Fishing gear space distribution positioning method based on unmanned aerial vehicle identification

Abstract

The invention relates to the field of fishing gear positioning, and discloses a fishing gear spatial distribution positioning method based on unmanned aerial vehicle identification, which comprises the following steps: front and rear end configuration is performed on an unmanned aerial vehicle, the flight latitude and longitude of the unmanned aerial vehicle after front and rear end configuration and image acquisition accuracy are ensured, and route planning and water area image acquisition are performed on the unmanned aerial vehicle; and finally, based on the acquired image feature information, matching with the standard feature information of the fishing gear, and identifying and determining the longitude and latitude of the fishing gear. According to the method, the image of the fishing gear in the water area can be collected through the unmanned aerial vehicle, the fishing gear is positioned based on the image recognition mode, the monitoring efficiency is greatly improved, compared with a traditional manual patrol mode, unmanned aerial vehicle monitoring can greatly reduce the labor cost and the time cost, and errors and omissions possibly occurring during manual patrol are avoided.

Description

Technical Field

[0001] The present invention relates to the field of fishing gear positioning, in particular to a method for positioning the spatial distribution of fishing gear based on UAV recognition. Background Art

[0002] When fishing gear is in use, it may damage fishery resources and the ecological environment. In order to effectively supervise and manage the use of fishing gear and prevent further damage to fishery resources and the ecological environment, it is necessary to perform spatial distribution positioning processing on it, such as using UAVs for spatial distribution positioning. UAVs for spatial distribution positioning can quickly cover large sea areas and monitor the laying situation of fishing gear in real time, greatly improving the monitoring efficiency. Compared with the traditional manual inspection method, UAV monitoring can greatly reduce labor costs and time costs, and avoid errors and omissions that may occur during manual inspections. Therefore, a method for positioning the spatial distribution of fishing gear based on UAV recognition is proposed. Summary of the Invention

[0003] The present invention overcomes the deficiencies of the prior art and provides a method for positioning the spatial distribution of fishing gear based on UAV recognition.

[0004] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0005] The first aspect of the present invention provides a method for positioning the spatial distribution of fishing gear based on UAV recognition, including the following steps:

[0006] Configure the front-end system and the back-end system of the UAV respectively to obtain a target-configured UAV;

[0007] Plan the flight path of the target-configured UAV and collect images of all positions in the target water area through the target-configured UAV;

[0008] Perform image analysis on all sub-images of the water area, extract the image features of the water area target, and match the image features of the water area target with the standard feature information of the fishing gear to identify and determine the longitude and latitude of the fishing gear.

[0009] Further, in a preferred embodiment of the present invention, the step of configuring the front-end system and the back-end system of the UAV respectively to obtain a target-configured UAV is specifically as follows:

[0010] Determine the UAV usage budget, and select the UAV with the highest camera resolution within the UAV usage budget and calibrate it as the target UAV;

[0011] Perform front-end system configuration inside the target unmanned aerial vehicle (UAV). Specifically, the front-end system is configured to collect samples using the image acquisition module inside the target UAV to obtain sample images, and preset the minimum image resolution. If the resolution of the sample image is less than the minimum image resolution, adjust the parameters of the image acquisition module to ensure that the resolution of the sample image is not less than the minimum image resolution;

[0012] Meanwhile, the front-end system configuration also includes GPS calibration for the target UAV. The GPS calibration process is as follows: first, obtain the control device of the target UAV, calibrate it as the UAV control device, place the target UAV in an area with known longitude and latitude coordinates, and at the same time, check the displayed longitude and latitude coordinates of the target UAV in the UAV control device. If the displayed longitude and latitude coordinates of the target UAV are not equal to the known longitude and latitude coordinates, perform automatic adjustment of the longitude and latitude coordinates of the target UAV until the displayed longitude and latitude coordinates of the target UAV are equal to the known longitude and latitude coordinates;

[0013] After performing the front-end system configuration on the target UAV, perform back-end system configuration on the target UAV. Specifically, the back-end system configuration is to configure an image processing module and a fishing gear recognition module;

[0014] Among them, the configuration of the fishing gear recognition module is to connect to the big data network in the fishing gear recognition module, retrieve the standard feature information of all fishing gears in the big data network, and at the same time store the standard feature information of all fishing gears in the fishing gear recognition module, so that the target UAV can perform information recognition on the feature information of the collected fishing gears;

[0015] When the target UAV completes the front-end system configuration and the back-end system configuration, the target configured UAV is obtained.

