A method for estimating the spatial distribution state of caged fish based on omnidirectional sonar

Abstract

The invention discloses a method for estimating the spatial distribution state of fish in a cage based on an omnidirectional sonar. The method comprises the following steps: the omnidirectional sonar is used to perform an overall scan on a tested cage to convert the scan into an overall sonar scan video of the cage, and then a fish automatic recognition model is trained to identify the position of each identified fish in each frame of a fish school sonar image of the overall sonar scan video of the cage; and at a target time when the tested cage needs to be estimated, the overall scan and the fish automatic recognition model are used to obtain the actual three-dimensional coordinates of each identified fish in each frame of a fish school sonar image of the overall sonar scan video of the target cage, and then a plurality of water layer data are set according to a preset water depth interval and a water depth overlap interval, and density clustering processing is performed on the actual three-dimensional coordinates of each water layer data, and the actual three-dimensional coordinates of each cage fish in the tested cage are calculated, and the spatial distribution state of the cage fish in the tested cage is estimated based on the coordinates.

Description

Technical Field

[0001] The invention relates to fishery resource estimation and measurement, in particular to a method for estimating the spatial distribution state of caged fish based on omnidirectional sonar. Background Art

[0002] In its 2022 report, the Food and Agriculture Organization of the United Nations (FAO) announced that due to the rapid growth of aquaculture, global fisheries and aquaculture production has reached a record high, making an increasingly significant contribution to ensuring food security and providing adequate nutrition in the 21st century. The role of aquatic food in the global food supply chain is becoming increasingly critical. Deep sea areas have a vast area and good water quality. Unlike terrestrial aquaculture, they have become an emerging space for the development of modern marine fishery economy. The full utilization of the deep sea is essential to meet the growing global food demand and provide a broader development prospect for the fishery industry. In this context, sonar detection technology, as the main means of underwater long-distance communication, has become an indispensable tool for ocean exploration. Sonar technology plays a vital role in seabed topography mapping, underwater target object identification and detection, and biological population monitoring. Its wide application provides strong support for the scientific research and sustainable management of deep-sea resources, promotes the sustainable development of marine fisheries, and contributes to global food security. This also highlights the key role of scientific and technological innovation in promoting new breakthroughs in the field of fisheries. As an important part of cage aquaculture, the monitoring of fish living habits and spatial distribution status has far-reaching significance in the following aspects: first, reasonable feeding and management strategies and accurate grasp of the distribution of fish in cages can allow farmers to divide the breeding areas according to the distribution of fish and improve efficiency; second, ignorance of the distribution of fish will lead to unreasonable feed delivery and waste of resources; third, lack of understanding of the living habits of fish will lead to unreasonable breeding operations and increase breeding losses.

[0003] In response to the above requirements, although some different fish distribution status monitoring methods have been proposed, their estimation effects on fish distribution and habits are not ideal and the cost is high. In addition, the monitoring requirements for breeding conditions are high, making it difficult to conveniently and effectively implement sonar monitoring and fish distribution assessment. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a method for estimating the spatial distribution state of caged fish based on omnidirectional sonar, so as to solve the problems of poor effect, high cost and high monitoring requirements of existing fish spatial distribution monitoring methods.

[0005] To solve the above technical problems, the technical solution adopted by the present invention is as follows:

[0006] A method for estimating the spatial distribution state of caged fish based on omnidirectional sonar, characterized by comprising:

[0007] Step S1, placing an omnidirectional sonar in the tested cage, and driving the omnidirectional sonar to move at a constant speed between the water surface and the bottom of the cage in the vertical direction, so as to perform an overall scan of the tested cage by the omnidirectional sonar during the movement, wherein the omnidirectional sonar from the water surface to the bottom of the cage, or from the bottom of the cage to the water surface can be regarded as an overall scan;

[0008] In addition, the data outputted by the omnidirectional sonar after an overall scan is played back through the supporting software of the omnidirectional sonar to be converted into a video of the overall sonar scan of the cage, and the video frames of the video of the overall sonar scan of the cage are recorded as the sonar image of the school of fish, see Figure 3 ;

[0009] Step S2: Scan the tested cage as a whole several times according to step S1 to obtain a sufficient number of cage sonar scanning videos as training data, so as to obtain a fish automatic recognition model through deep learning model training. The fish automatic recognition model can recognize the position of each identified fish in each frame of fish school sonar image in the cage sonar scanning video. Figure 4 ;

[0010] Step S3, at the target time, perform an overall scan of the tested cage according to step S1 to obtain an overall sonar scanning video of the target cage, and use the fish automatic recognition model to identify the position of each identified fish in each frame of the fish school sonar image of the overall sonar scanning video of the target cage;

