Fish habitat monitoring method, device and equipment, storage medium and product

Abstract

The invention provides a fish habitat monitoring method, device and equipment, a storage medium and a product, and relates to the technical field of fish habitat assessment. The method comprises the steps of determining a to-be-monitored target area according to habitat information of a to-be-monitored target fish, generating a target navigation route of the aerial robot, controlling the aerial robot to fly along the target navigation route, collecting multiple pieces of image data shot in the flight process, and splicing the multiple pieces of image data to obtain a spliced image; the method comprises the steps of obtaining an ortho-image map, performing image preprocessing, performing feature recognition on the preprocessed ortho-image map by using a convolutional neural network algorithm to obtain a feature region, and performing discrimination calculation on the feature region according to a calculation index corresponding to a to-be-monitored target fish so as to obtain an evaluation result of the ecological health condition of the target region. According to the method, the fish habitat monitoring efficiency is improved.

Description

Technical Field

[0001] The present application relates to the technical field of fish habitat assessment, and in particular to a method, device, equipment, storage medium and product for monitoring fish habitat. Background Art

[0002] Monitoring and assessment of fish habitats play an important role in environmental protection and ecological restoration. Traditional monitoring and assessment of fish habitats is mainly completed through manual ground surveys, which usually require a lot of manpower and time to obtain data on coastal zone types and river characteristics of fish habitats. In vast or inaccessible fish habitats, it is difficult to obtain comprehensive and accurate fish habitat data.

[0003] At present, fish habitat monitoring is mainly carried out through satellite remote sensing technology, or using drones to collect images and manually analyze image data. Among them, image data analysis has been replaced by image recognition technology in recent years, mainly achieved through computer vision and deep learning technology, such as image classification and feature extraction through convolutional neural networks.

[0004] In the existing technology, there is mainly a technical problem of low efficiency in fish habitat monitoring. Summary of the invention

[0005] The present application provides a method, device and storage for monitoring fish habitats, so as to solve the problem of low efficiency of monitoring fish habitats in the prior art.

[0006] In a first aspect, the present application provides a method for monitoring fish habitats, comprising:

[0007] According to the habitat information of the target fish to be monitored, the target area to be monitored is determined, and the target navigation route of the aerial robot is generated;

[0008] Control the aerial robot to fly along the target navigation route and collect multiple image data taken by the aerial robot during the flight;

[0009] Splice multiple image data to obtain an orthophoto map;

[0010] Performing image preprocessing on the orthophoto map to obtain a preprocessed image map;

[0011] Using a convolutional neural network algorithm, feature recognition is performed on the preprocessed image to obtain at least one feature region;

[0012] According to the calculation index corresponding to the target fish to be monitored, a discrimination calculation is performed on at least one characteristic area to obtain an assessment result of the ecological health status of the target area.

[0013] In a possible design, a convolutional neural network algorithm is used to perform feature recognition on the preprocessed image to obtain at least one feature area, including:

[0014] Normalizing the preprocessed image to obtain a normalized image;

[0015] Inputting the normalized image map into the trained improved convolutional neural network model to obtain a feature map through the output result of the trained improved convolutional neural network model;

[0016] According to a preset discrimination criterion, the feature map is classified and identified to obtain at least one feature region.

[0017] In a possible design, before the normalized image map is input into the trained improved convolutional neural network model to obtain the feature map through the output result of the trained improved convolutional neural network model, the method further includes:

[0018] Acquire a feature data set, wherein the feature data set includes a plurality of pre-labeled normalized image sample images;

[0019] The feature data set is used to train the initial improved convolutional neural network model, and the model parameters of the improved convolutional neural network model are optimized through the back propagation algorithm and the gradient descent method to obtain the trained improved convolutional neural network model.

[0020] In one possible design, the improved convolutional neural network model includes an input layer, a first convolutional layer, a spatial pyramid pooling layer, an improved residual block, a feature fusion layer, a fully connected layer, and an output layer;

[0021] The spatial pyramid pooling layer is used to capture multi-scale information;

[0022] The improved residual block includes multiple parallel convolutional layers to capture deep features.

[0023] In a possible design, the calculation indicators include the coastal zone type calculation indicator, the boulder bedrock ratio calculation indicator in the river channel, the river channel width calculation indicator and the flow state slowness and speed calculation indicator;

[0024] Accordingly, according to the calculation index corresponding to the target fish species to be monitored, a discrimination calculation is performed on at least one characteristic area to obtain an assessment result of the ecological health status of the target area, including:

[0025] According to the calculation index of coastal zone type, the proportion of boulder bedrock in the river channel, the river channel width and the flow rate, the proportion of coastal zone type, the proportion of boulder bedrock in the river channel, the river channel width coefficient and the flow rate coefficient of each characteristic area are calculated respectively;

[0026] According to the weights corresponding to the coastal zone type calculation index, river channel width calculation index and flow state speed calculation index, combined with the proportion of coastal zone types in each characteristic area, the proportion of boulder bedrock in the river channel, the river channel width coefficient and the flow state speed coefficient, the comprehensive score of each characteristic area is calculated;

[0027] Based on the comprehensive score of each characteristic area, the overall score of the ecological health status of the target area is calculated.

[0028] In a possible design, the target area to be monitored is determined based on the habitat information of the target fish to be monitored, and a target navigation route of the aerial robot is generated, including:

[0029] Identify target fish species to be monitored;

[0030] Obtaining habitat information of target fish, wherein the habitat information includes geographical location information, environmental condition information and biological characteristic information;

[0031] Determine the corresponding feature discrimination criteria based on habitat information;

[0032] Based on habitat information and feature identification criteria, geographic information system technology is used to generate feature collection maps;

[0033] Based on the feature collection graph, the target navigation route of the aerial robot is generated.

[0034] In a possible design, multiple image data are stitched together to obtain an orthophoto map, including:

[0035] Perform registration processing on multiple image data, identify common feature points in adjacent image data through feature point matching algorithm, and calculate the transformation matrix between image data;

[0036] According to the transformation matrix, adjacent image data are spliced ​​to obtain a spliced ​​image;

[0037] The stitched image is cropped and geometrically corrected to obtain an orthophoto.

[0038] In a possible design, the orthophoto image is preprocessed to obtain a preprocessed image, including:

[0039] Perform at least one of denoising, enhancement and color correction on the orthophoto to obtain a preprocessed image.

[0040] In a second aspect, the present application provides a monitoring device for fish habitats, comprising:

[0041] A first determination module is used to determine the target area to be monitored according to the habitat information of the target fish to be monitored, and generate a target navigation route of the aerial robot;

[0042] A first control module is used to control the aerial robot to fly along a target navigation route and collect a plurality of image data taken by the aerial robot during the flight;

[0043] A first processing module is used to perform splicing processing on multiple image data to obtain an orthophoto map;

[0044] The second processing module is used to perform image preprocessing on the orthophoto map to obtain a preprocessed image map;

[0045] A feature extraction module, used to perform feature recognition on the preprocessed image using a convolutional neural network algorithm to obtain at least one feature region;

[0046] The third processing module is used to perform discrimination calculation on at least one characteristic area according to the calculation index corresponding to the target fish to be monitored, so as to obtain an assessment result of the ecological health status of the target area.

[0047] In a third aspect, the present application provides a monitoring device for fish habitats, comprising:

[0048] at least one processor;

[0049] and a memory communicatively coupled to the at least one processor;

[0050] The memory stores instructions that can be executed by at least one processor, and the instructions are executed by at least one processor so that the at least one processor can execute the fish habitat monitoring method as described in any one of the first aspects.

[0051] In a fourth aspect, the present application provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the fish habitat monitoring method as described in any one of the first aspects.

[0052] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the fish habitat monitoring method as described in any one of the first aspects.

