A method and system for estimating the weight of live fish

Abstract

The present application belongs to the technical field of fish body weight detection, and particularly relates to a live fish body weight estimation method and system. The method comprises the following steps: S1, obtaining a fish body image; S2, identifying fish body color information and fish body key point information of the live fish based on the fish body image; S3, estimating the species of the live fish based on the fish body color information, the fish body key point information and a live fish species estimation model; and S4, estimating the body weight of the live fish based on the live fish species, the fish body key point information and a live fish body weight estimation model. The present application combines color information and fish body length-width ratio information to estimate the species, and further estimates the body weight through fish body species information and fish body size information, so that the fish body weight can be intelligently and accurately estimated.

Description

Technical Field

[0001] This invention belongs to the field of fish quality detection technology, specifically relating to a method and system for estimating the weight of live fish. Background Technology

[0002] China has ranked first in the world in aquatic product output for 32 consecutive years since 1989, making it the world's largest fishing nation. In 2020, aquatic product output reached 65.49 million tons, providing nearly one-third of the high-quality animal protein for urban and rural residents, playing a vital role in ensuring national food security and improving the nutritional health of the entire population. With the continuous improvement of people's living standards, the proportion of aquatic products in the dietary structure will continue to increase, and the demand for aquatic products will show a sustained growth trend.

[0003] Currently, in pond aquaculture, the weight of fish is generally determined manually based on experience to determine whether they have reached the saleable size. Fish that do not meet the saleable size need to be returned to the pond for further rearing, and fish that have reached the saleable size are also divided into several sales grades. However, the judgment process can be prone to errors due to insufficient human experience and a high degree of randomness, leading to fish being assigned to the wrong grade and causing unnecessary economic losses. Relying on manual experience to determine fish weight is time-consuming, labor-intensive, inefficient, and inaccurate. Furthermore, the number of experienced workers has decreased significantly in recent years, leading to a shortage of available personnel. Therefore, there is an urgent need to find new methods for measuring live fish weight. This paper proposes a method for estimating live fish weight to address the aforementioned problems. Summary of the Invention

[0004] To address the aforementioned problems in the existing technology, this invention provides a method and system for estimating the weight of live fish. This invention can combine color information and fish length-to-width ratio information to estimate the species, and further estimate the weight by combining fish species information and fish size information, which can more intelligently and accurately estimate the weight of fish.

[0005] The present invention adopts the following technical solution:

[0006] The first aspect of this invention provides a method for estimating the weight of live fish, comprising the following steps:

[0007] S1. Obtain the image of the fish;

[0008] S2. Obtain the color information and key point information of the live fish based on fish image recognition;

[0009] S3. Based on fish body color information, fish body key point information, and live fish species estimation model, the species of live fish is estimated.

[0010] S4. Based on the species of live fish, key information of the fish body, and the live fish weight estimation model, the weight of the live fish is estimated.

[0011] As a preferred embodiment, step S3 includes the following steps:

[0012] S3.1 Calculate the fish body length and width based on the key point information of the fish body;

[0013] S3.2. Based on the fish body color information, fish body length, fish body width, and live fish species estimation model, the live fish species are estimated.

[0014] As a preferred embodiment, step S3.2 includes the following steps:

[0015] S3.2.1 Calculate the ratio of the fish's body length to its body width;

[0016] S3.2.2. Based on the fish color information, the ratio of fish body length to fish body width, and the live fish species estimation model, the live fish species are estimated.

[0017] As a preferred embodiment, step S4 includes the following steps:

[0018] S4.1 Calculate the fish body height based on the key point information of the fish body;

[0019] S4.2. Based on the live fish species, fish body length, fish body width, fish body height, and live fish weight estimation model, the weight of the live fish is estimated.

[0020] As a preferred embodiment, the fish images in step S1 include a front view image of the fish, a top view image of the fish, and a side view image of the fish.

[0021] As a preferred method, the fish's length, width, and height are calculated based on key point information, including the following steps:

[0022] A. Identify key points of the fish in the front view, top view, and side view images of the fish, respectively.

[0023] B. Calculate the first initial body length and the first initial body width based on the key point information of the fish body in the top view image; calculate the second initial body length and the first initial body height based on the key point information of the fish body in the front view image; calculate the second initial body height and the second initial body width based on the key point information of the fish body in the side view image.

[0024] C. Based on the first initial body length, the second initial body length, the first initial body width, the second initial body width, the first initial body height, and the second initial body height, the fish body length, fish body width, and fish body height are calculated.