[0011] Furthermore, the actual three-dimensional coordinates (x, y, d) of each identified fish in each frame of the fish school sonar image of the overall sonar scanning video of the target cage are extracted and converted from the recognition results of the fish automatic recognition model, that is, from the pixel coordinates and water depth obtained by recognition, wherein x and y represent the horizontal plane abscissa and horizontal plane ordinate of the identified fish at the time of recognition (i.e., the time when the omnidirectional sonar scanning obtains the corresponding fish school sonar image), respectively, and d represents the water depth of the identified fish at the time of recognition;

[0012] The pixel coordinates in the fish school sonar image are converted into x, y in the actual three-dimensional coordinates by using the conversion method in the prior art, for example:

[0013] See also Figure 4, the fish automatic identification model uses a rectangular prediction box to frame the position of the identified fish, and the coordinates of the prediction box are in a coordinate system with the upper left corner of the sonar image as the origin, the right as the positive direction of the x-axis, and the downward as the positive direction of the y-axis. The lengths of the x and y coordinate axes are both [0,1], which is the ratio of the number of pixels in the prediction box to the length and width of the sonar image. Therefore, it is necessary to convert according to the size of the image generated by the sonar and the actual distance measured in reality, so that when processing the coordinates later, they can be assigned according to the actual length of the experimental fish. At the same time, it can also meet certain conditional constraints when clustering coordinate points taking into account realistic factors. Here, we adjust the two-dimensional coordinate system, converting the coordinate system with the upper left corner of the sonar image as the origin to the coordinate system with the center point of the sonar image as the origin. The directions of the x and y axes remain unchanged. The conversion formula is as follows:

[0014]

[0015] Where: x, y are the coordinates of the coordinate system with the upper left corner as the origin, X, Y are the coordinates of the coordinate system with the center point as the origin, and x, y, X, Y are the ratios of the image size to the actual size.

[0016] The pixel size of the sonar image used in this experiment is 804×604, and the range radius used by the sonar is 300 cm. Therefore, the formula for converting the coordinates in the image into the coordinate size in reality is:

[0017]

[0018] Where: x, y are the coordinates of the coordinate axis with the center point of the sonar image as the origin after the above processing, and X, Y are the actual coordinate data corresponding to x, y in reality.

[0019] For depth data, since the sonar in our experiment can be regarded as a uniform downward movement from the water surface (i.e., the starting depth is 0 cm), the sonar speed per second (v) (unit: cm / s) can be calculated based on the sonar detection time (T) and the detection depth (H). Then, based on the corresponding relationship between the total number of frames of the sonar video image (P) and the sonar time (T), the number of seconds per frame during sonar measurement (t) (unit: s) (i.e., the inverse of the frame rate) can be obtained. Then, the depth (h) corresponding to each frame during sonar measurement can be calculated based on the sonar movement speed and the time length of each sonar frame. The calculation formula is as follows:

[0020]

[0021] Step S4, starting from the water surface (i.e. d=0), multiple water layer data are set according to a preset water depth interval △d, and two adjacent water layer data have a preset water depth overlap interval δd, that is: dmax-dmin=△d, d1max-d2min=δd, dmax and dmin respectively represent the maximum water depth and the minimum water depth of any water layer data, d1max and d2min respectively represent the maximum water depth of the previous water layer data and the minimum water depth of the next water layer data in two adjacent water layer data; because it is considered that during measurement, fish may run to different water layers and generate similar signal data in different water layers, so ensuring that there is an overlap between two adjacent water layer data can minimize the influence of the coordinates of the center point fish after clustering; and because the subsequent step is clustering, even if the counting is repeated, it will not have much impact on the clustering result.

[0022] Furthermore, according to the water depth d, the actual three-dimensional coordinates (x, y, d) of each identified fish obtained in step S3 are respectively attributed to the corresponding water layer data, wherein the actual three-dimensional coordinates (x, y, d) within the overlapping interval δd of the water depth are attributed to both the upper and lower water layer data;

[0023] Step S5: for each of the water layer data, density clustering processing is performed on the actual three-dimensional coordinates (x, y, d) belonging to the water layer data to obtain a cluster of each water layer data; wherein each cluster can be identified as a fish in the corresponding water layer data range, and a cluster containing only one actual three-dimensional coordinate (x, y, d) is recorded as a noise point cluster, and a cluster containing at least two actual three-dimensional coordinates (x, y, d) is recorded as a core object cluster;

[0024] For example: Figure 5 This is the spatial distribution map of 7681 actual three-dimensional coordinates (x, y, d) contained in a certain water layer data. Figure 6 for Figure 5 The spatial distribution diagram of 17 clusters obtained after the water layer data is processed in step S5 is shown, wherein the 17 clusters, namely Cu0 to Cu16, are marked with different colors respectively.

