A fly identification system based on 3D imaging

Through a fly identification system based on 3D imaging, the feature pixel point extraction and dynamic shooting parameter adjustment of multi-frame and multi-angle images are solved, and the problem of inability to adjust the focal length and exposure time in the prior art is achieved efficient and accurate fly Identify the effect.

CN118658178BActive Publication Date: 2024-12-03FUJIAN INT TRAVEL HEALTH CARE CENT
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410720151.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-06-05
Publication Date
2024-12-03
Estimated Expiration
2044-06-05

AI Technical Summary

Technical Problem

In the existing image-based fly identification technology, the focal length and exposure time of the camera cannot be adjusted according to actual conditions, resulting in the inability to ensure the acquisition of clear three-dimensional image data, affecting the accuracy of fly identification.

Method used

Design a fly identification system based on 3D imaging, including a cloud monitoring center, data acquisition module, associated feature area extraction module, shooting parameter matching module and 3D imaging analysis module, through feature pixel points extraction and localization of multi-frame and multi-angle images Aggregation, similar clustering and interference removal, dynamically adjust shooting parameters, construct a 3D imaging model of flies and perform feature vector matching, judge the identification is successful or resolution update is performed.

Benefits of technology

Ensure sufficiently detailed three-dimensional data is obtained, and the accuracy, stability and efficiency of fly species identification are improved, and the impact of resolution adjustment on image clarity in traditional technologies is avoided.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118658178B_ABST
    Figure CN118658178B_ABST
Patent Text Reader

Abstract

A fly identification system based on 3D imaging relates to the field of image processing technology, and comprises a cloud monitoring center, wherein the cloud monitoring center is communicatively connected with a data acquisition module, a correlation feature region extraction module, a shooting parameter matching module and a 3D imaging analysis module; multiple frames and multiple angle images in a target area are collected; feature regions of the multiple frames and multiple angle images are extracted, similarity clustering and interference elimination are performed on the feature regions, and the correlation feature regions in the multiple frames and multiple angle grayscale images are obtained; the shooting parameters of the data acquisition module are dynamically adjusted according to the correlation feature regions in the multiple frames and multiple angle grayscale images; a 3D imaging model of flies in the target area is constructed, and a vector data set of feature parts to be identified of the 3D imaging model of flies is extracted for feature vector matching, and whether the identification is successful is judged according to the matching result; if the identification is not successful, a dynamic resolution update operation is performed to ensure the accuracy, stability and efficiency of fly species identification.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of image processing, and in particular to a fly identification system based on 3D imaging. Background Art

[0002] In the existing image-based fly identification technology, the focal length and exposure time of the camera are often not adjusted according to the actual situation, but the resolution of the camera is directly adjusted, which may not guarantee the acquisition of clear three-dimensional image data. The resolution adjustment affects the number of pixels in the image, but does not affect the clarity or depth information of the image. To obtain clear three-dimensional image data, it is usually necessary to ensure that the focal length and exposure time of the camera are suitable for the size and characteristics of the target object to ensure that sufficient details and depth information are captured. In order to solve the above technical problems, a fly identification system based on 3D imaging is now provided. Summary of the invention

[0003] In order to solve the above technical problems, the object of the present invention is to provide a fly identification system based on 3D imaging, including a cloud monitoring center, wherein the cloud monitoring center is communicatively connected with a data acquisition module, an associated feature region extraction module, a shooting parameter matching module and a 3D imaging analysis module;

[0004] The data acquisition module is used to acquire multi-frame multi-angle images within the target area;

[0005] The associated feature region extraction module is used to extract feature pixel points of multi-frame multi-angle images, locally aggregate the feature pixel points, generate feature regions, perform similarity clustering and interference elimination on the feature regions, and obtain associated feature regions in multi-frame multi-angle grayscale images;

[0006] The shooting parameter matching module is used to dynamically adjust the shooting parameters of the data acquisition module according to the associated feature areas in the multi-frame multi-angle grayscale images;

[0007] The 3D imaging analysis module is used to construct a 3D imaging model of flies in the target area, and extract a vector data set of feature parts to be identified of the 3D imaging model of flies for feature vector matching, and determine whether the identification is successful based on the matching result. If the identification is not successful, a dynamic resolution update operation is performed.