[0053] The present application provides a method, device, equipment, storage medium and product for monitoring fish habitats. According to the habitat information of the target fish to be monitored, the target area to be monitored is determined, and the target navigation route of the aerial robot is generated, the aerial robot is controlled to fly along the target navigation route, and multiple image data taken during the flight are collected, the multiple image data are spliced, and the orthophoto image is obtained and image preprocessing is performed. The preprocessed image image is feature-recognized using a convolutional neural network algorithm to obtain a feature area, and the feature area is discriminated and calculated according to the calculation index corresponding to the target fish to be monitored, so as to obtain an evaluation result of the ecological health status of the target area. The drone image data acquisition provided by the present application does not require manual intervention, which improves the efficiency of data acquisition. The present application provides an improved convolutional neural network model, which performs automated data processing and efficient and accurate feature recognition on the image image based on the spatial pyramid pooling layer and the improved residual to obtain the feature area. The present application performs a comprehensive and detailed calculation and evaluation of the feature area based on a variety of ecological indicators. The present method is used to monitor fish habitats, solves the above-mentioned technical problems, has the advantages of automation, convenience, high accuracy, high precision, high efficiency and comprehensiveness, and thus improves the efficiency of monitoring fish habitats. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0055] Figure 1 A schematic diagram of the architecture of a fish habitat monitoring system provided in an embodiment of the present application;

[0056] Figure 2 A flow chart of a method for monitoring fish habitat provided in an embodiment of the present application Figure 1 ;

[0057] Figure 3 A flow chart of a method for monitoring fish habitat provided in an embodiment of the present application Figure 2 ;

[0058] Figure 4 A schematic diagram of the structure of an improved convolutional neural network model provided in an embodiment of the present application;

[0059] Figure 5 A flow chart of a method for monitoring fish habitat provided in an embodiment of the present application Figure 3 ;

[0060] Figure 6 A flow chart of a method for monitoring fish habitat provided in an embodiment of the present application Figure 4 ;

[0061] Figure 7 A flow chart of a method for monitoring fish habitat provided in an embodiment of the present application Figure 5 ;

[0062] Figure 8 A flow chart of a method for monitoring fish habitat provided in an embodiment of the present application Figure 6 ;

[0063] Fig. 9 A schematic diagram of the structure of a fish habitat monitoring device provided in an embodiment of the present application;

[0064] Fig.10 A schematic diagram of the structure of a fish habitat monitoring device provided in an embodiment of the present application.

[0065] Reference numerals:

[0066] 11-data acquisition equipment; 12-processing equipment; 13-display equipment;

[0067] 41-input layer; 42-first convolutional layer; 43-spatial pyramid pooling layer; 44-improved residual block; 45-feature fusion layer; 46-fully connected layer; 47-output layer;

[0068] 91-first determination module; 92-first control module; 93-first processing module; 94-second processing module; 95-feature extraction module; 96-third processing module;

[0069] 100 - electronic device; 101 - processor; 102 - memory; 103 - communication component; 104 - bus. DETAILED DESCRIPTION

[0070] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the attached claims, rather than all embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0071] In the embodiments of the present application, words such as "first" and "second" are used to distinguish the same or similar items with substantially the same functions and effects. Those skilled in the art can understand that words such as "first" and "second" do not limit the quantity and execution order, and words such as "first" and "second" do not necessarily limit the difference.

[0072] It should be noted that in the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a concrete way. In the embodiments of the present application, "at least one" refers to one or more, and "more than one" refers to two or more.

[0073] It should be noted that the "at..." in the embodiments of the present application can be the instant when a certain situation occurs, or can be a period of time after a certain situation occurs, and the embodiments of the present application do not specifically limit this. In addition, the monitoring method of a fish habitat provided in the embodiments of the present application is only used as an example, and the monitoring method of a fish habitat can also include more or less content.

[0074] In order to clearly describe the technical solutions of the embodiments of the present application, some terms and technologies involved in the embodiments of the present application are briefly introduced below:

[0075] Geographic Information System (GIS): GIS is a computer system used to input, store, query, analyze and display geographic data. GIS has three main functions: map processing, database and spatial analysis. Figure 1 It can not only display geographic information in a simple way, but also store and manage large amounts of geographic data through database technology, and conduct in-depth analysis and mining of the data through spatial analysis functions.

[0076] Monitoring and assessment of fish habitats play an important role in environmental protection and ecological restoration. Traditional monitoring and assessment of fish habitats is mainly completed through ground survey methods to obtain data on coastal zone types and river characteristics of fish habitats. Ground surveys usually require a lot of manpower and time. In vast or inaccessible fish habitats, manual collection has limitations and it is difficult to obtain comprehensive and accurate fish habitat data. This method is not only inefficient, but also easily affected by human factors, resulting in insufficient data accuracy and consistency.

[0077] To address the above-mentioned drawbacks, existing technologies mainly use satellite remote sensing technology and drones for monitoring.

[0078] Among them, the fish habitat monitoring method based on satellite remote sensing technology has a wide coverage, but its spatial resolution and temporal resolution are low, which cannot meet the needs of fish habitat monitoring with high requirements for details. At the same time, satellite remote sensing data is greatly affected by weather conditions. Clouds and atmospheric conditions will have a significant impact on image quality, thus affecting the accurate calculation of fish habitat characteristics.

[0079] Among them, using drones to collect images and manually analyzing the collected images has improved the efficiency and accuracy of data collection to a certain extent, but there are still some shortcomings. The image data taken during the drone's cruising flight needs to be manually spliced ​​and processed, which is time-consuming and labor-intensive, and gaps and overlaps are prone to occur during the image stitching process, affecting the quality and accuracy of the final image. At the same time, the process of manually analyzing image data is cumbersome and error-prone, especially when processing large-scale data, the efficiency is low, and subjective factors have a greater impact.

[0080] In addition, with the development of computer vision and deep learning technology, image recognition technology has also been applied in the field of environmental monitoring. However, existing image recognition algorithms still have certain limitations in identifying fish habitat features. Although traditional convolutional neural networks perform well in image classification and feature extraction, their recognition accuracy and efficiency still need to be improved when dealing with multi-scale features in complex natural environments. Existing image recognition methods often only focus on the identification of a single indicator, lack a comprehensive assessment of the comprehensive ecological environment, and are difficult to provide scientific and comprehensive assessment results, which affects decisions on ecological protection and management.

[0081] Existing fish habitat monitoring methods have certain limitations in terms of automation, convenience, high accuracy, high precision, high efficiency and comprehensiveness, all of which lead to reduced efficiency in fish habitat monitoring.

[0082] Based on this, in order to solve the above-mentioned technical problems, the embodiments of the present application provide a method, device, equipment, storage medium and product for monitoring fish habitats, which can be used in the field of fish habitat assessment technology. The inventive concept of the present application is to collect image data by cruising over the target fish habitat with an unmanned aerial vehicle, obtain high-resolution image data, and then automatically process and identify the collected images through a convolutional neural network model, thereby avoiding the influence of human intervention. For the convolutional neural network model, a spatial pyramid pooling layer is introduced to enhance the recognition ability of multi-scale features in complex natural environments, and an improved residual block is introduced to improve the accuracy and efficiency of feature extraction through parallel convolutional layers and attention mechanisms. At the same time, the single coastal zone type calculation is changed to a variety of ecological indicators to conduct a comprehensive assessment of the fish habitat, and a scientific and comprehensive assessment result is obtained, thereby improving the efficiency of fish habitat monitoring.

[0083] Optional, Figure 1 This is a schematic diagram of a fish habitat monitoring system architecture provided by an embodiment of the present application, wherein the application data processing system is a computer device. Figure 1 In the above architecture, the above architecture includes at least one of a data acquisition device 11, a processing device 12 and a display device 13.

[0084] It is understandable that the structure illustrated in the embodiment of the present application does not constitute a specific limitation on the architecture of the fish habitat monitoring system. In other feasible implementations of the present application, the above architecture may include more or fewer components than shown in the figure, or combine certain components, or split certain components, or arrange the components differently, which can be determined according to the actual application scenario and is not limited here. Figure 1 The components shown may be implemented in hardware, software, or a combination of software and hardware.