[0016] Furthermore, in a preferred embodiment of the present invention, perform route planning on the target configured UAV, and use the target configured UAV to collect images of all positions in the target water area. Specifically:

[0017] Mark the water area where the fishing gear is located as the target water area, and determine the basic water area information of the target water area. Among them, the basic water area information includes the area, shape, and depth of the water area;

[0018] Determine the surrounding information of the target water area. Among them, the surrounding information includes the terrain conditions and electromagnetic environment conditions around the target water area;

[0019] Introduce the A* algorithm, and based on the basic water area information and surrounding information of the target water area, construct a three-dimensional space in the fishing gear recognition module, and construct a three-dimensional model of the target water area in it. Among them, different positions in the three-dimensional model of the target water area are marked, and coordinate points are set at different positions. At the same time, determine the maximum acquisition range when the target configured UAV collects images, and mark it as the maximum image acquisition range;

[0020] Determine the coordinate points of the starting position of the UAV, and plan all paths from the coordinate points of the starting position of the UAV to the end point in the three-dimensional model of the target water area through the A* algorithm, and label them as candidate navigation paths. Among them, the end point and the coordinate points of the starting position of the UAV are equal;

[0021] Among all the candidate navigation paths, based on the maximum range of image acquisition, select the path that traverses the entire target water area based on the maximum range of image acquisition and can collect images of all positions within the target water area, and label it as the secondary candidate navigation path;

[0022] Determine the coordinate points with abnormal surrounding terrain conditions and abnormal electromagnetic environment conditions in the three-dimensional model of the target water area, label them as non-coincident coordinate points, and among all the secondary candidate navigation paths, screen out the secondary candidate navigation paths that coincide with the non-coincident coordinate points, and after screening, select the secondary candidate navigation path with the shortest path among the remaining secondary candidate navigation paths, and label it as the target navigation path;

[0023] Import the target navigation path into the target-configured UAV, and control the target-configured UAV to navigate based on the target navigation path. At the same time, during the navigation process, based on the target-configured UAV, collect images of all positions within the target water area, and label them as water area sub-images.

[0024] Further, in a preferred embodiment of the present invention, the image analysis of all water area sub-images, extracting the water area target image features, and matching the water area target image features with the standard feature information of the fishing gear to identify and determine the longitude and latitude of the fishing gear is specifically as follows:

[0025] In the image processing module, perform image preprocessing on all water area sub-images. Among them, the image preprocessing is to first perform image gray conversion on the water area sub-images to obtain water area gray sub-images, and introduce a wavelet filter, and perform wavelet decomposition filtering processing on the water area gray sub-images through the wavelet filter;

[0026] Among them, the wavelet decomposition filtering processing is to pour the water area gray sub-images into the wavelet filter for inner product operation to realize the high and low frequency decomposition of the water area gray sub-images, obtain the approximation coefficient and detail coefficient of the water area gray sub-images, preset the initial coefficient in the wavelet filter, multiply the approximation coefficient and detail coefficient of the water area gray sub-images with the initial coefficient respectively, and extract every other new data obtained by multiplication, and label it as inner product data;

[0027] Extract all inner product data for data multi-layer segmentation and image multi-layer reconstruction to obtain the water area gray sub-images after wavelet decomposition filtering, and define them as water area gray filtered sub-images;

[0028] Introduce the Scale-Invariant Feature Transform (SIFT) method to extract features from the water area grayscale filtered sub-image, obtain the image features of the water area grayscale filtered sub-image, and label them as the water area target image features;

[0029] Match the water area target image features with the standard feature information of the fishing gear to identify and determine the longitude and latitude of the fishing gear.

[0030] Furthermore, in a preferred embodiment of the present invention, the introduction of the Scale-Invariant Feature Transform (SIFT) method to extract features from the water area grayscale filtered sub-image, obtain the image features of the water area grayscale filtered sub-image, and label them as the target image features is specifically as follows:

[0031] Construct multi-resolution of the water area grayscale filtered sub-image to obtain water area grayscale filtered sub-images with different resolutions. Based on the water area grayscale filtered sub-images with different resolutions, construct a water area grayscale filtered sub-image pyramid, where the resolutions of the water area grayscale filtered sub-images in different layers of the water area grayscale filtered sub-image pyramid are different;

[0032] Assign standard deviations to different layers of the water area grayscale filtered sub-image pyramid and ensure that the standard deviations of different layers are different to obtain a multi-scale water area grayscale filtered sub-image pyramid, and subtract adjacent two-layer images in the multi-scale water area grayscale filtered sub-image pyramid to obtain a Difference-of-Gaussians (DoG) pyramid;

[0033] Introduce a 3*3*3 pixel positioning window to locate extreme pixel points on the images of different layers of the Difference-of-Gaussians (DoG) pyramid. Obtain a set of extreme pixel points on the images of different layers of the Difference-of-Gaussians (DoG) pyramid, acquire the pixel values of the extreme pixel points, preset a target pixel threshold, and filter out the extreme pixel points whose pixel values are not within the target pixel threshold to obtain a filtered set of extreme pixel points;

[0034] On the images of different layers of the Difference-of-Gaussians (DoG) pyramid, calculate the pixel gradients of the filtered set of extreme pixel points through the Sobel operator to generate a gradient direction histogram of the filtered set of extreme pixel points, label it as the target gradient direction histogram, determine the highest peaks of different extreme pixel points on the target gradient direction histogram, and mark the gradient direction at the highest peak as the key direction;

[0035] Construct directed edges based on the key direction in the filtered set of extreme pixel points to obtain the directed edge features of the blurred image. Connect and construct the directed edge features of the blurred image on the images of different layers of the Difference-of-Gaussians (DoG) pyramid, and fuse the obtained image features to obtain the image features of the water area grayscale filtered sub-image, and label them as the water area target image features.