[0008] Furthermore, the process of extracting characteristic pixel points of multi-frame multi-angle images by the associated characteristic region extraction module includes:

[0009] The multi-frame multi-angle images of the target area are denoised and grayscaled to obtain multi-frame multi-angle grayscale images, and a multi-angle background grayscale image in the target area is set. The image alignment technology is used to align the pixels of each frame of the multi-angle grayscale image with the multi-angle background grayscale image respectively, and the grayscale values ​​of the pixels at the corresponding positions of each frame of the multi-angle grayscale image and the multi-angle background grayscale image are subtracted to obtain the grayscale difference value of the pixel points. A grayscale difference threshold is preset, and the pixel points whose grayscale difference values ​​are greater than the grayscale difference threshold are marked as feature pixels.

[0010] Furthermore, the associated feature region extraction module locally aggregates the feature pixels to generate a feature region, which includes:

[0011] A feature pixel point in each frame of the multi-angle grayscale image is randomly selected and marked as a target feature pixel point, and it is determined whether the adjacent pixel points of the target feature pixel point are feature pixels. If the adjacent pixel points are feature pixels, the adjacent pixel points are marked as target feature pixels, and it is determined whether the adjacent pixel points of the target feature pixel point are feature pixels, until there are no feature pixels adjacent to the target feature pixel point, a feature region is constructed according to the marked target feature pixel point, and when the construction of the above feature region is completed, any feature pixel point except the feature region in each frame of the multi-angle grayscale image is re-selected as the target feature pixel point, and the feature region is re-constructed, and so on, until all feature pixels in each frame of the multi-angle grayscale image are marked as target feature pixels.

[0012] Furthermore, the process of the associated feature region extraction module performing similarity clustering and interference elimination on the feature regions to obtain associated feature regions in the multi-frame multi-angle grayscale images includes:

[0013] The texture structure of each feature area in the multi-frame multi-angle grayscale image is obtained by using the gray-level co-occurrence matrix technology, and the size of each feature area is obtained at the same time. The texture structure and size of each feature area in the first frame multi-angle grayscale image in the multi-frame multi-angle grayscale image are used as the standard matching feature set of the second frame multi-angle grayscale image. The texture structure and size of each feature area in the second frame multi-angle grayscale image are feature matched with the standard matching feature set to obtain the texture structure similarity and size difference value of each feature area in the second frame multi-angle grayscale image and each feature area in the standard matching feature set;

[0014] A texture structure similarity threshold and a size difference threshold are preset, and if the texture structure similarity between a feature region in the sub-frame multi-angle grayscale image and a feature region in the standard matching feature set is greater than or equal to the texture structure similarity threshold and the size difference value is less than or equal to the size difference threshold, the feature region is marked as an associated feature region;

[0015] If the texture structure similarity between the feature region in the sub-frame multi-angle grayscale image and each feature region in the standard matching feature set is less than the texture structure similarity threshold or the size difference value is greater than the size difference threshold, the feature region is marked as an interference region and the interference region is removed;

[0016] The texture structure and size of the associated feature area in the multi-angle grayscale image of the next frame are used as the standard matching feature set of the multi-angle grayscale image of the next frame, and the associated feature area in the multi-angle grayscale image of the next frame is obtained. If there is an interference area in the multi-angle grayscale image of the next frame, the interference area is eliminated, and so on, until the associated feature area in the multi-angle grayscale image of the last frame is obtained, and the associated feature area in the multi-angle grayscale image of the last frame is used as the associated feature area in the multi-frame multi-angle grayscale images.

[0017] Furthermore, the process of the shooting parameter matching module dynamically adjusting the shooting parameters of the data acquisition module according to the associated feature areas in the multi-frame multi-angle grayscale images includes:

[0018] According to the size of the associated feature area in the multi-frame multi-angle grayscale image, the size of the fly corresponding to the associated feature area is obtained, the geometric center of the associated feature area is marked as the associated feature point, and the Euclidean distance between the associated feature point of each frame of the multi-angle grayscale image and the associated feature point of the previous frame of the multi-angle grayscale image is obtained, and the average movement speed of the flies corresponding to the associated feature area is obtained according to the Euclidean distance;

[0019] A focal length comparison table and a shooting frame rate comparison table are pre-constructed, and the multi-angle focal length is obtained according to the fly size in the multi-frame multi-angle grayscale image and the focal length comparison table, and the multi-angle shooting frame rate is obtained according to the average movement speed of the flies in the multi-frame multi-angle grayscale image and the shooting frame rate comparison table, and the adjusted shooting parameters are generated according to the multi-angle focal length and the multi-angle shooting frame rate.