[0085] In the specific implementation process, the data acquisition device 11 is a drone device, which may include a wireless data transmission device. The data acquisition device 11 may transmit the collected data to the processing device through the wireless data transmission device.

[0086] The processing device 12 can automatically process and identify the collected images through the convolutional neural network model, thereby avoiding the influence of human intervention. For the convolutional neural network model, a spatial pyramid pooling layer is introduced to enhance the recognition ability of multi-scale features in complex natural environments, and an improved residual block is introduced to improve the accuracy and efficiency of feature extraction through parallel convolution layers and attention mechanisms. At the same time, the single coastal zone type calculation is changed to a variety of ecological indicators to conduct a comprehensive assessment of fish habitats, and a scientific and comprehensive assessment result is obtained, thereby improving the efficiency of fish habitat monitoring.

[0087] The display device 13 may also be a touch screen or a screen of a terminal device, which is used to receive user instructions while displaying the above-mentioned content to achieve interaction with the user.

[0088] It should be understood that the above-mentioned processing device can be implemented by a processor reading instructions in a memory and executing the instructions, or it can be implemented by a chip circuit.

[0089] In addition, the network architecture and business scenarios described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Ordinary technicians in this field can know that with the evolution of network architecture and the emergence of new business scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0090] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0091] Figure 2 A flow chart of a method for monitoring fish habitat provided in an embodiment of the present application Figure 1 ,like Figure 2 As shown, the method includes:

[0092] S201. Determine the target area to be monitored based on the habitat information of the target fish to be monitored, and generate a target navigation route for the aerial robot.

[0093] In this embodiment, the target fish to be monitored is determined, the habitat characteristic information of the target fish to be monitored is collected according to the research purpose, the target area to be monitored is determined, and the target navigation route of the aerial robot is generated.

[0094] Among them, the target area to be monitored is mainly the river target investigation scope.

[0095] Among them, the aerial robot is a drone.

[0096] Among them, the target navigation route of the aerial robot includes flight path, flight altitude and overlap parameters.

[0097] S202: Control the aerial robot to fly along the target navigation route, and collect a plurality of image data captured by the aerial robot during the flight.

[0098] In this embodiment, the aerial robot is controlled to cruise along a target navigation route in a target area to be monitored, and multiple image data continuously captured by the aerial robot during the flight are collected.

[0099] S203, stitching the multiple image data to obtain an orthophoto.

[0100] In this embodiment, the collected multiple image data are spliced ​​to generate a continuous orthophoto map.

[0101] S204: Perform image preprocessing on the orthophoto image to obtain a preprocessed image.

[0102] Wherein, at least one of denoising, enhancement and color correction is performed on the orthophoto image to obtain a preprocessed image.

[0103] In this embodiment, the orthophoto image is subjected to denoising, enhancement, color calibration and brightness adjustment, and then the generated orthophoto image is verified through ground control points to obtain a pre-processed image.

[0104] S205: Using a convolutional neural network algorithm, perform feature recognition on the preprocessed image to obtain at least one feature region.

[0105] In this embodiment, a convolutional neural network algorithm is used to identify the habitat features of target fish species on the preprocessed image, and extract feature areas related to the calculation index.

[0106] Among them, the calculated index is the discrimination index of the characteristic area of ​​the target fish habitat.

[0107] Specifically, the number of calculation indicators is at least one, and correspondingly, the number of feature regions related to the calculation indicators is at least one. In the present application, generally, the number of calculation indicators is multiple, and the number of feature regions related to the calculation indicators is also multiple, thereby improving the comprehensiveness of the evaluation results.

[0108] S206: Perform discrimination calculation on at least one characteristic area according to the calculation index corresponding to the target fish to be monitored, so as to obtain an assessment result of the ecological health status of the target area.

[0109] In this embodiment, according to the calculation index corresponding to the target fish to be monitored, a discrimination calculation is performed on at least one corresponding characteristic area, and a comprehensive evaluation is performed in combination with the calculation results to obtain an evaluation result of the ecological health status of the target area.

[0110] The present embodiment provides a method for monitoring fish habitats. According to the habitat information of the target fish to be monitored, the target area to be monitored is determined, and the target navigation route of the aerial robot is generated. The aerial robot is controlled to fly along the target navigation route, and multiple image data taken during the flight are collected. The multiple image data are spliced ​​to obtain an orthophoto image and perform image preprocessing. The preprocessed image image is feature-recognized using a convolutional neural network algorithm to obtain a feature area. According to the calculation index corresponding to the target fish to be monitored, the feature area is discriminated and calculated to obtain an evaluation result of the ecological health status of the target area. The drone image data acquisition provided by the present application does not require manual intervention, which improves the efficiency of data acquisition. The present application provides an improved convolutional neural network model, which performs automated data processing and efficient and accurate feature recognition on the image image based on a spatial pyramid pooling layer and an improved residual to obtain a feature area. The present application performs a comprehensive and detailed calculation and evaluation of the feature area based on a variety of ecological indicators. The present method is used to monitor fish habitats, solves the above-mentioned technical problems, and has the advantages of automation, convenience, high accuracy, high precision, high efficiency and comprehensiveness, thereby improving the efficiency of monitoring fish habitats.

[0111] Figure 3 A flow chart of a method for monitoring fish habitat provided in an embodiment of the present application Figure 2 ,exist Figure 2 Based on the example, Figure 3 As shown, the specific implementation steps of the above S205 include:

[0112] S301 , normalizing the preprocessed image to obtain a normalized image.

[0113] In this embodiment, the orthophoto image is normalized to normalize the image pixel values ​​to a range of [0, 1] to obtain a normalized orthophoto image.

[0114] S302: Input the normalized image map into the trained improved convolutional neural network model to obtain a feature map through the output result of the trained improved convolutional neural network model.

[0115] Among them, the improved convolutional neural network model introduces a spatial pyramid pooling layer and an improved residual block, and combines multi-scale information capture and attention mechanism, so that the model can better identify and extract multi-scale features of fish habitats. The spatial pyramid pooling layer can effectively capture feature information of different scales, and enhance the model's ability to recognize multi-scale features in complex natural environments. The improved residual block improves the accuracy and efficiency of feature extraction through parallel convolutional layers and attention mechanism.

[0116] It should be noted that the improved convolutional neural network model includes an input layer, a first convolutional layer, a spatial pyramid pooling layer, an improved residual block, a feature fusion layer, a fully connected layer and an output layer.

[0117] Among them, the spatial pyramid pooling layer is used to capture multi-scale information.

[0118] Among them, the improved residual block includes multiple parallel convolutional layers to capture deep features.

[0119] As an example, Figure 4 A schematic diagram of the structure of the improved convolutional neural network model provided in the embodiment of the present application is shown in FIG. Figure 4 As shown, the improved convolutional neural network model specifically includes: an input layer 41, a first convolutional layer 42, a spatial pyramid pooling layer 43, an improved residual block 44, a feature fusion layer 45, a fully connected layer 46 and an output layer 47. The specific steps of running the improved convolutional neural network model are as follows:

[0120] S401. Design and construct an improved convolutional neural network model for target fish habitat feature recognition. The specific implementation steps are as follows:

[0121] Step 1: Design the input layer 41 to receive the normalized orthophoto map as input. The specific implementation formula is:

[0122] Input=I norm

[0123] Among them, Input is the input layer, which is used to receive the normalized orthophoto map; I norm It is an orthophoto.

[0124] Step 2: Design the first convolutional layer 42. The first convolutional layer is used to capture the basic characteristics of the target fish habitat. The specific implementation formula is:

[0125]

[0126] Among them, Conv1 is the first convolutional layer, which is used to capture the basic characteristics of the target fish habitat; Conv2D is a two-dimensional convolution operation; is the weight matrix of the first convolutional layer; is a bias term; ReLU is a rectified linear unit, a nonlinear activation function; BatchNorm is a batch normalization technique, a technique used to improve the speed and stability of neural network training.