[0036] Further, in a preferred embodiment of the present invention, the matching of the water area target image features with the standard feature information of the fishing gear, identifying and determining the longitude and latitude of the fishing gear is specifically as follows:

[0037] In the three-dimensional model of the target water area, perform feature mapping on the water area target image features to generate mapping coordinates of the water area target image features in the three-dimensional model of the target water area;

[0038] Pour the standard feature information of the fishing gear into the three-dimensional model of the target water area, calculate the Hamming distance between the standard feature information of the fishing gear and the water area target image features, and preset a standard Hamming distance threshold;

[0039] If the Hamming distance between the standard feature information of the fishing gear and the water area target image features is within the Hamming distance threshold, then in the three-dimensional model of the target water area, calibrate the mapping coordinates corresponding to the water area target image features as the fishing gear coordinates;

[0040] Based on the fishing gear coordinates, locate the longitude and latitude of the fishing gear in the target water area and calibrate it as the target longitude and latitude of the fishing gear.

[0041] The second aspect of the present invention also provides a fishing gear spatial distribution positioning system based on UAV recognition. The spatial distribution positioning system includes a memory and a processor. The memory stores a spatial distribution positioning method. When the spatial distribution positioning method is executed by the processor, the following steps are realized:

[0042] Configure the front-end system and the back-end system of the UAV respectively to obtain a target-configured UAV;

[0043] Plan the flight path of the target-configured UAV and collect images of all positions in the target water area through the target-configured UAV;

[0044] Perform image analysis on all water area sub-images, extract water area target image features, and match the water area target image features with the standard feature information of the fishing gear to identify and determine the longitude and latitude of the fishing gear.

[0045] The present invention solves the technical defects existing in the background technology. The present invention has the following beneficial effects: Configure the front-end and back-end of the UAV to ensure the accurate longitude and latitude of the UAV flight and image acquisition after the front-end and back-end configuration, plan the flight path of the UAV and collect water area images, and finally match the collected image feature information with the standard feature information of the fishing gear to identify and determine the longitude and latitude of the fishing gear. The present invention can collect images of fishing gear in the water area through the UAV and realize the positioning of the fishing gear based on image recognition, greatly improving the monitoring efficiency. Compared with the traditional manual inspection method, the UAV monitoring can greatly reduce the labor cost and time cost and avoid the errors and omissions that may occur during manual inspection. Description of the Drawings

[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings of embodiments can be obtained based on these drawings.

[0047] Figure 1 The flowchart of the fishing gear spatial distribution positioning method based on UAV recognition is shown;

[0048] Figure 2 The flowchart of the method for determining the longitude and latitude of the fishing gear is shown;

[0049] Figure 3 The program view of the fishing gear spatial distribution positioning system based on UAV recognition is shown. Detailed implementation manners

[0050] In order to more clearly understand the above objects, features and advantages of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners. It should be noted that, without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other.

[0051] Many specific details are set forth in the following description in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.

[0052] Figure 1 The flowchart of the fishing gear spatial distribution positioning method based on UAV recognition is shown, including the following steps:

[0053] S102: Configure the front-end system and the back-end system of the UAV respectively to obtain the target-configured UAV;

[0054] S104: Plan the flight path of the target-configured UAV, and collect images of all positions in the target water area through the target-configured UAV;

[0055] S106: Analyze all sub-images of the water area, extract the water area target image features, and match the water area target image features with the standard feature information of the fishing gear to identify and determine the longitude and latitude of the fishing gear.

[0056] Further, in a preferred embodiment of the present invention, the configuring the front-end system and the back-end system of the UAV respectively to obtain the target-configured UAV is specifically:

[0057] Determine the budget for using the drone, and select the drone with the highest camera resolution within the budget for using the drone, and calibrate it as the target drone;

[0058] Perform front-end system configuration in the target drone. Among them, the front-end system is configured to collect samples from the image acquisition module in the target drone to obtain sample images, and preset the minimum image resolution. If the resolution of the sample image is less than the minimum image resolution, adjust the parameters of the image acquisition module to ensure that the resolution of the sample image is not less than the minimum image resolution;

[0059] At the same time, the front-end system configuration also includes GPS calibration of the target drone. The GPS calibration is to first obtain the control device of the target drone, calibrate it as the drone control device, and place the target drone in an area with known longitude and latitude coordinates. At the same time, view the longitude and latitude coordinates of the target drone displayed in the drone control device. If the displayed longitude and latitude coordinates of the target drone are not equal to the known longitude and latitude coordinates, perform automatic longitude and latitude coordinate adjustment on the target drone until the displayed longitude and latitude coordinates of the target drone are equal to the known longitude and latitude coordinates;