[0025] Step S6, calculating the actual three-dimensional coordinates (x', y', d') of each fish in the tested cage:

[0026] For the noise point cluster, the only actual three-dimensional coordinate (x, y, d) contained in it is used as the actual three-dimensional coordinate (x', y', d') of the fish in the cage corresponding to the noise point cluster;

[0027] For the core object cluster, calculate the center point coordinates of all the actual three-dimensional coordinates (x, y, d) contained in it, as the actual three-dimensional coordinates (x', y', d') of the fish in the cage corresponding to the core object cluster; that is, x' is the average value of all the aforementioned x, y' is the average value of all the aforementioned y, and d' is the average value of all the aforementioned d;

[0028] Among them, x', y' and d' represent the abscissa of the actual horizontal plane of the fish in the cage, the ordinate of the actual horizontal plane of the fish and the actual water depth of the fish, respectively;

[0029] When the water depth overlap interval δd is set for the data of two adjacent water layers and δd≠0, the coordinates of the center point can be calculated more accurately after clustering.

[0030] Step S7: Estimate the spatial distribution state of the fish in the tested cage according to the actual three-dimensional coordinates (x', y', d') of the fish obtained in step S6.

[0031] Preferably: in step S1, the omnidirectional sonar is placed at the central axis position of the tested cage and is driven by a vertical lifting drive device to move at a uniform speed between the water surface and the bottom of the cage in the vertical direction; wherein the vertical lifting drive device can be a mechanical lifting device driven by a motor, or a rope driven by manpower, and a scale can be set on the rope to facilitate depth control and recording.

[0032] Among them, the omnidirectional sonar can be a sonar that can simultaneously achieve 360° omnidirectional scanning, or a sonar that achieves 360° omnidirectional scanning by driving the scanning beam to rotate in the horizontal plane. When using the latter, the vertical movement speed of the omnidirectional sonar cannot be too fast to avoid failure to fully scan the entire space of the measured cage.

[0033] Wherein, the beam opening angle range of the scanning beam of the omnidirectional sonar is [a°, b°];

[0034] In step S4, the water depth interval Δd is calculated according to the following formula:

[0035] △d=(△d1+△d2) / 2;

[0036] △d1=2*tan(a / 2)*R;

[0037] △d2=2*tan(b / 2)*R;

[0038] Where △d1 and △d2 are the water depth intervals when the beam opening angles are a° and b° respectively; R is the radius of the measured cage.

[0039] Therefore, the overall scanning of the tested network box by the omnidirectional sonar can be realized more quickly and effectively.

[0040] Wherein, in step S4, the water depth overlap interval δd is taken as the smaller value between the average fish length and the average fish width of the identified fish.

[0041] Thus, it is possible to avoid repeated counting of fish in the two adjacent water layer data due to fish wandering within the overlapping interval δd of the water depth between the two adjacent water layer data.

[0042] In step S5, the DBSCAN clustering analysis method is used to perform density clustering processing on the actual three-dimensional coordinates (x, y, d) belonging to the same water layer data; wherein the clustering parameters are set in the following manner:

[0043] The clustering radius (epsilon) takes values ​​in the range of 0.5N to 0.8N, where N is the larger value of the average fish length and the average fish width of the identified fish; when the clustering radius is larger within this range, the number of clusters obtained by clustering will decrease, but considering the movement of fish in the water and the position changes between fish, the error of the number after clustering is small; the increase in the clustering radius will cause the offset of the center point of the fish, but it will have little effect on the spatial distribution characteristics of the overall fish in the cage.

[0044] The minimum number of samples (min_samples) is set to 1, which means that the cluster obtained by clustering can contain at least one sample, that is, the actual three-dimensional coordinates (x, y, d).

[0045] Preferably, in step S2, the deep learning model adopts the YOLOv8 model, and the fish automatic identification model is trained in the following manner:

[0046] Step S2-1, intercepting a sufficient number of fish sonar images from the overall sonar scanning video of the cage, and annotating the position of each identified fish in each frame of the fish sonar image using the MakeSense online data annotation software to output an annotation result file;

[0047] Step S2-2: using the whole cage sonar scanning video as input data, using the annotation result file corresponding to the whole cage sonar scanning video as output data, and training the YOLOv8 model to obtain the fish automatic identification model.

[0048] Preferably, in step S7, the spatial distribution state estimation includes: drawing a horizontal distribution diagram of the fish in the tested cage according to the fish actual horizontal plane horizontal coordinate x' and the fish actual horizontal plane vertical coordinate y' obtained in step S6, see Figure 7 .