[0025] As a preferred option, step C specifically includes:

[0026] Calculate the average of the first initial body length and the second initial body length to determine the fish's body length;

[0027] Calculate the average of the first initial body width and the second initial body width to be used as the fish body width;

[0028] Calculate the average of the first initial body height and the second initial body height to be used as the fish's body height.

[0029] As a preferred embodiment, in step S3.2, the fish color information includes the fish color information in the front view image, the fish color information in the top view image, and the fish color information in the side view image.

[0030] As a preferred option, step S3.2.1 specifically includes:

[0031] The first initial ratio is calculated based on the first initial body length and the first initial body width;

[0032] The second initial ratio is calculated based on the second initial body length and the second initial body width.

[0033] Based on the first initial ratio and the second initial ratio, the ratio of the fish's body length to its body width is calculated.

[0034] The second aspect of this invention provides a live fish weight estimation system, based on a live fish weight estimation method provided in the first aspect of the invention, comprising a fish body image acquisition module, a fish body information recognition module, a species estimation module, and a weight estimation module connected in sequence, wherein the fish body information recognition module is also connected to the weight estimation module;

[0035] Fish image acquisition module, used to acquire fish images;

[0036] The fish body information recognition module is used to identify the color information and key point information of live fish.

[0037] The species estimation module estimates the species of live fish based on fish color information, key point information of the fish body, and a live fish species estimation model.

[0038] The weight estimation module estimates the weight of live fish based on the species of live fish, key information about the fish body, and a live fish weight estimation model.

[0039] The beneficial effects of this invention are:

[0040] This invention obtains the color information and key point information of live fish based on fish image recognition. Then, based on the color information, key point information, and a live fish species estimation model, the species of the live fish is estimated. Finally, based on the species, key point information, and weight estimation model, the weight of the live fish is estimated. In other words, this invention combines color information and fish aspect ratio information for species estimation, and further uses fish species information and fish size information for weight estimation, which can more intelligently and accurately estimate fish weight.

[0041] In this invention, the fish images include a front view image, a top view image, and a side view image. The fish length, width, and height information are calculated based on key point information in the three views, rather than from a single view. Therefore, the final estimation results are more accurate.

[0042] In this invention, the fish color information includes the color information from the front view image, the top view image, and the side view image, rather than being obtained from a single view. Therefore, the final estimation result is more accurate. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 This is a flowchart of a method for estimating the weight of live fish according to an embodiment of the present invention.

[0045] Figure 2 This is a schematic diagram of a live fish weight estimation system according to an embodiment of the present invention. Detailed Implementation

[0046] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.

[0047] Example 1:

[0048] Reference Figure 1As shown, this embodiment provides a method for estimating the weight of live fish, including the following steps:

[0049] S1. Obtain the image of the fish;

[0050] S2. Obtain the color information and key point information of the live fish based on fish image recognition;

[0051] S3. Based on fish body color information, fish body key point information, and live fish species estimation model, the species of live fish is estimated.

[0052] S4. Based on the species of live fish, key information of the fish body, and the live fish weight estimation model, the weight of the live fish is estimated.

[0053] As can be seen, in this embodiment, the color information and key point information of the live fish are obtained based on fish image recognition. Based on the color information, key point information, and live fish species estimation model, the species of the live fish is estimated. Finally, based on the species of the live fish, key point information, and live fish weight estimation model, the weight of the live fish is estimated. That is, the present invention combines color information and fish aspect ratio information for species estimation, and further uses fish species information and fish size information for weight estimation, which can more intelligently and accurately estimate the weight of the fish.

[0054] Specifically:

[0055] In step S1, the fish images include a front view image, a top view image, and a side view image.

[0056] Step S3 includes the following steps:

[0057] S3.1 Calculate the fish body length and width based on the key point information of the fish body;

[0058] S3.2. Based on the fish body color information, fish body length, fish body width, and live fish species estimation model, the live fish species are estimated.

[0059] Step S3.2 includes the following steps:

[0060] S3.2.1 Calculate the ratio of the fish's body length to its body width;

[0061] S3.2.2. Based on the fish color information, the ratio of fish body length to fish body width, and the live fish species estimation model, the live fish species are estimated.

[0062] In step S3.2, the fish color information includes the fish color information in the front view image, the fish color information in the top view image, and the fish color information in the side view image.