[0060] After performing the front-end system configuration on the target drone, perform the back-end system configuration on the target drone. Among them, the back-end system configuration is to configure the image processing module and configure the fishing gear recognition module;

[0061] Among them, the configuration of the fishing gear recognition module is to connect to the big data network in the fishing gear recognition module, retrieve the standard feature information of all fishing gears in the big data network, and store the standard feature information of all fishing gears in the fishing gear recognition module at the same time, so that the target drone can identify the feature information of the collected fishing gear;

[0062] When the target drone completes the front-end system configuration and the back-end system configuration, the target configured drone is obtained.

[0063] It should be noted that the fishing gear in this application is general fishing gear. Before the fishing gear is identified by the drone, the drone needs to be accurately configured, including accurately positioning the real-time longitude and latitude of the drone, aiming to ensure that the drone will not deviate, resulting in deviation when generating the longitude and latitude of the fishing gear in the drone and causing the spatial positioning of the fishing gear to shift. At the same time, an image processing module and a fishing gear recognition module need to be configured in the drone. Because after the water area image is collected, the image processing module is required to perform image preprocessing on the water area image to obtain image features, and the image features and the standard features of the fishing gear are compared in the fishing gear recognition module for the purpose of positioning the fishing gear. At the same time, the sampling analysis of the sample image can ensure that the resolution of the collected image is higher than the preset value, ensure that the collected image is clear, and reduce the recognition deviation. The flight precision positioning of the drone is achieved through the GPS positioning calibration method to ensure that the known coordinates of the drone are equal to the real-time coordinates, and the standard feature information of the fishing gear is stored in the fishing gear recognition module. After completing the front-end system configuration and the back-end system configuration, the target configured drone is obtained.

[0064] Further, in a preferred embodiment of the present invention, the route planning is performed on the target configured drone, and the images of all positions in the target water area are collected by the target configured drone, specifically:

[0065] The water area where the fishing gear is located is designated as the target water area, and the basic water area information of the target water area is determined, where the water area basic information includes the area, shape, and depth of the water area;

[0066] The surrounding information of the target water area is determined, where the surrounding information includes the terrain conditions and electromagnetic environment conditions around the target water area;

[0067] The A* algorithm is introduced, and based on the basic water area information and surrounding information of the target water area, a three-dimensional space is constructed in the fishing gear recognition module, and a three-dimensional model of the target water area is constructed therein. Different positions in the target water area three-dimensional model are marked, and coordinate points are set at different positions. At the same time, the maximum acquisition range when the target configured drone collects images is determined and designated as the maximum image acquisition range;

[0068] The coordinate point of the starting position of the drone is determined, and all paths from the coordinate point of the starting position of the drone to the end point are planned in the three-dimensional model of the target water area through the A* algorithm and designated as the candidate navigation paths, where the end point and the coordinate point of the starting position of the drone are equal;

[0069] Among all the candidate navigation paths, based on the maximum image acquisition range, a path that traverses the entire target water area based on the maximum image acquisition range and can collect images of all positions in the target water area is selected and designated as the secondary candidate navigation path;

[0070] Identify the coordinate points with abnormal surrounding terrain conditions and abnormal electromagnetic environment conditions in the three-dimensional model of the target water area, mark them as non-coincident coordinate points, and in all the secondary candidate navigation paths, filter out the secondary candidate navigation paths that coincide with the non-coincident coordinate points. After filtering, select the secondary candidate navigation path with the shortest path among the remaining secondary candidate navigation paths and mark it as the target navigation path;

[0071] Import the target navigation path into the target configured unmanned aerial vehicle (UAV), and control the target configured UAV to navigate based on the target navigation path. At the same time, during the navigation process, based on the target configured UAV, collect images of all positions in the target water area and mark them as water area sub-images.

[0072] It should be noted that the purpose of determining the basic information and surrounding information of the target water area is to construct a three-dimensional model of the target water area, and the purpose of constructing the model is to achieve the precise positioning of fishing gear. The A* algorithm is a heuristic search algorithm that combines the advantages of the Dijkstra algorithm and the efficiency of heuristic methods. It uses a cost function to estimate the cost from the current node to the target node and selects the optimal path accordingly. The A* algorithm is commonly used for path search in complex environments in route planning. Through the A* algorithm, multiple flight paths of UAVs can be planned in the model, but not all of these flight paths support collecting images of all positions in the water area. Therefore, the path that can traverse the entire target water area based on the maximum range of image acquisition and can collect images of all positions in the target water area is selected. The purpose is to screen out the paths that can traverse the entire water area and can take pictures of all positions without dead angles and output them, that is, to obtain the secondary candidate navigation paths. Due to the influence of the terrain, if there is terrain on the water area that prevents the UAV from flying, or there is an area with abnormal electromagnetic intensity, the UAV cannot fly to the corresponding coordinate area. Therefore, filter out the secondary candidate navigation paths that coincide with the non-coincident coordinate points to ensure the safety of the UAV, and among the remaining paths, select the shortest path. The purpose is to save energy and efficiently take pictures. Finally, output the target navigation path. After obtaining the target navigation path, control the UAV to fly according to the target navigation path, and then the images of all positions in the target water area can be taken and marked as water area sub-images.