[0049] Preferably, in step S7, the spatial distribution state estimation includes: drawing a spatial distribution map of the caged fish in the water depth range corresponding to each water layer data according to the actual three-dimensional coordinates (x', y', d') of the fish obtained in step S6, see Figure 8 .

[0050] Preferably, in step S7, the spatial distribution state estimation includes: drawing a spatial distribution map of the fish in the tested cage according to the actual three-dimensional coordinates (x', y', d') of the fish obtained in step S6, see Fig. 9 .

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] First, the present invention uses an omnidirectional sonar 1 to perform an overall scan on the tested net cage 2 to convert it into an overall sonar scanning video of the net cage, and then trains to obtain a fish automatic recognition model that can recognize the position of each identified fish in each frame of the fish school sonar image of the overall sonar scanning video of the net cage; and, at the target time when the tested net cage 2 needs to be estimated, the actual three-dimensional coordinates (x, y, d) of each identified fish in each frame of the fish school sonar image of the target net cage overall sonar scanning video are obtained through overall scanning and the fish automatic recognition model, and then a plurality of water layer data are set according to the preset water depth interval △d and the water depth overlap interval δd, and the actual three-dimensional coordinates (x, y, d) of each water layer data are subjected to density clustering processing, and the actual three-dimensional coordinates (x', y', d') of each cage fish in the tested net cage 2 are calculated, and based on this, the spatial distribution state of the cage fish in the tested net cage 2 is estimated, so that the breeder can timely understand the distribution of the cage fish in the tested net cage 2.

[0053] Second, since the present invention only needs to use the omnidirectional sonar 1 to obtain all the data required for the implementation method according to step S1, the requirements for scanning the omnidirectional sonar 1 are relatively low. It only needs to scan the complete measured network box 2 at a uniform speed in the vertical direction. It has the advantages of being simple, convenient, efficient and fast, and can solve the problems of the prior art in terms of high requirements for data monitoring and collection of the measured network box 2, expensive monitoring equipment, complex operation and high cost.

[0054] Third, since the present invention does not have strict requirements on the quality of the data obtained by the omnidirectional sonar 1 scanning, even if the fish in the tested cage 2 are scanned repeatedly, the density clustering processing in step S5 can be used to reduce the impact of repeated scanning, so as to obtain a higher estimation accuracy of the spatial distribution state of fish in the cage, thereby solving the problem that the estimation effect of the prior art is not ideal.

[0055] Fourth, the present invention estimates the spatial distribution status of the caged fish in the tested cage 2, which can help the farmers to feed the fish at designated locations according to the spatial distribution status and living habits of the fish in the cage, and adjust the breeding plan in time, thereby reducing feed waste, effectively reducing breeding costs, improving breeding efficiency, and maximizing profits and minimizing losses in breeding.

[0056] Fifth, the present invention can conveniently check the breeding density of fish in the cages by drawing out the spatial distribution map of the tested cages 2 in the water depth range corresponding to each water layer data. This can greatly increase the understanding of the fish breeding density and reduce the losses caused by misestimation of breeding density. In addition, the breeders know the distribution of fish underwater, so they only need to place feed at fixed points in the water layer where fish are more concentrated, which can reduce feed losses and improve feed conversion rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments:

[0058] Figure 1 It is a flowchart of the method for estimating the spatial distribution state of fish in cages of the present invention;

[0059] Figure 2 It is a schematic diagram of the omnidirectional sonar of the present invention being placed in a tested net box;

[0060] Figure 3 is a schematic diagram of a sonar image of a school of fish in step S1 of the present invention;

[0061] Figure 4 A schematic diagram of the fish automatic identification model identifying the position of the identified fish in the fish school sonar image in step S2 of the present invention;

[0062] Figure 5 It is the spatial distribution map of 7681 actual three-dimensional coordinates (x, y, d) contained in a certain water layer data;

[0063] Figure 6 for Figure 5 The spatial distribution map of the clusters obtained after the water layer data is processed in step S5;

[0064] Figure 7 It is the horizontal distribution map of fish in a tested cage;

[0065] Figure 8 The spatial distribution map of the caged fish in the water depth range corresponding to each water layer data of a tested cage divided into 11 water layer data;

[0066] Fig. 9 This is the spatial distribution map of fish in a tested cage. DETAILED DESCRIPTION

[0067] The present invention is described in detail below in conjunction with the embodiments and the accompanying drawings to help those skilled in the art better understand the inventive concept of the present invention. However, the protection scope of the claims of the present invention is not limited to the following embodiments. For those skilled in the art, all other embodiments obtained without creative work without departing from the inventive concept of the present invention belong to the protection scope of the present invention.