[0020] Furthermore, the process of the 3D imaging analysis module constructing a 3D imaging model of flies in the target area and extracting a vector data set of feature parts to be identified of the 3D imaging model of flies for feature vector matching includes:

[0021] Acquire multiple frames and multiple angles of images in the target area acquired by the data acquisition module according to the adjusted shooting parameters, construct a 3D imaging model of flies in the target area according to the multiple frames and multiple angles of images, construct a fly feature extraction model, import the fly feature extraction model according to the fly 3D imaging model to perform feature extraction, obtain a number of vector data of feature parts to be identified, and construct a vector data set of feature parts to be identified according to the number of vector data of feature parts to be identified;

[0022] A fly feature database is preset, and the fly feature database includes feature part vector data sets corresponding to different fly types. The feature part vector data set to be identified is matched with the feature part vector data set of each fly type one by one, and the comprehensive similarity of each fly type is obtained. The fly type with the highest comprehensive similarity is obtained, and the fly type is marked as a first fly type. A comprehensive similarity threshold is preset, and it is determined whether the comprehensive similarity of the first fly type is greater than the comprehensive similarity threshold. If it is greater, it means that the identification is successful. If not, a dynamic resolution update operation is performed.

[0023] Furthermore, the 3D imaging analysis module matches the feature vector data set of the feature part to be identified with the feature vector data set of each fly type one by one to obtain the comprehensive similarity of each fly type, including:

[0024] Obtain the characteristic part types corresponding to several characteristic part vector data in the characteristic part vector data set corresponding to each fly type, and the characteristic part types corresponding to several characteristic part vector data to be identified in the characteristic part vector data set to be identified, perform feature vector comparison between the characteristic part vector data of the same characteristic part type and the characteristic part vector data to be identified, obtain the feature vector similarity corresponding to the characteristic part type, and so on, obtain the feature vector similarity corresponding to each characteristic part type corresponding to each fly type, and obtain the comprehensive similarity of each fly type according to the feature vector similarity corresponding to each characteristic part type corresponding to each fly type.

[0025] Furthermore, the process of the 3D imaging analysis module performing a dynamic resolution update operation includes:

[0026] Obtain the similarity of feature vectors corresponding to each feature part type corresponding to the first fly type, obtain the feature part type with the lowest feature vector similarity, obtain the coupling coefficient between the feature part type and each shooting angle, screen out the shooting angle with the largest coupling coefficient with the feature part type, mark the shooting angle as the first shooting angle, determine the resolution according to the similarity of the feature vectors corresponding to the feature part type, and adjust the resolution of the first shooting angle according to the resolution.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] 1. Compared with the traditional method of directly adjusting the resolution of the camera without adjusting the focal length and exposure time of the camera, the present invention first extracts characteristic pixel points of multi-frame multi-angle images, locally aggregates the characteristic pixel points, generates characteristic regions, performs similarity clustering and interference elimination on the characteristic regions, obtains associated characteristic regions in the multi-frame multi-angle grayscale images, obtains the size and characteristics of flies according to the associated characteristic regions in the multi-frame multi-angle grayscale images, and dynamically adjusts the shooting parameters of the data acquisition module according to the size and characteristics of the flies to ensure that sufficiently detailed three-dimensional data can be obtained, which is helpful for subsequent 3D imaging and analysis.