[0073] Figure 2 The flowchart of the method for determining the longitude and latitude of fishing gear is shown, including the following steps:

[0074] S202: Perform image analysis on all water area sub-images, extract the image features of the water area target images, and match the image features of the water area target images with the standard feature information of the fishing gear to identify and determine the longitude and latitude of the fishing gear;

[0075] S204: Introduce the Scale-Invariant Feature Transform (SIFT) method to extract the features of the water area grayscale filtered sub-images, and obtain the image features of the water area grayscale filtered sub-images, marked as target image features;

[0076] S206: Match the water area target image features with the standard feature information of the fishing gear to identify and determine the longitude and latitude of the fishing gear.

[0077] Further, in a preferred embodiment of the present invention, the image analysis of all water area sub-images, extracting the water area target image features, and matching the water area target image features with the standard feature information of the fishing gear to identify and determine the longitude and latitude of the fishing gear are specifically as follows:

[0078] In the image processing module, perform image preprocessing on all water area sub-images. Among them, the image preprocessing is as follows: First, perform image gray conversion on the water area sub-images to obtain water area gray sub-images, and introduce a wavelet filter. Perform wavelet decomposition filtering on the water area gray sub-images through the wavelet filter.

[0079] Among them, the wavelet decomposition filtering process is as follows: Pour the water area gray sub-image into the wavelet filter for inner product operation to achieve high-frequency and low-frequency decomposition of the water area gray sub-image, obtain the approximation coefficient and detail coefficient of the water area gray sub-image. Preset initial coefficients in the wavelet filter, multiply the approximation coefficient and detail coefficient of the water area gray sub-image with the initial coefficients respectively, and extract every other newly obtained multiplied data and label it as inner product data.

[0080] Extract all inner product data for multi-layer data segmentation and multi-layer image reconstruction to obtain the water area gray sub-image after wavelet decomposition filtering, defined as the water area gray filtered sub-image.

[0081] Introduce the Scale-Invariant Feature Transform (SIFT) method to extract features from the water area gray filtered sub-image to obtain the image features of the water area gray filtered sub-image, labeled as the water area target image features.

[0082] Match the water area target image features with the standard feature information of the fishing gear to identify and determine the longitude and latitude of the fishing gear.

[0083] It should be noted that before extracting the image features of the water area sub-image, comparing the features with the standard features of fishing gear, and judging the accurate positioning of fishing gear in the water area, it is necessary to preprocess the image to ensure that the features extracted from the image are accurate and prevent deviations in the positioning of fishing gear. The image preprocessing includes image grayscale processing and image noise reduction processing. The image after grayscale processing is clearer and some unnecessary noises are removed. After obtaining the grayscale image, image noise reduction still needs to be carried out to ensure that the image is not affected by noise and is more accurate. Image noise reduction is carried out by means of wavelet noise reduction. The steps of preprocessing the image by the wavelet transform algorithm usually involve the decomposition and reconstruction processes of the image. These steps are aimed at extracting the features of the image, removing noise, or compressing, etc. Among them, the approximation coefficient is low-frequency information, and the detail coefficient is high-frequency information. Each number on the initial data is multiplied by each number on the filter data. The result of taking the inner product with the low-pass filter is the approximation coefficient, and the result of taking the inner product with the high-pass filter is the detail coefficient. Multiplying every other one is to generate inner product data and achieve the purpose of data noise reduction. The data is multi-layer segmented for the inner product data obtained after wavelet transform. Based on the resolution characteristics and the localization characteristics of wavelet analysis, the domain features at different scales are segmented and a reconstruction filter is given to implement wavelet reconstruction, that is, the results of taking the inner product of the approximation coefficient and the detail coefficient are combined to obtain the reconstructed image, that is, the grayscale filtered sub-image of the water area obtained after filtering.