[0127] Step 3: Introduce the spatial pyramid pooling layer 43 to capture the multi-scale information of the target fish habitat characteristics. The specific implementation formula is:

[0128]

[0129] Among them, SPP is the spatial pyramid pooling process; L is the number of pooling operations of different scales in the spatial pyramid pooling layer, which is the number of layers or branches of the pyramid; Pool l It is the first pooling layer operation.

[0130] Specifically, spatial pyramid pooling allows the network to accept input images of any size without cropping or scaling the image, thereby retaining the original information of the image and being flexible; spatial pyramid pooling can capture the features of images at different scales through pooling operations at different scales, thereby improving the performance of the network; the spatial pyramid pooling layer can accept input images of different sizes, which is equivalent to data enhancement of the input image, which helps to reduce overfitting and improve efficiency.

[0131] Step 4: Design an improved residual block 44 targeting the characteristics of the target fish habitat. The improved residual block contains multiple parallel convolutional layers and is combined with an attention mechanism. The specific implementation formula is:

[0132] ResBlock improved =Conv 3x3 +Conv 5x5 +Attention(Conv 1x1 )

[0133] Specific:

[0134] Conv 3x3 =ReLU(BatchNorm(Conv2D(SPP,W 3x3 )))+b 3x3

[0135] Conv 5x5 =ReLU(BatchNorm(Conv2D(SPP,W 5x5 )))+b 5x5

[0136]

[0137] Among them, ResBlock improved It is an improved residual block; Conv 3x3 Is a 3x3 convolutional layer; Conv 5x5 is a 5x5 convolutional layer; Attention is the application of the attention mechanism; Conv 1x1 is a 1x1 convolutional layer; ReLU is a nonlinear activation function that sets all negative values ​​to zero and keeps positive values ​​unchanged, expressed as f(x)=max(0,x); Conv2D is a two-dimensional convolutional layer; W 3x3is the 3x3 weight matrix of the convolution kernel (filter) used in the Conv2D layer; b 3x3 is the bias term of the Conv2D layer; W 5x5 is the 5x5 weight matrix of the convolution kernel (filter) used in the Conv2D layer; b 5x5 is the bias term of the Conv2D layer; softmax is a function that compresses a real vector into the interval (0,1) and the sum of all elements is 1; It's K 1x1 The transpose of Q 1x1 is the query matrix of the 1x1 convolutional layer; K 1x1 is the key matrix of the 1x1 convolutional layer; is; V 1x1 is the value matrix of the 1x1 convolutional layer.

[0138] Specifically, the attention mechanism allows the model to focus more on the important parts of the input when processing data. By passing the output of the 1x1 convolutional layer to the attention mechanism, the weights of different channels (or features) can be dynamically adjusted, thereby enhancing the model's ability to capture important information.

[0139] Step 5: Design a feature fusion layer 45 to fuse the features output by the improved residual block to enhance the expression of the target fish habitat features. The specific implementation formula is:

[0140] FeatureFusion=Concat(ResBlock improved )

[0141] Among them, FeatureFusion is the feature fusion layer; Concat is the feature fusion operation.

[0142] Step 6: After the feature fusion layer 45, a fully connected layer 46 is introduced to flatten the extracted features and classify them through the fully connected layer 46. The specific implementation formula is:

[0143]

[0144] Among them, FC i is a fully connected layer; is the weight matrix of the i-th fully connected layer; is the bias term.

[0145] Step 7: Design the output layer 47 to output the recognition results of the target fish habitat characteristics. The specific implementation formula is:

[0146] Output = softmax(FC n )

[0147] Among them, Output is the output layer; softmax is a normalization function used for multi-classification problems; FC n is the nth fully connected layer.

[0148] Step 8: Combine the layers from step 1 to step 7 to form a complete improved convolutional neural network model for target fish habitat feature recognition. The specific implementation formula is:

[0149] CNN improved

[0150] =(Input→Conv1→SPP→ResBlock improved

[0151] →FeatureFusion→FC1→Output)

[0152] Among them, CNN improved It is a convolutional neural network model.

[0153] S402. Using a pre-labeled target fish habitat feature data set to train an improved convolutional neural network model, optimizing model parameters through a back propagation algorithm and a gradient descent method, so that the improved convolutional neural network model can identify the target fish habitat features.

[0154] It can be understood that the pre-labeled target fish habitat characteristic data set here can be determined according to actual conditions, and the embodiments of the present application do not impose specific limitations on this.

[0155] S403, using the trained convolutional neural network model to extract features from the normalized orthophoto image to obtain a convolutional neural network model feature map.

[0156] S303: Classify and identify the convolutional neural network model feature map according to a preset discrimination criterion to obtain at least one feature region.

[0157] First, according to the preset discrimination criteria, the convolutional neural network model feature map is classified to identify the feature areas related to the discrimination criteria. The specific implementation formula is:

[0158] R = {r i ∣r i ∈F,classify(r i ,S criteria )=true}

[0159] Among them, R is the set of identified feature regions; r i is the i-th region in the convolutional neural network model feature map; F is the convolutional neural network model feature map; S criteriais the discrimination criterion; classify is the feature region classification function; true is a Boolean value used to represent the logical "true" or "yes" state.

[0160] It can be understood that the preset judgment criteria here can be determined according to actual conditions, and the embodiments of the present application do not impose specific limitations on this.

[0161] Thirdly, the identified feature areas are extracted from the convolutional neural network model feature map to form a feature area map of the target fish habitat. The specific implementation formula is:

[0162]

[0163] Among them, T habitat is the characteristic area map of the target fish habitat; |R| is the size of the characteristic area set.

[0164] In this embodiment, the convolutional neural network model is improved based on the spatial pyramid pooling layer, the improved residual block, the multi-scale information capture and the attention mechanism, so as to eliminate the input size limitation of the convolutional neural network model and make the model flexible; through the multi-pooling operation, the model's ability to extract image features is improved, thereby improving the accuracy of image classification and target detection; avoiding repeated feature extraction processes and improving computational efficiency; through multi-scale feature extraction, the model can learn more abstract features and enhance the generalization ability of the model; through the residual block, the model accuracy is improved and the model training is accelerated.

[0165] Figure 5 A flow chart of a method for monitoring fish habitat provided in an embodiment of the present application Figure 3 ,exist Figure 2 and Figure 3 Based on the example, Figure 5 As shown, the specific implementation steps before S302 also include:

[0166] S501: Acquire a feature data set, wherein the feature data set includes a plurality of pre-labeled normalized image sample images.

[0167] Among them, the feature data set includes image data of the target river section in different seasons and under different meteorological conditions.

[0168] It can be understood that the pre-labeled normalized image sample map here can be determined according to actual conditions, and the embodiments of the present application do not impose specific limitations on this.

[0169] S502: using the feature data set to train the initial improved convolutional neural network model, and optimizing the model parameters of the improved convolutional neural network model through the back propagation algorithm and the gradient descent method to obtain a trained improved convolutional neural network model.

[0170] Specifically, back propagation optimizes the parameters by calculating the gradient of the loss function for each layer of parameters and back propagating these gradients to the previous layer of the network. The gradient descent method can gradually adjust the weights and biases in the model so that the model's predicted output is closer to the true label. It is simple and easy to implement, and can also perform well for large-scale data sets.

[0171] In this embodiment, a feature data set is obtained based on a large number of image sample images, and then based on the feature data set, the model parameters of the convolutional neural network model are optimized and improved through the back propagation algorithm and the gradient descent method, thereby improving the efficiency, flexibility, automated feature extraction, strong generalization ability, adaptability and convenience of the model.