[0029] 2. Construct a 3D imaging model of flies in the target area, and extract the vector data set of the feature parts to be identified of the 3D imaging model of flies for feature vector matching. Determine whether the identification is successful based on the matching results. If the identification is not successful, perform a dynamic resolution update operation to improve the clarity of key shooting angles and ensure the accuracy, stability and efficiency of fly species identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is a schematic diagram of a fly identification system based on 3D imaging according to an embodiment of the present application. DETAILED DESCRIPTION

[0031] The following is a clear and complete description of the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0032] like Figure 1 As shown, a fly identification system based on 3D imaging includes a cloud monitoring center, wherein the cloud monitoring center is communicatively connected with a data acquisition module, an associated feature region extraction module, a shooting parameter matching module and a 3D imaging analysis module;

[0033] The data acquisition module is used to acquire multi-frame multi-angle images within the target area;

[0034] The associated feature region extraction module is used to extract feature pixel points of multi-frame multi-angle images, locally aggregate the feature pixel points, generate feature regions, perform similarity clustering and interference elimination on the feature regions, and obtain associated feature regions in multi-frame multi-angle grayscale images;

[0035] The shooting parameter matching module is used to dynamically adjust the shooting parameters of the data acquisition module according to the associated feature areas in the multi-frame multi-angle grayscale images;

[0036] The 3D imaging analysis module is used to construct a 3D imaging model of flies in the target area, and extract a vector data set of feature parts to be identified of the 3D imaging model of flies for feature vector matching, and determine whether the identification is successful based on the matching result. If the identification is not successful, a dynamic resolution update operation is performed.

[0037] It should be further explained that, in the specific implementation process, the present invention collects multiple frames of multi-angle images in each target area through multiple high-resolution cameras to construct 3D imaging of flies. The shooting angles of the multi-angle images include:

[0038] Horizontal viewing angle: capture images at all angles in the horizontal direction, including images of the front of the fly to capture the head and facial features of the fly; images of the side of the fly to capture the body outline and side features of the fly; images of the back of the fly to capture features such as the back and wings of the fly;

[0039] Bird's-eye view: Capture images shot from above, looking down, to capture the overall shape and outline of the fly;

[0040] Upward perspective: capture images taken from below to capture the fly's lower features and body structure;

[0041] The combination of these multi-angle images can provide a more comprehensive and three-dimensional perspective, which is helpful for subsequent 3D imaging and analysis. In the present invention, the comparison process between each multi-angle image is a comparison between images at the same angle, for example, image comparison between overhead perspectives and image comparison between upward perspectives.

[0042] It should be further explained that, in a specific implementation process, the process of extracting characteristic pixel points of multi-frame multi-angle images by the associated feature region extraction module includes:

[0043] Denoising and grayscale preprocessing are performed on the multi-frame multi-angle images of the target area to obtain multi-frame multi-angle grayscale images. It should be noted that the image data captured by the data acquisition module may have noise problems. Therefore, in one embodiment of the present invention, after the image data is obtained, it is necessary to perform image preprocessing operations on it to improve the quality of the image data so as to obtain image data for subsequent analysis. Improving image quality through image preprocessing operations is an image processing method well known to those skilled in the art. In the embodiment of the invention, the noise in the image is removed by a Gaussian filtering method, and then the image data is equalized to improve the contrast within the image, and the image data is grayscaled to convert the color image into a grayscale image. The grayscale process is usually achieved by averaging the weights of the color channels, wherein denoising and equalization processing, semantic segmentation network and grayscale are technical means well known to those skilled in the art, and the specific steps are not repeated here. Other image preprocessing means may also be used in other embodiments of the present invention, which are not limited here.

[0044] A multi-angle background grayscale image is set in the target area, and the image alignment technology is used to align the pixels of each frame of the multi-angle grayscale image with the multi-angle background grayscale image. The grayscale values ​​of the pixels at the corresponding positions of each frame of the multi-angle grayscale image and the multi-angle background grayscale image are subtracted to obtain the grayscale difference value of the pixel points. A grayscale difference threshold is preset, and the pixel points whose grayscale difference values ​​are greater than the grayscale difference threshold are marked as feature pixels.