[0084] Furthermore, in a preferred embodiment of the present invention, the scale-invariant feature transform (SIFT) method is introduced to extract the features of the grayscale filtered sub-image of the water area, and the image features of the grayscale filtered sub-image of the water area are obtained, which are labeled as target image features. Specifically:

[0085] Construct a multi-resolution of the grayscale filtered sub-image of the water area to obtain grayscale filtered sub-images of different resolutions. Based on the grayscale filtered sub-images of different resolutions, construct a pyramid of grayscale filtered sub-images of the water area. Among them, the resolutions of the grayscale filtered sub-images of different layers in the pyramid of grayscale filtered sub-images of the water area are different;

[0086] Assign standard deviations to different layers of the pyramid of grayscale filtered sub-images of the water area and ensure that the standard deviations of different layers are different to obtain a multi-scale pyramid of grayscale filtered sub-images of the water area, and subtract the images of adjacent layers in the multi-scale pyramid of grayscale filtered sub-images of the water area to obtain a difference-of-Gaussians pyramid;

[0087] Introduce a 3*3*3 pixel point positioning window to locate the extreme pixel points on the images of different layers of the difference-of-Gaussians pyramid. An extreme pixel point set is obtained on the images of different layers of the difference-of-Gaussians pyramid. Obtain the pixel values of the extreme pixel points, preset a target pixel threshold, and filter out the extreme pixel points whose pixel values are not within the target pixel threshold to obtain a filtered extreme pixel point set;

[0088] On the images of different layers of the difference Gaussian pyramid, pixel - point gradients are calculated for the filtered set of extreme - value pixel points through the Sobel operator, generating a gradient - direction histogram of the filtered set of extreme - value pixel points, which is labeled as the target gradient - direction histogram. On the target gradient - direction histogram, the highest peaks of different extreme - value pixel points are determined, and the gradient direction at the highest peak is marked as the key direction;

[0089] Directed edges are constructed based on the key direction in the filtered set of extreme - value pixel points to obtain the directed - edge features of the blurred image. The directed - edge features of the blurred image are connected and constructed on the images of different layers of the difference Gaussian pyramid, and the obtained image features are fused to obtain the image features of the water - area gray - scale filtered sub - image, which are labeled as the water - area target image features.

[0090] It should be noted that the Scale - Invariant Feature Transform (SIFT) method is a machine - vision algorithm for detecting and describing local features in images. By constructing a Gaussian multi - scale pyramid, monitoring key points and assigning directions to key points, and generating directed edges, feature extraction is achieved. The image pyramid is a set of representations of the same image at different resolutions. Gaussian blurring needs to be performed on the pyramid, that is, different standard deviations are assigned to different layers to achieve blurred images of different scales in the same layer. Finally, combined with multi - layer and multi - scale pyramids, a multi - scale water - area gray - scale filtered sub - image pyramid is constructed. The difference between adjacent layers of the multi - scale water - area gray - scale filtered sub - image pyramid gives the difference Gaussian pyramid, which aims to provide conditions for monitoring key points. Among them, the combined key points are the feature information of fishing gear. The 3×3×3 pixel - point positioning window is a window for finding extreme values in the image. Traversing the image, it is similar to a filtering window. The 3×3×3 pixel - point positioning window is a sub - pixel - level precise positioning window. After obtaining the extreme points, the extreme points are filtered to obtain a set of key points, that is, the filtered set of extreme - value pixel points. Since the direction of the key points is not determined and the direction needs to be determined to construct directed edges, the Sobel operator is introduced to calculate the gradient to achieve the direction positioning of the key points. After constructing the histogram, the angle corresponding to the highest peak of the histogram is the direction of the key points, that is, the key direction. Directed edges are constructed based on the key direction, and feature linking is performed on all constructed directed edges. Finally, all images are fused to generate high - resolution and clear image features, that is, the water - area target image features.

[0091] Furthermore, in a preferred embodiment of the present invention, the matching of the water - area target image features with the standard feature information of fishing gear, identifying and determining the longitude and latitude of the fishing gear is specifically as follows:

[0092] In the three - dimensional model of the target water area, feature mapping is performed on the water - area target image features, so that mapping coordinates of the water - area target image features are generated within the three - dimensional model of the target water area;

[0093] Pour the standard feature information of the fishing gear into the three-dimensional model of the target water area, calculate the Hamming distance between the standard feature information of the fishing gear and the feature of the water area target image, and preset the standard Hamming distance threshold.

[0094] If the Hamming distance between the standard feature information of the fishing gear and the feature of the water area target image is within the Hamming distance threshold, then in the three-dimensional model of the target water area, calibrate the mapping coordinates corresponding to the water area target image feature as the fishing gear coordinates.

[0095] Based on the fishing gear coordinates, locate the longitude and latitude of the fishing gear in the target water area and calibrate it as the target longitude and latitude of the fishing gear.

[0096] It should be noted that in the three-dimensional model of the target water area, the purpose of feature mapping of the water area target image feature is to provide conditions for the spatial positioning of the fishing gear. Since there are coordinate data in the model, the longitude and latitude can be obtained according to the coordinate data. The Hamming distance is an algorithm for judging the similarity between data. If the Hamming distance between the feature data is within the preset range, it proves that the feature data has a high similarity. In other words, if the Hamming distance between the standard feature information of the fishing gear and the feature of the water area target image is within the Hamming distance threshold, then the mapping coordinates corresponding to the water area target image feature are the coordinates of the fishing gear.