[0068] like Figures 1 to 4 As shown, the present invention discloses a method for estimating the spatial distribution state of fish in cages based on omnidirectional sonar, comprising:

[0069] Step S1, placing the omnidirectional sonar 1 in the tested cage 2, and driving the omnidirectional sonar 1 to move at a constant speed between the water surface 2a and the cage bottom 2b in the vertical direction, so as to perform an overall scan of the tested cage 2 through the omnidirectional sonar 1 during the movement, wherein the omnidirectional sonar 1 from the water surface 2a to the cage bottom 2b, or from the cage bottom 2b to the water surface 2a can be regarded as an overall scan;

[0070] In addition, the data outputted by the omnidirectional sonar 1 after an overall scan is played back through the supporting software of the omnidirectional sonar 1 to be converted into a video of the overall sonar scan of the cage, and the video frames of the video of the overall sonar scan of the cage are recorded as the sonar image of the school of fish, see Figure 3 ;

[0071] Preferably: in step S1, the omnidirectional sonar 1 is placed at the central axis position of the measured cage 2 and is driven by a vertical lifting drive device to move at a uniform speed between the water surface 2a and the bottom surface 2b of the cage in the vertical direction; wherein the vertical lifting drive device can be a mechanical lifting device driven by a motor, or a rope driven by manpower, and a scale can be set on the rope to facilitate depth control and recording.

[0072] Among them, the omnidirectional sonar 1 can adopt a sonar that can simultaneously realize 360° omnidirectional scanning, or it can adopt a sonar that realizes 360° omnidirectional scanning by driving the scanning beam 1a to rotate in the horizontal plane. When using the latter, the vertical movement speed of the omnidirectional sonar 1 cannot be too fast to avoid failure to fully scan the entire space of the measured cage 2.

[0073] Step S2: Scan the tested cage 2 as a whole for multiple times according to step S1, and obtain a sufficient number of cage overall sonar scanning videos as training data, so as to obtain a fish automatic recognition model through deep learning model training, and the fish automatic recognition model can recognize the position of each identified fish in each frame of fish school sonar image of the cage overall sonar scanning video, see Figure 4 ;

[0074] Preferably, in step S2, the deep learning model adopts the YOLOv8 model, and the fish automatic identification model is trained in the following manner:

[0075] Step S2-1, intercepting a sufficient number of fish sonar images from the overall sonar scanning video of the cage, and annotating the position of each identified fish in each frame of the fish sonar image using the MakeSense online data annotation software to output an annotation result file;

[0076] Step S2-2: using the whole cage sonar scanning video as input data, using the annotation result file corresponding to the whole cage sonar scanning video as output data, and training the YOLOv8 model to obtain the fish automatic identification model.

[0077] Step S3, at the target time, perform an overall scan of the tested cage 2 according to step S1 to obtain an overall sonar scanning video of the target cage, and use the fish automatic recognition model to identify the position of each identified fish in each frame of the fish school sonar image of the overall sonar scanning video of the target cage;

[0078] Furthermore, the actual three-dimensional coordinates (x, y, d) of each identified fish in each frame of the fish school sonar image of the overall sonar scanning video of the target cage are extracted and converted from the recognition result of the fish automatic recognition model, that is, from the pixel coordinates and water depth obtained by recognition, wherein x and y represent the horizontal plane abscissa and horizontal plane ordinate of the identified fish at the time of recognition (i.e., the time when the omnidirectional sonar 1 scans and obtains the corresponding fish school sonar image), and d represents the water depth of the identified fish at the time of recognition;

[0079] The pixel coordinates in the fish school sonar image are converted into x, y in the actual three-dimensional coordinates by using the conversion method in the prior art, for example:

[0080] See also Figure 4, the fish automatic identification model uses a rectangular prediction box to frame the position of the identified fish, and the coordinates of the prediction box are in a coordinate system with the upper left corner of the sonar image as the origin, the right as the positive direction of the x-axis, and the downward as the positive direction of the y-axis. The lengths of the x and y coordinate axes are both [0,1], which is the ratio of the number of pixels in the prediction box to the length and width of the sonar image. Therefore, it is necessary to convert according to the size of the image generated by the sonar and the actual distance measured in reality, so that when processing the coordinates later, they can be assigned according to the actual length of the experimental fish. At the same time, it can also meet certain conditional constraints when clustering coordinate points taking into account realistic factors. Here, we adjust the two-dimensional coordinate system, converting the coordinate system with the upper left corner of the sonar image as the origin to the coordinate system with the center point of the sonar image as the origin. The directions of the x and y axes remain unchanged. The conversion formula is as follows:

[0081]

[0082] Where: x, y are the coordinates of the coordinate system with the upper left corner as the origin, X, Y are the coordinates of the coordinate system with the center point as the origin, and x, y, X, Y are the ratios of the image size to the actual size.

