conventional fish tracking algorithms presented in the literature such as: ·DIDSON-based Acoustic Counting Method for Juvenile Ayu Plecoglossus altivelis Migrating Upstream by Jun HAN et. al. ·Particle Filter-Based Predictive Tracking for Robust Fish Counting by Erikson Morais et. al. b) Applying classical image matching algorithms such as template matching and point feature matching.

[0049] The distance calculation unit 140, 240, 540 performs camera-to-part distance calculations between the cameras 105, 205a and 205b, 505a and 505b and the same tracked part of the fish (preferably the tail) on multiple images.

[0050] The objective of the present invention is to extract the strobe at the time t when the fish is not bending its body. A strobe should be understood by those skilled in the art as a burst of light that allows the camera to capture an image. 340, 440 in Fig. 3 and Fig. 4 illustrate the camera-tail distance over time. Different strobes, e.g., strobe 1, strobe 2, strobe 3, are depicted and the distance between the camera and the tail is represented at different times t on the X-axis. The camera-tail distance on each frame (each strobe is considered as a frame that captures a photo) is measured by calculating the 2D position of the tail on the paired images and the 3D position of the tail using triangulation techniques to calculate the distance between the tail and the camera.

[0051] The image selection unit 150 extracts frames where the camera-to-tail distance is at an intermediate distance 440a between the maximum and minimum values. The extracted frames are considered to be frames where the fish is not bending its body, i.e., frames where the true fork length 460a at 460 in FIG. 4 is calculated or strobes are photographed.

[0052] In one embodiment of the present invention, there may be some frames at intermediate distances between the maximum and minimum camera-tail distances, so an additional extraction process may be required. The image selector 150 performs a frame extraction process to select images that are considered as intermediate images of the same tracked fish from among the multiple images of the same tracked fish based on the camera-tail distances of the same tracked fish. Figure 6 discloses frames that can be considered before extraction, including, but not limited to, the following: a) All frames from the frame with the maximum (minimum) distance to the frame with the next minimum (maximum) distance b) All frames from the frame with the minimum (maximum) distance to the frame with the next minimum (maximum) distance c) Every frame of a tracked fish The present invention discloses three possible processes for extracting frames, including but not limited to the following: 1) Extract the frame in the middle of the distance 2) Extract the frame whose distance is closest to the average distance 3) Extract the frame whose time is closest to the average time

[0053] As an example of the above three processes 1), 2), and 3), consider the case where n frames are candidates after performing any of the above steps a), b), and c). In process 1), the median of the n camera-to-tail distances is calculated from the n frames, and the frame corresponding to the calculated median is the extracted frame. In process 2), the average of the n camera-to-tail distances is calculated from the n frames, and the frame with the camera-to-tail distance closest to the calculated average is the extracted frame. In process 3), the average time of n frames is calculated from the n frames, and the frame with the time closest to the calculated average is the extracted frame.

[0054] In a preferred embodiment, Figure 6 is a graphical representation of the results after considering the frames according to step a) above (the three frames (frames 3-5) between frame 2 with the maximum camera-to-tail distance and frame 6 with the next minimum camera-to-tail distance) and performing the frame extraction process according to step 1) above. 640 represents the frame with the median camera-to-tail distance out of the three frames.

[0055] After selecting an appropriate frame based on the distance between camera parts of the same tracked fish, the body length is calculated.

[0056] The body length calculation unit 160 calculates the body length of the tracked same fish from the selected image. The fork length is calculated at 560 in FIG. 5. The body length calculation unit 160 performs 3D reconstruction and body length calculation. The fork length is calculated from the fish on the extracted frame as follows: a) Detection of two body parts of a fish, e.g. the head and tail b) Performing a 3D reconstruction of the 2D positions of the two parts using triangulation techniques c) Calculate the distance between the two parts

[0057] In another embodiment of the present invention, a frame rate control process is disclosed. In the present invention, fish can be detected in Fig. 7 and Fig. 8. When a fish is detected, the camera goes into an active state 740c and the distance between the camera and the tail is measured in time series. On the other hand, if no fish is detected for a certain period of time, the camera goes into an idle state 740d. In this embodiment, a shooting frequency control unit 800 controls the time interval between multiple images taken by the camera system.

[0058] The photographing frequency control unit 800 shortens the time interval when the fish detection unit detects a fish, and lengthens the time interval when the fish detection unit does not detect a fish.

[0059] In a preferred embodiment, when the fish detection unit detects a fish, the photographing frequency control unit 800 calculates a frequency component from the camera-to-camera distance and controls the time interval based on the frequency component. In another preferred embodiment, the photographing frequency control unit controls the time interval so that the sampling frequency corresponding to the time interval is set to at least twice the frequency component.