[0045] It should be further explained that, in a specific implementation process, the associated feature region extraction module locally aggregates feature pixels to generate a feature region, which includes:

[0046] A feature pixel point in each frame of the multi-angle grayscale image is randomly selected and marked as a target feature pixel point, and it is determined whether the adjacent pixel points of the target feature pixel point are feature pixels, and the adjacent pixel points are specifically eight adjacent pixel points in the up, down, left, right and diagonal directions of the target feature pixel point. If the adjacent pixel points are feature pixels, the adjacent pixel points are marked as target feature pixels, and it is continuously determined whether the adjacent pixel points of the target feature pixel point are feature pixels, until there are no feature pixels in the adjacent pixel points of the target feature pixel point, a feature region is constructed according to the marked target feature pixel points, and when the construction of the above feature region is completed, any feature pixel point except the feature region in each frame of the multi-angle grayscale image is re-selected as the target feature pixel point, and the feature region is re-constructed, and so on, until all feature pixels in each frame of the multi-angle grayscale image are marked as target feature pixels.

[0047] It should be further explained that, in a specific implementation process, the associated feature region extraction module performs similarity clustering and interference elimination on the feature regions, and the process of obtaining associated feature regions in multi-frame multi-angle grayscale images includes:

[0048] The texture structure of each feature area in the multi-frame multi-angle grayscale image is obtained by using the grayscale co-occurrence matrix technology. It should be noted that the specific process of using the grayscale co-occurrence matrix technology to obtain the texture structure of each feature area in the multi-frame multi-angle grayscale image is to calculate the grayscale co-occurrence matrix around each pixel of the surface image data for each angle collected by sliding the window. The specific details are as follows: the grayscale level of the surface image data is quantized into discrete grayscale levels, the grayscale value range is divided into several levels, for example, an 8-bit image can be divided into 16, 32, and 64 levels, and the parameters required to define the grayscale co-occurrence matrix, including distance (d) and direction (θ), are determined according to actual application requirements. The specific steps will not be repeated. Subsequently, based on the calculated grayscale co-occurrence matrix, a series of texture structure features are extracted, and a texture structure is constructed according to the texture structure features. The texture structure values ​​in the texture structure include:

[0049] Contrast: It is a statistical feature that describes the contrast of pixels at different gray levels in an image. It is a feature that measures the roughness of the image texture and reflects the clarity of the image and the depth of the texture grooves. The deeper the texture grooves, the greater the contrast and the clearer the visual effect. On the contrary, when the contrast is small, the grooves are shallow and the effect is blurred.

[0050] Energy: describes the uniformity of the grayscale distribution of pixels in an image, measures the randomness contained in the image, and shows the complexity of the image. When all values ​​of the co-occurrence matrix are equal or the pixel values ​​show the greatest randomness, the entropy is the largest;

[0051] Homogeneity: describes the similarity of the gray levels of adjacent pixels in the image; reflects the homogeneity of the image texture and measures the local changes in the image texture. A large value indicates that there is a lack of changes between different areas of the image texture and the local texture is very uniform.

[0052] Inverse variance: reflects the clarity and regularity of the texture. The texture is clear and regular.

[0053] At the same time, the size of each feature area is obtained, and the texture structure and size of each feature area of ​​the first frame multi-angle grayscale image in the multi-frame multi-angle grayscale image are used as the standard matching feature set of the second frame multi-angle grayscale image. The texture structure and size of each feature area in the second frame multi-angle grayscale image are feature matched with the standard matching feature set to obtain the texture structure similarity and size difference value of each feature area in the second frame multi-angle grayscale image and each feature area in the standard matching feature set;

[0054] The calculation formula for obtaining the texture structure similarity between each feature area in the sub-frame multi-angle grayscale image and each feature area in the standard matching feature set is:

[0055] ;

[0056] in, It represents the texture structure similarity between the i-th feature region in the multi-angle grayscale image of the sub-frame and the j-th feature region in the standard matching feature set. represents the contrast between the local area of ​​the i-th pixel and the local area of ​​the j-th pixel, represents the energy of the i-th feature region in the multi-angle grayscale image of the sub-frame, represents the energy of the jth feature region in the standard matching feature set, represents the homogeneity of the i-th feature region in the multi-angle grayscale image of the sub-frame, represents the homogeneity of the jth feature region in the standard matching feature set, represents the inverse variance of the i-th feature region in the multi-angle grayscale image of the sub-frame, represents the inverse variance of the jth feature region in the quasi-matching feature set, , , represents the weight factor;

[0057] A texture structure similarity threshold and a size difference threshold are preset. If the texture structure similarity between a feature region in the sub-frame multi-angle grayscale image and a feature region in the standard matching feature set is greater than or equal to the texture structure similarity threshold and the size difference value is less than or equal to the size difference threshold, the feature region is marked as an associated feature region. If the texture structure similarity between the feature region in the sub-frame multi-angle grayscale image and each feature region in the standard matching feature set is less than the texture structure similarity threshold or the size difference value is greater than the size difference threshold, the feature region is marked as an interference region and the interference region is removed.