[0083] The pixel size of the sonar image used in this experiment is 804×604, and the range radius used by the sonar is 300 cm. Therefore, the formula for converting the coordinates in the image into the coordinate size in reality is:

[0084]

[0085] Where: x, y are the coordinates of the coordinate axis with the center point of the sonar image as the origin after the above processing, and X, Y are the actual coordinate data corresponding to x, y in reality.

[0086] For the depth data, since the sonar in our experiment can be regarded as a uniform downward movement from the water surface, i.e., the starting depth is 0 cm, the sonar speed v per second can be calculated based on the sonar detection time T and the detection depth H. Then, based on the correspondence between the total number of frames P of the sonar video image and the sonar time T, the number of seconds per frame during sonar measurement can be obtained. The unit is s, i.e., the inverse of the frame rate. Then, the depth h corresponding to each frame during sonar measurement can be calculated based on the sonar movement speed and the time length of each sonar frame. The calculation formula is as follows:

[0087]

[0088] Step S4, starting from the water surface (i.e. d=0), multiple water layer data are set according to a preset water depth interval △d, and two adjacent water layer data have a preset water depth overlap interval δd, that is: dmax-dmin=△d, d1max-d2min=δd, dmax and dmin respectively represent the maximum water depth and the minimum water depth of any water layer data, d1max and d2min respectively represent the maximum water depth of the previous water layer data and the minimum water depth of the next water layer data in two adjacent water layer data; because it is considered that during measurement, fish may run to different water layers and generate similar signal data in different water layers, so ensuring that there is an overlap between two adjacent water layer data can minimize the influence of the coordinates of the center point fish after clustering; and because the subsequent step is clustering, even if the counting is repeated, it will not have much impact on the clustering result.

[0089] Furthermore, according to the water depth d, the actual three-dimensional coordinates (x, y, d) of each identified fish obtained in step S3 are respectively attributed to the corresponding water layer data, wherein the actual three-dimensional coordinates (x, y, d) within the overlapping interval δd of the water depth are attributed to both the upper and lower water layer data;

[0090] Wherein, the beam opening angle range of the scanning beam 1a of the omnidirectional sonar 1 is [a°, b°];

[0091] In step S4, the water depth interval Δd is calculated according to the following formula:

[0092] △d=(△d1+△d2) / 2;

[0093] △d1=2*tan(a / 2)*R;

[0094] △d2=2*tan(b / 2)*R;

[0095] Wherein, △d1 and △d2 are the water depth intervals when the beam opening angles are a° and b° respectively; R is the radius of the measured cage 2.

[0096] Thus, the overall scanning of the measured network box 2 by the omnidirectional sonar 1 can be realized more quickly and effectively.

[0097] Wherein, in step S4, the water depth overlap interval δd is taken as the smaller value between the average fish length and the average fish width of the identified fish.

[0098] Thus, it is possible to avoid repeated counting of fish in the two adjacent water layer data due to fish wandering within the overlapping interval δd of the water depth between the two adjacent water layer data.

[0099] Step S5: for each of the water layer data, density clustering processing is performed on the actual three-dimensional coordinates (x, y, d) belonging to the water layer data to obtain a cluster of each water layer data; wherein each cluster can be identified as a fish in the corresponding water layer data range, and a cluster containing only one actual three-dimensional coordinate (x, y, d) is recorded as a noise point cluster, and a cluster containing at least two actual three-dimensional coordinates (x, y, d) is recorded as a core object cluster;

[0100] For example: Figure 5 This is the spatial distribution map of 7681 actual three-dimensional coordinates (x, y, d) contained in a certain water layer data. Figure 6 for Figure 5 The spatial distribution diagram of 17 clusters obtained after the water layer data is processed in step S5 is shown, wherein the 17 clusters, namely Cu0 to Cu16, are marked with different colors respectively.

[0101] In step S5, the DBSCAN clustering analysis method is used to perform density clustering processing on the actual three-dimensional coordinates (x, y, d) belonging to the same water layer data; wherein the clustering parameters are set in the following manner:

[0102] The clustering radius (epsilon) takes values ​​in the range of 0.5N to 0.8N, where N is the larger value of the average fish length and the average fish width of the identified fish; when the clustering radius is larger within this range, the number of clusters obtained by clustering will decrease, but considering the movement of fish in the water and the position changes between fish, the error of the number after clustering is small; the increase in the clustering radius will cause the offset of the center point of the fish, but it will have little effect on the spatial distribution characteristics of the overall fish in the cage.