[0060] In one embodiment of the present invention, the active state frame rate is calculated from the tail beat frequency, which can be estimated in a variety of ways, including: a) Estimation from swimming speed and fish species b) Estimation of the distance between the camera and the tail from undersampling values

[0061] In the preferred embodiment, the tailbeat frequency is used to determine the frame rate in the "active state" and adjust the oversampling process accordingly. The calculated tailbeat frequency is then used to finally adjust the frame rate to be consistent with the sampling theorem.

[0062] In another embodiment of the present invention, in Fig. 10, the device for measuring fish length includes stereo cameras 105, 205a and 205b, 505a and 505b, fish detection units 110, 210, 510, part detection units 120, 220a, 520a, tracking units 130, 230, 530, part position calculation units 1040, image selection units 1050, and body length calculation units 160, 560. Compared with the above embodiment and Fig. 1, the distance calculation units 140, 240, 540 are replaced by the part position calculation unit 1040, the image selection unit 150 is replaced by the image selection unit 1050, and the others remain the same.

[0063] The part position calculation unit 1040 calculates the position of the part (preferably the tail) of the same tracked fish on the multiple images. The part position is calculated by a triangulation technique from the two-dimensional position of the part on the paired images.

[0064] Figure 11 shows an example where the positions 1101, 1102, 1103, 1104, 1105, 1106 of the tail of the same fish tracked by calculation from multiple images are plotted. In this example, position 1106 is the most recent position of the fish and position 1101 is the oldest position of the fish. In Figure 11, the position of the fish tail is projected onto a two-dimensional horizontal plane XZ, where the Z axis is parallel to the lenses of the stereo camera.

[0065] The part position calculation unit 1040 further calculates a line 1141 that best fits the positions 1101, 1102, 1103, 1104, 1105, 1106 of the tail of the same fish tracked on the multiple images. For example, but not limited to, the line 1141 can be calculated by regression analysis using the least squares method. The line 1141 may be a three-dimensional line calculated in three dimensions from the three-dimensional position of the tail. The line 1141 may be a two-dimensional line calculated in two dimensions after projecting the three-dimensional position of the tail onto a two-dimensional horizontal plane XZ. The line 1141 represents the trajectory of the same tracked fish.

[0066] The image selection unit 1050 extracts frames (images) in which the fish is not bending its body based on the relative positions of the tail and the line 1141. More specifically, as an example, the image selection unit 1050 may select a frame in which the position of the tail is close to the line 1141, i.e., less than a predetermined distance from the line 1141. Preferably, the image selection unit 1050 may select a frame in which the position of the tail is closest to the line 1141, for example, position 1103 in FIG. 11.

[0067] In this embodiment, the site is preferably the tail or part of the tail, however, the site may be another part of the fish, preferably closer to the tail than the mouth of the fish.

[0068] The fish length measuring device of this embodiment may include the photographing frequency control unit 800 of the above-described embodiment that controls the time interval between multiple images taken by the camera system. The photographing frequency control unit 800 performs control to shorten the time interval when the fish detection unit detects a fish and to lengthen the time interval when the fish detection unit does not detect a fish.

[0069] Next, the technical effects of the embodiment of the present invention will be described. Conventionally, the size of a fish is calculated by taking pictures of the fish with a camera at different times and calculating the fork length. However, tail beat movement and systematic errors are not taken into account in the calculation of the fork length. As a result, the calculated result of the fork length is smaller than the true value, with an error of more than the allowable 2%.

[0070] To solve this problem, in an embodiment of the present disclosure, a frame in which the fish is not bending its body is extracted. When measuring the fork length, the present invention measures the distance between the camera and the part for each frame, and extracts a frame that is at an intermediate distance between the maximum and minimum values ​​of the distance. Also, when measuring the fork length, the present invention measures the part position of each frame, and extracts a frame in which the part position is close to the trajectory of the fish.

[0071] As a result, it is possible to calculate the fork length and therefore the fish length more accurately. [Explanation of symbols]

[0072] 105, 205a, 205b, 505a, 505b Stereo camera 110,210,510 Fish detector 120, 220a, 520a Part detection section 130,230,530 Tracking Department 140,240,540 Distance calculation part 150,1050 Image selection section 160,560 Body Length Calculation Section 205c 505c Images 220b,520b Pairing section 340,440 Camera tail distance 440a Mid-range 360,460 Fork length 460a True fork length 640 Median distance between camera tails 740c active 740d idle 800 Shooting frequency control section 91 True value of fork length 92 Fork length measurement 1040 Part position calculation section 1101,1102,1103,1104,1105,1106 Fish tail position 1141 Line