[0097] As Figure 3 shown, the second aspect of the present invention also provides a fishing gear spatial distribution positioning system based on UAV recognition. The spatial distribution positioning system includes a memory 31 and a processor 32. The memory 31 stores a spatial distribution positioning method. When the spatial distribution positioning method is executed by the processor 32, the following steps are implemented:

[0098] Configure the front-end system and the back-end system of the UAV respectively to obtain a target-configured UAV.

[0099] Plan the flight path of the target-configured UAV and collect images of all positions in the target water area through the target-configured UAV.

[0100] Perform image analysis on all water area sub-images, extract the water area target image features, and match the water area target image features with the standard feature information of the fishing gear to identify and determine the longitude and latitude of the fishing gear.

[0101] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A method for spatial distribution and positioning of fishing gear based on drone recognition, characterized in that: The following steps are involved: Perform front-end system configuration and back-end system configuration on the UAV respectively to obtain the target configuration UAV; Plan the route for the target-configured UAV and collect images of all locations in the target waters through the target-configured UAV; Image analysis is performed on all water area sub-images to extract water area target image features, which are then matched with standard feature information of fishing gear to identify and determine the longitude and latitude of the fishing gear.

2. The method for spatial distribution and positioning of fishing gear based on drone recognition according to claim 1, characterized in that: The front-end system configuration and the back-end system configuration are respectively performed on the drone to obtain the target configuration drone, specifically: Determine the budget for drone use, and select a drone with the highest camera resolution within the budget and mark it as the target drone; Performing a front-end system configuration in the target UAV, wherein the front-end system is configured to collect samples from an image acquisition module in the target UAV to obtain a sample image and preset a minimum image resolution. If the sample image resolution is less than the minimum image resolution, adjusting parameters of the image acquisition module to ensure that the sample image resolution is not less than the minimum image resolution; At the same time, the front-end system configuration also includes GPS calibration of the target drone, wherein the GPS calibration is to first obtain the control device of the target drone, calibrate it as the drone control device, and place the target drone in an area with known longitude and latitude coordinates, and at the same time check the longitude and latitude coordinates of the displayed target drone in the drone control device. If the longitude and latitude coordinates of the displayed target drone are not equal to the known longitude and latitude coordinates, the longitude and latitude coordinates of the target drone are automatically adjusted until the longitude and latitude coordinates of the displayed target drone are equal to the known longitude and latitude coordinates; After the front-end system is configured for the target UAV, the back-end system is configured for the target UAV, wherein the back-end system is configured to configure an image processing module and a fishing gear recognition module; The configuration of the fishing gear identification module is to access the big data network in the fishing gear identification module, retrieve the standard feature information of all fishing gears in the big data network, and store the standard feature information of all fishing gears in the fishing gear identification module, so that the target drone can identify the collected feature information of the fishing gear; When the target UAV completes the front-end system configuration and the back-end system configuration, the target configured UAV is obtained.

3. The method for spatial distribution and positioning of fishing gear based on drone recognition according to claim 1, characterized in that: The route planning of the target configuration drone and the collection of images of all positions of the target waters by the target configuration drone are specifically as follows: Marking the water area where the fishing gear is located as the target water area, and determining the basic water area information of the target water area, wherein the basic water area information includes the area, shape and depth of the water area; Determine the surrounding information of the target water area, wherein the surrounding information includes the terrain conditions and electromagnetic environment conditions around the target water area; The A* algorithm is introduced, and based on the basic information of the target water area and the surrounding information, a three-dimensional space is constructed in the fishing gear identification module, and a three-dimensional model of the target water area is constructed, wherein different positions of the target water area are marked in the three-dimensional model of the target water area, and coordinate points are set at different positions, and the maximum acquisition range when the target configuration drone collects images is determined, which is calibrated as the maximum image acquisition range; Determine the coordinate point of the starting position of the UAV, and plan all paths from the coordinate point of the starting position of the UAV to the end point in the three-dimensional model of the target waters through the A* algorithm, and mark them as the selected navigation path, where the coordinate point of the end point is equal to the starting position of the UAV; Among all the candidate navigation paths, based on the maximum image acquisition range, a path that traverses the entire target waters based on the maximum image acquisition range and can acquire images of all positions in the target waters is selected and calibrated as the secondary candidate navigation path; In the three-dimensional model of the target waters, coordinate points with abnormal surrounding terrain conditions and abnormal electromagnetic environment conditions are determined and marked as non-overlapping coordinate points. Among all the secondary candidate navigation paths, the secondary candidate navigation paths that overlap with the non-overlapping coordinate points are screened out, and after the screening, the secondary candidate navigation path with the shortest path is selected from the saved secondary candidate navigation paths and marked as the target navigation path; The target navigation path is imported into the target configuration UAV, and the target configuration UAV is controlled to navigate based on the target navigation path. Meanwhile, during the navigation process, images of all positions in the target waters are collected based on the target configuration UAV and calibrated as waters sub-images.