[0103] The minimum number of samples (min_samples) is set to 1, which means that the cluster obtained by clustering can contain at least one sample, that is, the actual three-dimensional coordinates (x, y, d).

[0104] Step S6, calculate the actual three-dimensional coordinates (x', y', d') of each fish in the tested cage 2:

[0105] For the noise point cluster, the only actual three-dimensional coordinate (x, y, d) contained in it is used as the actual three-dimensional coordinate (x', y', d') of the fish in the cage corresponding to the noise point cluster;

[0106] For the core object cluster, calculate the center point coordinates of all the actual three-dimensional coordinates (x, y, d) contained in it, as the actual three-dimensional coordinates (x', y', d') of the fish in the cage corresponding to the core object cluster; that is, x' is the average value of all the aforementioned x, y' is the average value of all the aforementioned y, and d' is the average value of all the aforementioned d;

[0107] Among them, x', y' and d' represent the abscissa of the actual horizontal plane of the fish in the cage, the ordinate of the actual horizontal plane of the fish and the actual water depth of the fish, respectively;

[0108] When the water depth overlap interval δd is set for the data of two adjacent water layers and δd≠0, the coordinates of the center point can be calculated more accurately after clustering.

[0109] Step S7: Estimate the spatial distribution state of the fish in the tested cage 2 according to the actual three-dimensional coordinates (x', y', d') of the fish obtained in step S6.

[0110] Preferably, in step S7, the spatial distribution state estimation includes: drawing a horizontal distribution diagram of the fish in the tested cage 2 according to the actual horizontal plane abscissa x' and the actual horizontal plane ordinate y' of the fish obtained in step S6, see Figure 7 .

[0111] Preferably, in step S7, the spatial distribution state estimation includes: drawing a spatial distribution map of the caged fish in the water depth range corresponding to each water layer data of the tested cage 2 according to the actual three-dimensional coordinates (x', y', d') of the fish obtained in step S6, see Figure 8 .

[0112] Preferably, in step S7, the spatial distribution state estimation includes: drawing a spatial distribution map of the fish in the tested cage 2 according to the actual three-dimensional coordinates (x', y', d') of the fish obtained in step S6, see Fig. 9 .

[0113] The present invention is not limited to the above-mentioned specific implementation modes. According to the above-mentioned contents, in accordance with the common technical knowledge and customary means in the field, without departing from the above-mentioned basic technical ideas of the present invention, the present invention can also make other various forms of equivalent modifications, replacements or changes, all of which fall within the protection scope of the present invention.

Claims

1. A method for estimating the spatial distribution state of caged fish based on omnidirectional sonar, characterized in that: include: Step S1, placing an omnidirectional sonar (1) in a tested cage (2), and driving the omnidirectional sonar (1) to move at a constant speed between the water surface (2a) and the cage bottom surface (2b) in a vertical direction, so as to perform an overall scan of the tested cage (2) through the omnidirectional sonar (1) during the movement; Furthermore, the data outputted by the omnidirectional sonar (1) after an overall scan is played back through the supporting software of the omnidirectional sonar (1) to be converted into a video of the overall sonar scan of the net cage, and the video frames of the video of the overall sonar scan of the net cage are recorded as the sonar image of the school of fish; Step S2, performing multiple overall scans on the tested cage (2) according to step S1, obtaining a sufficient number of cage overall sonar scanning videos as training data, so as to obtain a fish automatic identification model through deep learning model training, wherein the fish automatic identification model can identify the position of each identified fish in each frame of the fish school sonar image of the cage overall sonar scanning video; Step S3, at the target time, perform an overall scan of the tested cage (2) according to step S1 to obtain an overall sonar scanning video of the target cage, and use the fish automatic recognition model to identify the position of each identified fish in each frame of the fish school sonar image of the overall sonar scanning video of the target cage; Furthermore, the actual three-dimensional coordinates (x, y, d) of each identified fish in each frame of the fish school sonar image of the overall sonar scanning video of the target cage are extracted and converted from the identification results of the fish automatic identification model, wherein x and y represent the horizontal plane abscissa and horizontal plane ordinate of the identified fish at the time of identification, respectively, and d represents the water depth of the identified fish at the time of identification; Step S4, starting from the water surface, multiple water layer data are set according to a preset water depth interval △d, and two adjacent water layer data have a preset water depth overlap interval δd, that is: dmax-dmin=△d, d1max-d2min=δd, dmax and dmin respectively represent the maximum water depth and minimum water depth of any water layer data, d1max and d2min respectively represent the maximum water depth of the previous water layer data and the minimum water depth of the next water layer data in two adjacent water layer data; Furthermore, according to the water depth d, the actual three-dimensional coordinates (x, y, d) of each identified fish obtained in step S3 are respectively attributed to the corresponding water layer data; Step S5: for each of the water layer data, density clustering processing is performed on the actual three-dimensional coordinates (x, y, d) belonging to the water layer data to obtain a cluster of each water layer data; wherein a cluster containing only one actual three-dimensional coordinate (x, y, d) is recorded as a noise point cluster, and a cluster containing at least two actual three-dimensional coordinates (x, y, d) is recorded as a core object cluster; Step S6, calculating the actual three-dimensional coordinates (x', y', d') of each fish in the tested cage (2): For the noise point cluster, the only actual three-dimensional coordinate (x, y, d) contained in it is used as the actual three-dimensional coordinate (x', y', d') of the fish in the cage corresponding to the noise point cluster; For the core object cluster, calculate the center point coordinates of all the actual three-dimensional coordinates (x, y, d) contained in it, as the actual three-dimensional coordinates (x', y', d') of the fish in the cage corresponding to the core object cluster; Among them, x', y' and d' represent the abscissa of the actual horizontal plane of the fish in the cage, the ordinate of the actual horizontal plane of the fish and the actual water depth of the fish, respectively; Step S7: Estimate the spatial distribution state of the fish in the tested cage (2) based on the actual three-dimensional coordinates (x', y', d') of the fish obtained in step S6.