[0172] Figure 6 A flow chart of a method for monitoring fish habitat provided in an embodiment of the present application Figure 4 ,exist Figure 2 , Figure 3 as well as Figure 5 Based on the example, Figure 6 As shown, the calculation indexes include the coastal zone type calculation index, the boulder bedrock proportion calculation index in the river channel, the river channel width calculation index and the flow state calculation index. The specific implementation steps of the above S206 include:

[0173] S601. According to the calculation index of coastal zone type, the calculation index of the proportion of boulder bedrock in the river channel, the calculation index of river channel width and the calculation index of flow pattern speed, determine the proportion of coastal zone type, the proportion of boulder bedrock in the river channel, the river channel width coefficient and the flow pattern speed coefficient of each characteristic area respectively.

[0174] The coastal zone type is to quantitatively analyze the coastal zone types within 50 meters on both sides of the river channel (the 50 meters here is only an example, in the specific implementation process, it can be any range, and the embodiment of the present application does not impose specific restrictions on this) , including the proportion of different types of buildings, vegetation, and silt deposition belts, calculate the area of ​​each type in the coastal zone, and calculate the proportion of each type:

[0175]

[0176] Among them, P type,k represents the proportion of the kth coastal zone type; R type,k represents the area of ​​the kth coastal zone type; K is the total number of coastal zone types; Area represents the function for calculating the area of ​​the region;

[0177] According to the characteristic area R in the identified river channel area boulders , calculate the area percentage of boulders and bedrock in the river channel:

[0178]

[0179] Among them, P boulders represents the area ratio of boulders and bedrock, R boulders Characteristic areas representing boulders and bedrock, R river It represents the overall water surface area of ​​the river, and Area represents the function for calculating the area of ​​the region.

[0180] Secondly, according to the river morphological characteristics of the characteristic area, the river width is calculated. The specific implementation formula is:

[0181]

[0182] Among them, W river is the width of the river; d i is the river width at the ith measuring point; N is the number of measuring points.

[0183] Among them, the river width calculation indicators include river width, river width at each measuring point, number of measuring points and other related data.

[0184] Finally, according to the flow characteristics of the characteristic area, the flow urgency coefficient is calculated. The specific implementation formula is:

[0185]

[0186] Among them, F flow is the flow state coefficient, v j is the flow velocity at the jth measurement point; d j is the depth of the jth measurement point; M is the number of measurement points.

[0187] Among them, the flow speed calculation indicators include flow speed coefficient, flow velocity at each measuring point, depth of each measuring point, number of measuring points and other related data.

[0188] It should be noted that the proportion of coastal zone types, river channel width coefficient and flow speed coefficient are the indicator values ​​corresponding to the three calculation indicators of coastal zone type, river channel width and flow speed coefficient respectively.

[0189] S602. Calculate the comprehensive score of each characteristic area based on the weights corresponding to the coastal zone type calculation index, the boulder bedrock proportion in the river channel calculation index, the river channel width calculation index and the flow rate calculation index, combined with the coastal zone type proportion, boulder bedrock proportion in the river channel, river channel width coefficient and flow rate coefficient of each characteristic area.

[0190] Specifically, the index values ​​of all feature areas are combined and the comprehensive score of the feature area is calculated according to the preset judgment criteria. The specific implementation formula is:

[0191]

[0192] Among them, S score is the comprehensive score of the feature area; w k is the weight of the kth indicator; V k is the kth indicator value; k is the number of indicators.

[0193] Among them, the index values ​​include the proportion of coastal zone types, the proportion of boulder bedrock in the river channel, the river channel width coefficient and the flow speed coefficient.

[0194] It can be understood that the preset judgment criteria here can be determined according to actual conditions, and the embodiments of the present application do not impose specific limitations on this.

[0195] S603. Calculate the total ecological health score of the target area based on the comprehensive score of each characteristic area.

[0196] First, the ecological health status of fish habitats was evaluated based on the comprehensive scores of characteristic areas combined with discriminant indicators.

[0197] Secondly, other ecological indicators were calculated using a multiple regression analysis model in combination with the assessment data.

[0198] Among them, other ecological indicators include vegetation diversity index, water quality indicators, and climate conditions.

[0199] Finally, based on the discriminant indicators and other ecological indicators, the analytic hierarchy process was used to construct an ecological health assessment model to conduct a comprehensive assessment of the ecological health status of fish habitats and calculate the total score of the ecological health status of the target area.

[0200] In this embodiment, it is not limited to the calculation of a single coastal zone type, but also combines multiple ecological indicators to conduct a comprehensive assessment of fish habitats. Through characteristic area identification and indicator calculation, a detailed characteristic area map and a comprehensive score are generated to provide scientific and comprehensive assessment results.

[0201] Figure 7 A flow chart of a method for monitoring fish habitat provided in an embodiment of the present application

[0202] Five, in Figure 2 , Figure 3 , Figure 5 as well as Figure 6 Based on the example, Figure 7 As shown, the above S201

[0203] The specific implementation steps include:

[0204] S701, determine the target fish to be monitored.

[0205] Specifically, according to the research purpose, select the target fish species to be monitored, and the specific implementation formula is:

[0206]

[0207] Among them, F target is the target fish; arg max is the function, the parameter (or variable) that makes (x) reach the maximum; It is the collection of all potential target fishes; For all the f in P f,importance The importance of fish research; f,availability is the availability of fish data; P f,conservation is the conservation value of fish; ω1 is the weight coefficient; ω2 is the weight coefficient; ω3 is the weight coefficient.

[0208] S702: Acquire habitat information of target fish, wherein the habitat information includes geographical location information, environmental condition information, and biological characteristic information.

[0209] Specifically, the implementation formula is:

[0210]

[0211] Among them, I habitat It is a comprehensive indicator of the habitat of target fish; G i,geo is the geographical location; j,env is the environmental condition; B k,bio is a biological characteristic; α i is the weight coefficient; β j is the weight coefficient; γ k is the weight coefficient.

[0212] It should be noted that, after step S702, the following steps are also included:

[0213] Optionally, the collected characteristic information of the target fish habitat is sorted and entered into a database to form structured data. The specific implementation formula is:

[0214] D structured =T(I habitat ,F target )

[0215]

[0216] Among them, D structuredis structured data; T is a data conversion function that converts these elements into structured data pairs; H i is the i-th element of the habitat characteristic data of the target fish; F j is the j-th element of the target fish data.

[0217] Further optionally, based on the survival needs of the target fish, key target fish habitat characteristic information is extracted from the database, including coastal zone type, river width, flow speed, distribution of deep pools and shallows, straightness, curvature, and bifurcation of the river channel, and bottom characteristics.

[0218] Among them, boulders with a particle size greater than 20 cm, pebbles with a diameter of 5-20 cm, and fine sand and silt less than 5 cm were extracted.

[0219] The specific implementation formula is:

[0220] K features ={δ1C1+δ2W2+δ3F3+δ4T4+δ5M5+δ6B6}

[0221] Among them, K features is a set of key target fish habitat characteristics; C1 is the coastal zone type; W2 is the river width; F3 is the flow speed; T4 is the distribution of deep pools and shallows; M5 is the river morphology; B6 is the bottom characteristics; δ1 is the characteristic weight; δ2 is the characteristic weight; δ3 is the characteristic weight; δ4 is the characteristic weight; δ5 is the characteristic weight; δ6 is the characteristic weight.

[0222] S703: Determine corresponding feature discrimination criteria according to habitat information.

[0223] In this embodiment, based on the extracted habitat feature information, a discrimination criterion for the target fish habitat feature is formulated for image recognition and habitat assessment. The specific implementation formula is:

[0224] S criteria =(K features ×R standards )

[0225] Among them, S criteria is habitat information; R standards is a matrix of standard values ​​of features.

[0226] S704. Generate a feature collection map based on habitat information and feature identification criteria using geographic information system technology.