[0058] The texture structure and size of the associated feature area in the multi-angle grayscale image of the next frame are used as the standard matching feature set of the multi-angle grayscale image of the next frame, and the associated feature area in the multi-angle grayscale image of the next frame is obtained. If there is an interference area in the multi-angle grayscale image of the next frame, the interference area is eliminated, and so on, until the associated feature area in the multi-angle grayscale image of the last frame is obtained, and the associated feature area in the multi-angle grayscale image of the last frame is used as the associated feature area in the multi-frame multi-angle grayscale images.

[0059] It should be further explained that, in a specific implementation process, the process of the shooting parameter matching module dynamically adjusting the shooting parameters of the data acquisition module according to the associated feature areas in the multi-frame multi-angle grayscale images includes:

[0060] According to the size of the associated feature area in the multi-frame multi-angle grayscale image, the size of the fly corresponding to the associated feature area is obtained, the geometric center of the associated feature area is marked as the associated feature point, and the Euclidean distance between the associated feature point of each frame of the multi-angle grayscale image and the associated feature point of the previous frame of the multi-angle grayscale image is obtained, and the average movement speed of the flies corresponding to the associated feature area is obtained according to the Euclidean distance;

[0061] The calculation formula for obtaining the average movement rate of flies corresponding to the associated feature area according to the Euclidean distance is:

[0062] ;

[0063] Where v represents the average movement speed of flies, represents the Euclidean distance between the associated feature points of the k-th frame of the multi-angle grayscale image and the associated feature points of the previous frame of the multi-angle grayscale image; n represents the total number of frames of the multi-frame multi-angle grayscale image, and T represents the shooting time of the multi-frame multi-angle grayscale image;

[0064] A focal length comparison table and a shooting frame rate comparison table are pre-constructed, wherein the focal length comparison table includes focal lengths corresponding to different fly sizes, and the shooting frame rate includes shooting frame rates corresponding to different average movement rates of flies. The multi-angle focal length is obtained according to the fly sizes in the multi-frame multi-angle grayscale images and the focal length comparison table, the multi-angle shooting frame rate is obtained according to the average movement rate of the flies in the multi-frame multi-angle grayscale images and the shooting frame rate comparison table, and the adjusted shooting parameters are generated according to the multi-angle focal length and the multi-angle shooting frame rate.

[0065] It should be further explained that, in the specific implementation process, the 3D imaging analysis module constructs a 3D imaging model of flies in the target area, and extracts a vector data set of feature parts to be identified of the 3D imaging model of flies for feature vector matching, including:

[0066] Acquire multiple frames and multiple angles of images in the target area acquired by the data acquisition module according to the adjusted shooting parameters, and construct a 3D imaging model of flies in the target area according to the multiple frames and multiple angles of images. In the present invention, the 3D imaging model of flies is constructed by inputting the multiple frames and multiple angles of images into RealityCapture. RealityCapture is an image-based 3D modeling software with the characteristics of fastness and high efficiency, and can be used to generate high-quality 3D models, construct a fly feature extraction model, import the fly feature extraction model according to the fly 3D imaging model to perform feature extraction, obtain a number of vector data of feature parts to be identified, and construct a vector data set of feature parts to be identified according to the number of vector data of feature parts to be identified;

[0067] The vector data of the characteristic parts to be identified include the size parameters such as length, width, height, and color, texture structure, and other characteristics of the body, the size parameters such as length, width, thickness, and other size parameters of the wings, and the color, texture structure, and other characteristics of the wing surface, the size and geometric parameters such as the size, shape, and arrangement of the eyes, and the color, texture structure, and other characteristics of the eyes, the length, shape, number, and other parameters of the antennae, and the color, texture structure, and other characteristics of the antennae, and other organs: such as the size, shape, texture structure, and other characteristic parameters of the mouthparts, legs, etc.;

[0068] A fly feature database is preset, and the fly feature database includes feature part vector data sets corresponding to different fly types. The feature part vector data set to be identified is matched with the feature part vector data set of each fly type one by one, and the comprehensive similarity of each fly type is obtained. The fly type with the highest comprehensive similarity is obtained, and the fly type is marked as a first fly type. A comprehensive similarity threshold is preset, and it is determined whether the comprehensive similarity of the first fly type is greater than the comprehensive similarity threshold. If it is greater, it means that the identification is successful. If not, a dynamic resolution update operation is performed.