[0063] Step S4 includes the following steps:

[0064] S4.1 Calculate the fish body height based on the key point information of the fish body;

[0065] S4.2. Based on the live fish species, fish body length, fish body width, fish body height, and live fish weight estimation model, the weight of the live fish is estimated.

[0066] More specifically:

[0067] The fish's length, width, and height are calculated based on key point information, including the following steps:

[0068] A. Identify key points of the fish in the front view, top view, and side view images of the fish, respectively.

[0069] B. Calculate the first initial body length and the first initial body width based on the key point information of the fish body in the top view image; calculate the second initial body length and the first initial body height based on the key point information of the fish body in the front view image; calculate the second initial body height and the second initial body width based on the key point information of the fish body in the side view image.

[0070] C. Based on the first initial body length, the second initial body length, the first initial body width, the second initial body width, the first initial body height, and the second initial body height, the fish body length, fish body width, and fish body height are calculated.

[0071] Step C specifically includes:

[0072] Calculate the average of the first initial body length and the second initial body length to determine the fish's body length;

[0073] Calculate the average of the first initial body width and the second initial body width to be used as the fish body width;

[0074] Calculate the average of the first initial body height and the second initial body height to be used as the fish's body height.

[0075] In step S3.2.1, specifically:

[0076] The first initial ratio is calculated based on the first initial body length and the first initial body width;

[0077] The second initial ratio is calculated based on the second initial body length and the second initial body width.

[0078] Based on the first initial ratio and the second initial ratio, the ratio of the fish's body length to its body width is calculated.

[0079] As can be seen, the fish images in this embodiment include a front view image, a top view image, and a side view image. The fish length, width, and height information are calculated based on key point information in the three views, rather than from a single view. Therefore, the final estimation result is more accurate.

[0080] Furthermore, in this embodiment, the fish color information includes the fish color information in the front view image, the fish color information in the top view image, and the fish color information in the side view image. Instead of obtaining it from a single view, the final estimation result is more accurate.

[0081] More specifically:

[0082] In step S1, after acquiring the fish image, the acquired fish image needs to undergo preprocessing such as extracting color component values, grayscale conversion, binarization, image filtering, and image cropping.

[0083] In step S2, the preprocessed image is input into the ResNet50 feature extraction network in the Mask R-CNN segmentation network. The deep learning-based instance segmentation Mask R-CNN algorithm separates the fish outline from the image. Since the fish body occupies most of the information in the image, the RPN increases the image segmentation and detection analysis of large targets in the multi-scale analysis based on feature maps, changes the receptive field range of the model on the feature map, and optimizes Mask R-CNN by adjusting the receptive field range of the backbone network ResNet50 and the RPN in the algorithm, thereby improving the detection accuracy of the model.

[0084] Fish color and body shape features derived from key points are two of the most important visual features of an image. Color, an attribute of an object's surface, is the most direct visual feature describing image content and a crucial piece of information for distinguishing different objects. Body shape, on the other hand, is a physical attribute of a fish. By extracting and analyzing the color and body shape features of fish images, the aim is to establish a fish species recognition model to identify fish species.

[0085] The keypoint detection algorithm is built upon the Keypoint-RCNN algorithm, which adds a keypoint detection branch to Mask-RCNN. The body length, height, and width data are calculated based on the spacing between keypoints. Specifically, body length is the vertical distance between two keypoints from the snout to the base of the caudal fin; body height is the vertical distance between two keypoints from the highest to the lowest point of the body; and body width is the distance between two keypoints on the left and right sides of the body.

[0086] The weight estimation model described in step 6 is built based on a Stacking feature extraction network. The Stacking algorithm is a hierarchical ensemble method capable of integrating heterogeneous models. During the training of the Stacking ensemble model, k-fold cross-validation is typically used to partition the dataset and train the model to reduce the risk of overfitting. The Stacking model has strict requirements for the primary trainer, demanding high diversity and accuracy. After screening, a suitable model is selected as the primary trainer for the Stacking model, thus forming the initial structure of the Stacking model. After obtaining the original feature dataset, data cleaning and normalization are performed, and appropriate model evaluation metrics are selected. Then, the model is trained simply, and network search and cross-validation methods are used to optimize the model parameters. Specifically, the fish species, body length, body height, and body width are used as inputs, and the weight value is used as the output. The training data is then divided into training and test sets in a 7:3 ratio, and the model is trained using 5-fold cross-validation, ultimately obtaining the Stacking model structure. The trained weight estimation model estimates the fish's weight and outputs the weight.