[0227] In this embodiment, according to the feature discrimination criteria and habitat feature information, GIS technology is used to generate a feature collection map of the target fish habitat for planning the flight path and altitude of the drone. The specific implementation formula is:

[0228] M map =G(S criteria ,I habitat ,F target )

[0229]

[0230] Among them, M map is the characteristic collection map of the target fish habitat; G is the function of GIS technology to generate the map; ∫ A is the integral over the area; ⊙ is the successive multiplication of elements; A is the area of ​​the target area; F target,k is the impact of the target fish at the kth habitat feature point.

[0231] S705: Generate a target navigation route for the aerial robot based on the feature collection graph.

[0232] Among them, the aerial robot is a drone.

[0233] Among them, the target navigation route is controlled by the flight path, flight altitude and overlap parameters.

[0234] It should be noted that, after step S705, the following steps are also included:

[0235] Optionally, the image data is continuously captured during the flight. The specific implementation formula is:

[0236]

[0237] Among them, I capture (t) is the image data set captured at time point t; P uav (t i ) is the UAV at time t i is the location information; Δt is the time interval; V uav (t i ) is the UAV at time t i Speed; F camera (t i ) is the camera device at time t i focal length and exposure parameters; R image (t i ) is the image resolution and quality parameter.

[0238] In this embodiment, target fish are selected based on multiple angles, which improves the scientific nature of target fish screening; multiple influencing factors are introduced based on weight coefficients, which improves flexibility and pertinence; the UAV cruising path and altitude are formulated based on the characteristic collection map of the target fish habitat, which improves the efficiency of UAV cruising; based on UAV photography, high-resolution data images are obtained, which avoids the tediousness and errors of manual intervention, saves labor costs, and improves the accuracy and efficiency of data processing.

[0239] Figure 8 A flow chart of a method for monitoring fish habitat provided in an embodiment of the present application Figure 6 ,exist Figure 2 , Figure 3 , Figure 5 , Figure 6 as well as Figure 7 Based on the example, Figure 8 As shown, the specific implementation steps of the above S203 include:

[0240] S801, performing registration processing on multiple image data, identifying common feature points in adjacent image data through a feature point matching algorithm, and calculating a transformation matrix between the image data.

[0241] In this embodiment, the specific implementation formula is:

[0242]

[0243] Among them, T i,j is the transformation matrix between images; argmin is a function used to find the parameter value that makes a function reach the minimum value; is the kth feature point in the i-th image; is the kth feature point in the jth image; M is the number of feature points.

[0244] It is mainly noted that before step S801, the following steps are also included:

[0245] Optionally, the image data captured by the drone during the cruise flight is extracted from the storage device to form a preliminary image data set. The specific implementation formula is:

[0246] I raw = {I i |i=1,2,…,N}

[0247] Among them, I raw is a preliminary image data set; I i is the i-th image data; N is the total number of images.

[0248] S802: splicing adjacent image data according to the transformation matrix to obtain a spliced ​​image.

[0249] In this embodiment, adjacent images are spliced ​​together using the calculated transformation matrix to generate a spliced ​​image.

[0250] Among them, the stitched image is I seamless .

[0251] Specifically, the implementation steps of step S802 are as follows:

[0252] Step 1: Use a feature point detection algorithm to detect feature points on each image in the preliminary image data set to generate a feature point set.

[0253] Step 2: Use the feature point matching algorithm to match the feature point sets of adjacent images to generate a set of matching pairs. The specific implementation formula is:

[0254]

[0255] Among them, M i,i+1 is the set of matching pairs of feature points of the i-th and i+1-th images; P i is a set of feature points; P i+1 is with P i A set of feature points of adjacent images; match is a feature point matching function.

[0256] Step 3: Calculate the transformation matrix between images based on the matching pair set. The specific implementation formula is:

[0257]

[0258] Among them, T i,i+1 is the transformation matrix between the i-th and i+1-th images; ∥x∥ is the Euclidean distance.

[0259] Step 4: According to the transformation matrix, transform the i+1th image to align it with the i-th image to obtain the transformed image. The specific implementation formula is:

[0260] I i+1,transformed =T i,i+1 I i+1

[0261] Among them, I i+1,transformed is the i+1th image after transformation.

[0262] Step 5: Calculate the weight matrix based on the image overlapping area to smooth the image edges and optimize the stitching gap. The specific implementation formula is:

[0263]

[0264] Where W is the weight matrix; wij is the element of the weight matrix; d ij is the distance from the i,jth pixel in the overlapping area of ​​the image to the overlapping boundary; overlap is the overlapping area of ​​the image.

[0265] Step 6: Use the calculated weight matrix to perform weight smoothing on the transformed image so that it transitions smoothly with the i-th image. The specific implementation formula is:

[0266] I smooth,i,i+1 =W·I i,overlap +(1-W)·I i+1,transformed,overlap

[0267] Among them, I smooth,i,i+1 is the smoothed stitched image; I i,overlap is the i-th image; I i+1,transformed,overlap is the part of the transformed i+1th image in the overlapping area; · is the element-by-element multiplication.

[0268] Step 7: Perform color correction on the smoothed stitched image, using a color histogram matching algorithm to ensure color consistency in the stitched area.

[0269] Step 8: Fuse the color-corrected images to generate the final seamless stitching image.

[0270] Among them, the seamless stitching image is also called a seamless stitching image.

[0271] S803: crop the stitched image and perform geometric correction processing to obtain an orthophoto image.

[0272] First, the stitched image is cropped to remove the overlapping parts and invalid areas to obtain the cropped image data.

[0273] Secondly, the cropped image data is geometrically corrected, and the geometric distortion in the image is corrected to the real geographic coordinates using geographic reference points and known geographic coordinates.

[0274] Finally, the geometrically corrected image data are integrated to generate a continuous orthophoto map.

[0275] Among them, the image data is I cropped ; The orthophoto is I ortho .

[0276] In this embodiment, based on image registration and stitching, a continuous and seamless orthophoto map is obtained, which improves the continuity and integrity of the data; based on cropping, overlapping parts and invalid areas are removed, which improves the compactness and efficiency of the orthophoto map; based on geometric correction, the geometric distortion in the image is corrected to the real geographic coordinates, which improves the geographic precision and positioning accuracy of the image; based on ground control points, the generated orthophoto map is verified, which improves the reliability and accuracy of the orthophoto map.

[0277] The present application also provides a preferred embodiment, which selects a section of fish habitat in the upper reaches of a river as the research object, and the specific location is a section of a mountain river in S Province, and the length of the section is about 5 kilometers. This area has typical river bends, slow and fast flow patterns, and diverse vegetation coverage characteristics, and is an important habitat for a variety of freshwater fish. Traditional ground survey methods are difficult to fully and accurately obtain data on the type and ecological status of the coastal zone in this area, and are time-consuming and laborious.

[0278] Step 1: High-resolution images of the target river section were collected using drones.

[0279] Specifically, the drone cruises along the preset flight path and altitude, continuously capturing image data over the entire river section. The drone flight parameters are set as follows: flight altitude of 50 meters, aerial resolution of 5 cm / pixel, and image coverage of 90%, ensuring high coverage and high-resolution data collection.

[0280] Step 2: Preprocess the collected image data, including image correction, denoising and stitching, to generate continuous orthophotos.

[0281] Specifically, these images are normalized to normalize the image pixel values ​​to the range of [0,1] for subsequent image recognition.

[0282] Step 3: Design and build an improved convolutional neural network model for fish habitat feature recognition.

[0283] Specifically, the model captures multi-scale fish habitat features through spatial pyramid pooling layers and improved residual blocks, and combines attention mechanisms to improve recognition accuracy and efficiency. The model is trained using a pre-labeled fish habitat feature dataset, which contains image data of the river section in different seasons and under different meteorological conditions, totaling 5,000 image samples.

[0284] Step 4: Use the trained improved convolutional neural network model to extract features from the normalized orthophoto image and identify the characteristic areas of the target fish habitat.