2. The method for estimating the spatial distribution state of caged fish based on omnidirectional sonar according to claim 1, characterized in that: In step S1, the omnidirectional sonar (1) is placed at the central axis position of the tested cage (2) and is driven by a vertical lifting drive device to move at a uniform speed between the water surface (2a) and the bottom surface (2b) of the cage in a vertical direction.

3. The method for estimating the spatial distribution state of caged fish based on omnidirectional sonar according to claim 2, characterized in that: The beam opening angle range of the scanning beam (1a) of the omnidirectional sonar (1) is [a°, b°]; In step S4, the water depth interval Δd is calculated according to the following formula: △d=(△d1+△d2) / 2; △d1=2*tan(a / 2)*R; △d2=2*tan(b / 2)*R; Wherein, △d1 and △d2 are the water depth intervals when the beam opening angles are a° and b° respectively; R is the radius of the measured cage (2).

4. The method for estimating the spatial distribution state of caged fish based on omnidirectional sonar according to claim 2, characterized in that: In step S4, the water depth overlap interval δd is taken as the smaller value between the average fish length and the average fish width of the identified fish.

5. The method for estimating the spatial distribution state of caged fish based on omnidirectional sonar according to claim 1, characterized in that: In step S5, the DBSCAN clustering analysis method is used to perform density clustering processing on the actual three-dimensional coordinates (x, y, d) belonging to the same water layer data; wherein the clustering parameters are set in the following manner: The cluster radius ranges from 0.5N to 0.8N, where N is the larger value of the average fish length and average fish width of the identified fish; The minimum number of samples is set to 1.

6. The method for estimating the spatial distribution state of caged fish based on omnidirectional sonar according to any one of claims 1 to 5, characterized in that: In step S2, the deep learning model adopts the YOLOv8 model, and the fish automatic identification model is trained in the following manner: Step S2-1, intercepting a sufficient number of fish sonar images from the overall sonar scanning video of the cage, and annotating the position of each identified fish in each frame of the fish sonar image using the MakeSense online data annotation software to output an annotation result file; Step S2-2: using the whole cage sonar scanning video as input data, using the annotation result file corresponding to the whole cage sonar scanning video as output data, and training the YOLOv8 model to obtain the fish automatic identification model.

7. The method for estimating the spatial distribution state of caged fish based on omnidirectional sonar according to any one of claims 1 to 5, characterized in that: In the step S7, the spatial distribution state estimation includes: drawing a horizontal distribution diagram of the fish in the tested cage (2) according to the actual horizontal plane abscissa x' and the actual horizontal plane ordinate y' of the fish obtained in the step S6.

8. The method for estimating the spatial distribution state of caged fish based on omnidirectional sonar according to any one of claims 1 to 5, characterized in that: In step S7, the spatial distribution state estimation includes: drawing a spatial distribution map of the caged fish in the tested cage (2) respectively in the water depth range corresponding to each water layer data according to the actual three-dimensional coordinates (x', y', d') of the fish obtained in step S6.

9. The method for estimating the spatial distribution state of caged fish based on omnidirectional sonar according to any one of claims 1 to 5, characterized in that: In the step S7, the spatial distribution state estimation includes: drawing a spatial distribution map of the fish in the tested cage (2) according to the actual three-dimensional coordinates (x', y', d') of the fish obtained in step S6.

Images