4. The method for spatial distribution and positioning of fishing gear based on drone recognition according to claim 1, characterized in that: The image analysis is performed on all water area sub-images to extract water area target image features, and the water area target image features are matched with the standard feature information of the fishing gear to identify and determine the longitude and latitude of the fishing gear, specifically: In the image processing module, all water area sub-images are subjected to image preprocessing, wherein the image preprocessing is to firstly perform image grayscale conversion on the water area sub-image to obtain a water area grayscale sub-image, and then introduce a wavelet filter to perform wavelet decomposition filtering processing on the water area grayscale sub-image through the wavelet filter; The wavelet decomposition filtering process is as follows: pouring the water gray sub-image into the wavelet filter for inner product operation to achieve high- and low-frequency decomposition of the water gray sub-image, obtaining the approximate coefficient and detail coefficient of the water gray sub-image, presetting the initial coefficient in the wavelet filter, multiplying the approximate coefficient and detail coefficient of the water gray sub-image with the initial coefficient respectively, and extracting new data obtained by every other multiplication, and marking them as inner product data; Extract all inner product data to perform multi-layer data segmentation and multi-layer image reconstruction, and obtain the water area grayscale sub-image after wavelet decomposition filtering, which is defined as the water area grayscale filtered sub-image; The scale-invariant feature-variant (SIFT) method is introduced to extract features from the water area grayscale filter sub-image, and the image features of the water area grayscale filter sub-image are obtained and calibrated as the water area target image features. Match the image features of the water target with the standard feature information of the fishing gear to identify and determine the longitude and latitude of the fishing gear.

5. The method for spatial distribution and positioning of fishing gear based on drone recognition according to claim 4, characterized in that: The scale-invariant feature-variant (SIFT) method is introduced to extract features from the water grayscale filter sub-image, and the image features of the water grayscale filter sub-image are obtained and calibrated as target image features, specifically: Performing multi-resolution construction on the water grayscale filter sub-image to obtain water grayscale filter sub-images of different resolutions, and constructing a water grayscale filter sub-image pyramid based on the water grayscale filter sub-images of different resolutions, wherein the water grayscale filter sub-images of different layers in the water grayscale filter sub-image pyramid have different resolutions; Standard deviation values ​​are assigned to different layers of the water grayscale filter sub-image pyramid, and the standard deviation values ​​of different layers are ensured to be different, so as to obtain a multi-scale water grayscale filter sub-image pyramid, and two adjacent layers of images are subtracted in the multi-scale water grayscale filter sub-image pyramid to obtain a differential Gaussian pyramid; A 3*3*3 pixel positioning window is introduced to locate extreme pixel points on images of different layers of the differential Gaussian pyramid, and extreme pixel point sets are obtained on images of different layers of the differential Gaussian pyramid. The pixel values ​​of the extreme pixel points are obtained, and a target pixel threshold is preset. The extreme pixel points whose pixel values ​​are not within the target pixel threshold are filtered to obtain a filtered extreme pixel point set. On images of different layers of the differential Gaussian pyramid, pixel gradients are calculated on the filtered extreme pixel set by using the Sobel operator to generate a gradient direction histogram of the filtered extreme pixel set, which is calibrated as a target gradient direction histogram, and the highest peaks of different extreme pixel points are determined on the target gradient direction histogram, and the gradient direction at the highest peak is marked as a key direction; Directed edges are constructed based on the key direction in the filtered extreme pixel point set to obtain directed edge features of the blurred image. The directed edge features of the blurred image are connected and constructed on images of different layers of the differential Gaussian pyramid, and the image features obtained after the construction are fused to obtain image features of the water area grayscale filtered sub-image, which are calibrated as water area target image features.

6. The method for spatial distribution and positioning of fishing gear based on drone recognition according to claim 4, characterized in that: The matching of the water target image features with the standard feature information of the fishing gear to identify and determine the longitude and latitude of the fishing gear is specifically as follows: In the target water area three-dimensional model, feature mapping is performed on the water area target image features, so that mapping coordinates of the water area target image features are generated in the target water area three-dimensional model; Pour the standard feature information of the fishing gear into the three-dimensional model of the target water area, calculate the Hamming distance between the standard feature information of the fishing gear and the image feature of the water area target, and preset the standard Hamming distance threshold; If the Hamming distance between the standard feature information of the fishing gear and the water target image feature is within the Hamming distance threshold, then in the target water three-dimensional model, the mapping coordinates corresponding to the water target image feature are calibrated as the fishing gear coordinates; Based on the coordinates of the fishing gear, the longitude and latitude of the fishing gear are located in the target waters and calibrated as the target longitude and latitude of the fishing gear.

7. The fishing gear spatial distribution positioning system based on drone recognition is characterized by: The spatial distribution positioning system includes a memory and a processor, wherein the memory stores a spatial distribution positioning method program. When the spatial distribution positioning method program is executed by the processor, the steps of the spatial distribution positioning method for fishing gear based on drone identification as described in any one of claims 1 to 6 are implemented.

Images