[0285] Specifically, these characteristic areas include coastal zone type area, river channel bottom composition (boulders and bedrock ratio) area, river channel width area and flow regime slowness and rapidity area, etc. Based on the identified characteristic areas and the preset discrimination criteria, the indicators such as the proportion of coastal zone type, the proportion of boulder bedrock in the river channel, the river channel width and the flow regime slowness and rapidity coefficient are calculated.

[0286] It should be noted that in order to verify the effectiveness of the method of this application, the evaluation results were compared with the traditional ground survey method. The traditional ground survey method conducted a one-week survey in the same river section, using manual measurement and recording data. In terms of evaluation results, the method provided by this application obtained the following Table 1:

[0287] Table 1 Data proof and comparison

[0288]

[0289] Among them, by comparing the data, we can clearly see that the method provided by this application has significant advantages over traditional methods in many aspects.

[0290] Specifically, the method provided in this application estimates that buildings account for 12.3% of the coastal zone, vegetation accounts for 63.4%, and silt deposits account for 24.3%. Traditional survey methods estimate that buildings account for 10.8%, vegetation accounts for 60.5%, and silt deposits account for 28.7% of the coastal zone. This shows that the method provided in this application is more accurate in distinguishing the types of coastal zones of fish habitats and can capture the status of fish habitats more comprehensively. This is mainly due to the high-precision feature extraction capability of the improved convolutional neural network model.

[0291] More specifically, the proportion of boulder bedrock in the river channel assessed by the method provided in this application is 23.4%, while the proportion of boulder bedrock in the river channel measured by the traditional method is 10.4%, indicating that the method provided in this application is more accurate in identifying the composition of the river channel bottom.

[0292] More specifically, the average width of the river channel evaluated by the method provided by this application is 24.6 meters, while the average width of the river channel measured by the traditional method is 23.9 meters. Although the difference between the two is not large, the automated data processing and recognition of the method provided by this application significantly reduces human errors and ensures the consistency and reliability of the measurement results.

[0293] More specifically, the method provided by this application obtains a flow rate coefficient of 1.8, while the traditional method is 1.7. The calculation of the flow rate coefficient requires accurate measurement of flow velocity and river depth. The method provided by this application achieves accurate recognition and calculation of these features through high-resolution drone images and improved convolutional neural network models.

[0294] More specifically, the data collection time of the method provided by this application is only 2 hours, while the traditional method takes 5 days. In addition, the image processing and feature recognition time is 4 hours, while the traditional method takes 2 days. In terms of total time, the method provided by this application only takes 6 hours, while the traditional method takes 7 days. This significant difference is mainly due to the application of drone technology and automated image recognition algorithms, which improve the efficiency of data collection and processing.

[0295] In this embodiment, the method provided in this application performs significantly better than traditional methods in fish habitat monitoring, which not only improves the accuracy and efficiency of the assessment, but also significantly reduces working time and labor costs, providing strong support for the scientific management and protection of fish habitats.

[0296] Fig. 9 A schematic diagram of the structure of a fish habitat monitoring device provided in an embodiment of the present application is shown in FIG. Fig. 9 As shown, the device includes: a first determination module 91, a first control module 92, a first processing module 93, a second processing module 94, a feature extraction module 95 and a third processing module 96.

[0297] The first determination module 91 is used to determine the target area to be monitored according to the habitat information of the target fish to be monitored, and generate a target navigation route of the aerial robot.

[0298] The first control module 92 is used to control the aerial robot to fly along the target navigation route and collect a plurality of image data taken by the aerial robot during the flight.

[0299] The first processing module 93 is used to perform splicing processing on multiple image data to obtain an orthophoto map.

[0300] The second processing module 94 is used to perform image preprocessing on the orthophoto image to obtain a preprocessed image.

[0301] The feature extraction module 95 is used to perform feature recognition on the preprocessed image using a convolutional neural network algorithm to obtain at least one feature area.

[0302] The third processing module 96 is used to perform discrimination calculation on at least one characteristic area according to the calculation index corresponding to the target fish to be monitored, so as to obtain an assessment result of the ecological health status of the target area.

[0303] In a possible design, the feature extraction module 95 is further used to normalize the preprocessed image map to obtain a normalized image map;

[0304] Inputting the normalized image map into the trained improved convolutional neural network model to obtain a feature map through the output result of the trained improved convolutional neural network model;

[0305] According to a preset discrimination criterion, the feature map is classified and identified to obtain at least one feature region.

[0306] In a possible design, the feature extraction module 95 is further used to, before inputting the normalized image map into the trained improved convolutional neural network model to obtain the feature map through the output result of the trained improved convolutional neural network model:

[0307] Acquire a feature data set, wherein the feature data set includes a plurality of pre-labeled normalized image sample images;

[0308] The feature data set is used to train the initial improved convolutional neural network model, and the model parameters of the improved convolutional neural network model are optimized through the back propagation algorithm and the gradient descent method to obtain the trained improved convolutional neural network model.

[0309] In one possible design, the improved convolutional neural network model includes an input layer, a first convolutional layer, a spatial pyramid pooling layer, an improved residual block, a feature fusion layer, a fully connected layer, and an output layer;

[0310] The spatial pyramid pooling layer is used to capture multi-scale information;

[0311] The improved residual block includes multiple parallel convolutional layers to capture deep features.

[0312] In a possible design, the calculation index includes a coastal zone type calculation index, a boulder bedrock ratio calculation index in the river channel, a river channel width calculation index, and a flow state speed calculation index. Accordingly, the third processing module 96 is further used to determine the coastal zone type ratio, the boulder bedrock ratio in the river channel, the river channel width coefficient, and the flow state speed coefficient of each characteristic area according to the coastal zone type calculation index, the boulder bedrock ratio calculation index in the river channel, the river channel width calculation index, and the flow state speed calculation index.

[0313] According to the weights corresponding to the coastal zone type calculation index, the boulder bedrock proportion calculation index in the river channel, the river channel width calculation index and the flow state speed calculation index, the comprehensive score of each characteristic area is calculated by combining the coastal zone type proportion, boulder bedrock proportion in the river channel, river channel width coefficient and flow state speed coefficient of each characteristic area;

[0314] Based on the comprehensive score of each characteristic area, the overall score of the ecological health status of the target area is calculated.

[0315] In a possible design, the first determination module 91 is also used to determine the target fish to be monitored;

[0316] Obtaining habitat information of target fish, wherein the habitat information includes geographical location information, environmental condition information and biological characteristic information;

[0317] Determine the corresponding feature discrimination criteria based on habitat information;

[0318] Based on habitat information and feature identification criteria, geographic information system technology is used to generate feature collection maps;

[0319] Based on the feature collection graph, the target navigation route of the aerial robot is generated.

[0320] In a possible design, the first processing module 93 is further used to perform registration processing on multiple image data, identify common feature points in adjacent image data through a feature point matching algorithm, and calculate a transformation matrix between the image data;

[0321] According to the transformation matrix, adjacent image data are spliced ​​to obtain a spliced ​​image;

[0322] The stitched image is cropped and geometrically corrected to obtain an orthophoto.

[0323] In a possible design, the second processing module 94 is further used to perform at least one of denoising, enhancement and color correction on the orthophoto image to obtain a preprocessed image.

[0324] A fish habitat monitoring device provided in this embodiment can execute a fish habitat monitoring method of the above embodiment, and its implementation principle and technical effect are similar, which will not be described in detail in this embodiment.

[0325] In the specific implementation of the aforementioned method for monitoring a fish habitat, each module can be implemented as a processor, and the processor can execute computer-executable instructions stored in the memory, so that the processor executes the aforementioned method for monitoring a fish habitat.

[0326] Fig.10 This is a schematic diagram of the structure of a fish habitat monitoring device provided in an embodiment of the present application. Fig.10 As shown, the electronic device 100 includes: at least one processor 101 and a memory 102. The electronic device 100 also includes a communication component 103. The processor 101, the memory 102 and the communication component 103 are connected via a bus 104.