Claims

1. A camera system that captures multiple images in time series; a fish detection unit for detecting fish on the plurality of images; a part detection unit that detects parts of the detected fish from the plurality of images; a tracking unit that tracks a single fish among the fish detected on the plurality of images; a distance calculation unit for calculating a camera-to-camera part distance between the camera system and the tracked same fish part on the plurality of images; an image selection unit that searches for an intermediate image at an intermediate distance, which has a smaller inter-camera part distance than other images having a maximum inter-camera part distance and a larger inter-camera part distance than other images having a minimum inter-camera part distance, from among the multiple images obtained by tracking the same fish, and selects the intermediate image as a selected image; a body length calculation unit for calculating a body length of the tracked same fish from the selected image; Equipped with A fish length measuring device characterized by the above.

2. The fish length measuring device according to claim 1, When a plurality of intermediate images are found, the image selection unit selects an image having a camera-part distance at a median distance from among the plurality of intermediate images as a selected image. A fish length measuring device characterized by the above.

3. The fish length measuring device according to claim 1, When a plurality of intermediate images are found, the image selection unit selects, from among the plurality of intermediate images, an image having a camera-part distance that is closest to an average distance as a selected image. A fish length measuring device characterized by the above.

4. The fish length measuring device according to claim 1, When a plurality of intermediate images are found, the image selection unit selects, from among the plurality of intermediate images, an image whose time is closest to the average time as a selected image. A fish length measuring device characterized by the above.

5. The fish length measuring device according to any one of claims 1 to 4, The part detection unit detects a part of the fish closer to the tail than to the mouth as the part of the fish. A fish length measuring device characterized by the above.

6. The fish length measuring device according to any one of claims 1 to 5, The part detection unit detects the tail of the fish as the part. A fish length measuring device characterized by the above.

7. The fish length measuring device according to any one of claims 1 to 6, a photographing frequency control unit that controls a time interval between a plurality of images photographed by the camera system so as to decrease when the fish detection unit detects a fish and to increase when the fish detection unit does not detect a fish; A fish length measuring device characterized by the above.

8. The fish length measuring device according to claim 7, When the fish detection unit detects a fish, the photography frequency control unit Calculating a frequency component from the camera-to-part distance; controlling the time interval based on the frequency components; A fish length measuring device characterized by the above.

9. The fish length measuring device according to claim 8, the photographing frequency control unit controls the time interval so that a sampling frequency corresponding to the time interval is at least twice as high as the frequency component; A fish length measuring device characterized by the above.

10. A camera system for capturing a plurality of images in time series; a fish detection unit for detecting fish on the plurality of images; a part detection unit that detects parts of the detected fish from the plurality of images; a tracking unit that tracks a single fish among the fish detected on the plurality of images; a part position calculation unit that calculates the positions of the tracked parts of the same fish on a plurality of images and calculates a line that best fits the positions of the tracked parts of the same fish; an image selection unit that selects an image of the same tracked fish from among a plurality of images in which the same fish is tracked, based on the relative positions of the parts and the lines; a body length calculation unit for calculating a body length of the tracked same fish from the selected image; Equipped with A fish length measuring device characterized by the above.

11. The fish length measuring device according to claim 10, the image selection unit selects an image of the same tracked fish from among a plurality of images in which the position of the part is less than a predetermined distance from the line; A fish length measuring device characterized by the above.

12. In the fish length measuring device according to claim 10 or 11, the image selection unit selects an image of the same tracked fish from among a plurality of images in which a position of a part is closest to the line; A fish length measuring device characterized by the above.

13. Multiple images are captured in chronological order. Detecting fish in the plurality of images; detecting parts of the detected fish on the plurality of images; Tracking a same fish among the fish detected on the plurality of images; Calculating a camera-to-camera location distance between a camera system capturing the plurality of images in time sequence and the tracked same fish location on the plurality of images; Searching for an intermediate image at an intermediate distance, which has a smaller inter-camera part distance than other images having a maximum inter-camera part distance and a larger inter-camera part distance than other images having a minimum inter-camera part distance, from among the plurality of images obtained by tracking the same fish; selecting the intermediate image as a selected image; calculating a body length of the tracked identical fish from the selected images; A method for measuring fish length.

14. Taking a plurality of images in time series, Detecting fish in the plurality of images; detecting parts of the detected fish on the plurality of images; Tracking a same fish among the fish detected on the plurality of images; Calculating the positions of the tracked fish parts on a plurality of images; Calculate a line of best fit to the locations of the tracked identical fish parts; selecting an image of the same tracked fish from among a plurality of images in which the same fish is tracked based on the relative positions of the parts and the line; calculating a body length of the tracked identical fish from the selected images; A method for measuring fish length.

Images