[0087] Example 2:

[0088] Reference Figure 2 As shown, this embodiment provides a live fish weight estimation system, based on the live fish weight estimation method described in Embodiment 1, including a fish body image acquisition module, a fish body information recognition module, a species estimation module, and a weight estimation module connected in sequence. The fish body information recognition module is also connected to the weight estimation module.

[0089] Fish image acquisition module, used to acquire fish images;

[0090] The fish body information recognition module is used to identify the color information and key point information of live fish.

[0091] The species estimation module estimates the species of live fish based on fish color information, key point information of the fish body, and a live fish species estimation model.

[0092] The weight estimation module estimates the weight of live fish based on the species of live fish, key information about the fish body, and a live fish weight estimation model.

[0093] That is, the fish color information and key point information of the live fish are obtained based on fish image recognition. Based on the fish color information, key point information and live fish species estimation model, the species of the live fish is estimated. Finally, based on the live fish species, key point information and live fish weight estimation model, the weight of the live fish is estimated. In other words, the present invention combines color information and fish length-to-width ratio information for species estimation, and further uses fish species information and fish size information for weight estimation, which can more intelligently and accurately estimate fish weight.

[0094] It should be noted that the live fish weight estimation system provided in this embodiment is similar to that in Embodiment 1, and will not be described in detail here.

[0095] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope of the present invention.

Claims

1. A method for estimating the weight of live fish, characterized in that, Including the following steps: S1. Obtain the image of the fish; S2. Obtain the color information and key point information of the live fish based on fish image recognition; S3. Based on fish body color information, fish body key point information, and live fish species estimation model, the species of live fish is estimated. S4. Based on the species of live fish, key information of the fish body, and the live fish weight estimation model, the weight of the live fish is estimated. Step S3 includes the following steps: S3.1 Calculate the fish body length and width based on the key point information of the fish body; S3.

2. Based on fish color information, fish length, fish width, and live fish species estimation model, the live fish species are estimated. Step S3.2 includes the following steps: S3.2.1 Calculate the ratio of the fish's body length to its body width; S3.2.

2. Based on fish color information, the ratio of fish length to fish width, and a live fish species estimation model, the species of live fish is estimated. Step S4 includes the following steps: S4.1 Calculate the fish body height based on the key point information of the fish body; S4.

2. Based on the live fish species, fish body length, fish body width, fish body height and live fish weight estimation model, the weight of the live fish is estimated. The fish images in step S1 include a front view image, a top view image, and a side view image. The fish's length, width, and height are calculated based on key point information, including the following steps: A. Identify key points of the fish in the front view, top view, and side view images of the fish, respectively. B. Calculate the first initial body length and the first initial body width based on the key point information of the fish body in the top view image; calculate the second initial body length and the first initial body height based on the key point information of the fish body in the front view image; calculate the second initial body height and the second initial body width based on the key point information of the fish body in the side view image. C. Based on the first initial body length, the second initial body length, the first initial body width, the second initial body width, the first initial body height, and the second initial body height, the fish body length, fish body width, and fish body height are calculated.

2. The method for estimating the weight of live fish according to claim 1, characterized in that, Step C specifically includes: Calculate the average of the first initial body length and the second initial body length to determine the fish's body length; Calculate the average of the first initial body width and the second initial body width to be used as the fish body width; Calculate the average of the first initial body height and the second initial body height to be used as the fish's body height.

3. The method for estimating the weight of live fish according to claim 1, characterized in that, In step S3.2, the fish color information includes the fish color information in the front view image, the fish color information in the top view image, and the fish color information in the side view image.

4. The method for estimating the weight of live fish according to claim 1, characterized in that, In step S3.2.1, specifically: The first initial ratio is calculated based on the first initial body length and the first initial body width; The second initial ratio is calculated based on the second initial body length and the second initial body width. Based on the first initial ratio and the second initial ratio, the ratio of the fish's body length to its body width is calculated.

5. A live fish weight estimation system, based on the live fish weight estimation method according to any one of claims 1-4, characterized in that, It includes a fish image acquisition module, a fish information recognition module, a species estimation module, and a weight estimation module connected in sequence. The fish information recognition module is also connected to the weight estimation module. Fish image acquisition module, used to acquire fish images; The fish body information recognition module is used to identify the color information and key point information of live fish. The species estimation module estimates the species of live fish based on fish color information, key point information of the fish body, and a live fish species estimation model. The weight estimation module estimates the weight of live fish based on the species of live fish, key information about the fish body, and a live fish weight estimation model.

Images