[0327] In the specific implementation process, at least one processor 101 executes the computer execution instructions stored in the memory 102, so that at least one processor 101 executes a fish habitat monitoring method belonging to claims 1-7 executed by the above fish habitat monitoring device side.

[0328] The specific implementation process of the processor 101 can be found in the above method embodiment, and its implementation principle and technical effect are similar, so this embodiment will not be repeated here. In the above embodiment, it should be understood that the processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. The general-purpose processor can be a microprocessor or the processor can be any conventional processor, etc. The steps of the method disclosed in the invention can be directly embodied as being executed by a hardware processor, or can be executed by a combination of hardware and software modules in the processor.

[0329] The memory may include a high-speed RAM memory, and may also include a non-volatile storage NVM, such as at least one disk storage.

[0330] The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, the bus in the drawings of the present application is not limited to only one bus or one type of bus.

[0331] The functions implemented by the above-mentioned monitoring equipment and main control equipment for fish habitats are introduced to the scheme provided by the embodiment of the present invention. It can be understood that in order to realize the above-mentioned functions, the monitoring equipment or main control equipment for fish habitats includes hardware structures and / or software modules corresponding to the execution of each function. In combination with the units and algorithm steps of each example described in the embodiment disclosed in the embodiment of the present invention, the embodiment of the present invention can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art may use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the technical solution of the embodiment of the present invention.

[0332] The present application also provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the above-mentioned method for monitoring fish habitats.

[0333] The above-mentioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The readable storage medium can be any available medium that can be accessed by a general or special-purpose computer.

[0334] An exemplary readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (Application Specific Integrated Circuits, referred to as: ASIC). Of course, the processor and the readable storage medium can also exist as discrete components in an electronic device or a main control device.

[0335] The present application also provides a computer program product, including a computer program, wherein the computer program is stored in a readable storage medium, and at least one processor of an electronic device can read the computer program from the readable storage medium, and at least one processor executes the computer program so that the electronic device executes the solution provided by any of the above embodiments.

[0336] Those skilled in the art can understand that all or part of the steps of the above method embodiments can be completed by hardware related to program instructions. The above program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above method embodiments are executed; and the above storage medium includes: ROM, RAM, disk or optical disk and other media that can store program codes.

[0337] So far, the technical solution of the present application has been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present application is obviously not limited to these specific embodiments, and the above embodiments are only used to illustrate the technical solution of the present application rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for monitoring fish habitat, characterized in that: include: According to the habitat information of the target fish to be monitored, the target area to be monitored is determined, and the target navigation route of the aerial robot is generated; Controlling the aerial robot to fly along the target navigation route, and collecting a plurality of image data captured by the aerial robot during the flight; Performing splicing processing on the multiple image data to obtain an orthophoto map; Performing image preprocessing on the orthophoto map to obtain a preprocessed image map; Using a convolutional neural network algorithm, performing feature recognition on the preprocessed image to obtain at least one feature region; According to the calculation index corresponding to the target fish to be monitored, the at least one characteristic area is discriminated and calculated to obtain an assessment result of the ecological health status of the target area.

2. The method according to claim 1, characterized in that: The method of using a convolutional neural network algorithm to perform feature recognition on the preprocessed image to obtain at least one feature area includes: Normalizing the preprocessed image to obtain a normalized image; Inputting the normalized image map into the trained improved convolutional neural network model to obtain a feature map through the output result of the trained improved convolutional neural network model; According to a preset discrimination criterion, the feature map is classified and identified to obtain at least one feature region.

3. The method according to claim 2, characterized in that Before inputting the normalized image map into the trained improved convolutional neural network model to obtain the feature map through the output result of the trained improved convolutional neural network model, the method further includes: Acquire a feature data set, wherein the feature data set includes a plurality of pre-labeled normalized image sample images; The feature data set is used to train an initial improved convolutional neural network model, and the model parameters of the improved convolutional neural network model are optimized by a back propagation algorithm and a gradient descent method to obtain a trained improved convolutional neural network model.

4. The method according to claim 3, characterized in that The improved convolutional neural network model includes an input layer, a first convolutional layer, a spatial pyramid pooling layer, an improved residual block, a feature fusion layer, a fully connected layer and an output layer; The spatial pyramid pooling layer is used to capture multi-scale information; The improved residual block includes multiple parallel convolutional layers to capture deep features.

5. The method according to any one of claims 1 to 4, characterized in that: The calculation indicators include the calculation indicators of coastal zone type, the calculation indicators of the proportion of boulder bedrock in the river channel, the calculation indicators of river channel width and the calculation indicators of flow state slowness and speed; Accordingly, the at least one characteristic area is discriminated and calculated according to the calculation index corresponding to the target fish to be monitored to obtain the evaluation result of the ecological health status of the target area, including: According to the calculation index of the coastal zone type, the calculation index of the proportion of boulder bedrock in the river channel, the calculation index of the river channel width and the calculation index of the flow state speed, the proportion of the coastal zone type, the proportion of boulder bedrock in the river channel, the river channel width coefficient and the flow state speed coefficient of each characteristic area are determined respectively; According to the weights corresponding to the coastal zone type calculation index, the boulder bedrock proportion calculation index in the river channel, the river channel width calculation index and the flow state speed calculation index, combined with the coastal zone type proportion, boulder bedrock proportion in the river channel and the flow state speed calculation index of each characteristic area, the calculation index of the coastal zone type calculation index, the boulder bedrock proportion in the river channel and the flow state speed calculation index of each characteristic area, the calculation index of the river channel width ... The bedrock proportion, river width coefficient and flow state coefficient are used to calculate the comprehensive score of each characteristic area; Based on the comprehensive score of each characteristic area, the total score of the ecological health status of the target area is calculated.

6. The method according to any one of claims 1 to 4, characterized in that: The method of determining the target area to be monitored based on the habitat information of the target fish to be monitored and generating the target navigation route of the aerial robot includes: Identify target fish species to be monitored; Acquiring habitat information of the target fish, wherein the habitat information includes geographical location information, environmental condition information, and biological characteristic information; Determining corresponding characteristic discrimination criteria according to the habitat information; Generate a feature collection map using geographic information system technology based on the habitat information and the feature identification criteria; Based on the feature collection graph, a target navigation route of the aerial robot is generated.

7. The method according to any one of claims 1 to 4, characterized in that: The step of performing splicing processing on the plurality of image data to obtain an orthophoto map includes: Performing registration processing on the multiple image data, identifying common feature points in adjacent image data through a feature point matching algorithm, and calculating a transformation matrix between the image data; splicing adjacent image data according to the transformation matrix to obtain a spliced ​​image; The stitched image is cropped and geometrically corrected to obtain an orthophoto.

8. The method according to claim 7, characterized in that The performing image preprocessing on the orthophoto image to obtain a preprocessed image image includes: At least one of denoising, enhancement and color correction is performed on the orthophoto image to obtain a preprocessed image.

9. A monitoring device for fish habitat, characterized in that: include: A first determination module is used to determine the target area to be monitored according to the habitat information of the target fish to be monitored, and generate a target navigation route of the aerial robot; A first control module, used to control the aerial robot to fly along the target navigation route and collect a plurality of image data taken by the aerial robot during the flight; A first processing module is used to perform splicing processing on the plurality of image data to obtain an orthophoto map; A second processing module is used to perform image preprocessing on the orthophoto image to obtain a preprocessed image image; A feature extraction module, used to perform feature recognition on the preprocessed image using a convolutional neural network algorithm to obtain at least one feature region; The third processing module is used to perform discrimination calculation on the at least one characteristic area according to the calculation index corresponding to the target fish to be monitored, so as to obtain an assessment result of the ecological health status of the target area.

10. A monitoring device for fish habitats, characterized in that: include: at least one processor; and a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can perform the fish habitat monitoring method as described in any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method for monitoring fish habitats according to any one of claims 1 to 8.

12. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for monitoring fish habitats according to any one of claims 1 to 8 is implemented.

Images