[0069] It should be further explained that, in the specific implementation process, the feature vector dataset of the feature part to be identified is matched with the feature vector dataset of the feature part of each fly type one by one, and the process of obtaining the comprehensive similarity of each fly type includes:

[0070] Obtain the characteristic part types corresponding to several characteristic part vector data in the characteristic part vector data set corresponding to each fly type, and the characteristic part types corresponding to several characteristic part vector data to be identified in the characteristic part vector data set to be identified, perform characteristic vector comparison between the characteristic part vector data of the same characteristic part type and the characteristic part vector data to be identified, obtain the characteristic vector similarity corresponding to the characteristic part type, and so on, obtain the characteristic vector similarity corresponding to each characteristic part type corresponding to each fly type, and obtain the comprehensive similarity of each fly type according to the characteristic vector similarity corresponding to each characteristic part type corresponding to each fly type;

[0071] The calculation formula for obtaining the comprehensive similarity of each fly type based on the similarity of the feature vectors corresponding to each characteristic part type of each fly type is:

[0072] ;

[0073] in, represents the comprehensive similarity of fly types, Indicates the similarity of the feature vector corresponding to the feature part type m, Represents the weight factor corresponding to the feature part type m.

[0074] It should be further explained that, in a specific implementation process, the process of the 3D imaging analysis module performing a dynamic resolution update operation includes:

[0075] Obtain the similarity of feature vectors corresponding to each feature part type corresponding to the first fly type, obtain the feature part type with the lowest feature vector similarity, obtain the coupling coefficient between the feature part type and each shooting angle, screen out the shooting angle with the largest coupling coefficient with the feature part type, mark the shooting angle as the first shooting angle, determine the resolution according to the similarity of the feature vectors corresponding to the feature part type, and adjust the resolution of the first shooting angle according to the resolution.

[0076] The specific process of obtaining the coupling coefficient between the characteristic part type and each shooting angle is as follows:

[0077] The total number of times each characteristic part type of flies appears in images at different shooting angles and the total number of times images at each shooting angle are taken during the historical acquisition process of the data acquisition module. The coupling coefficient of each shooting angle is obtained according to the total number of times each characteristic part type of flies appears in images at different shooting angles and the total number of times images at each shooting angle are taken. The calculation formula is:

[0078] ;

[0079] in, represents the coupling coefficient between the feature part type p and the shooting angle d, The total number of shots at shooting angle d, Represents the total number of times the characteristic part type p of the fly appears in the shooting angle d.

[0080] It should be further explained that, in the specific implementation process, the process of determining the resolution according to the similarity of the feature vectors corresponding to the feature part types is as follows:

[0081] A resolution recommendation model is constructed based on deep learning, and the historical comprehensive similarities of the feature part vector data sets to be identified in several target areas under different shooting parameters and different resolution conditions and the feature part vector data sets corresponding to different fly types in the fly feature database are collected, as well as the historical feature vector similarities corresponding to each feature part type in the feature part vector data set to be identified and each feature part type corresponding to different fly types, and the historical comprehensive similarities and historical feature vector similarities are used as training sets and test sets;

[0082] Input the training set into the resolution recommendation model for training until the loss function training is stable, save the model parameters, test the resolution recommendation model with the test set until it meets the preset requirements, and output the resolution recommendation model;

[0083] The feature vector similarity corresponding to the feature part type with the lowest feature vector similarity in the first fly type and the comprehensive similarity of the first fly type are input into the resolution recommendation model, and the resolution corresponding to the feature part type is output according to the resolution recommendation model.

[0084] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A fly identification system based on 3D imaging, comprising a cloud monitoring center, characterized in that: The cloud monitoring center is communicatively connected to a data acquisition module, an associated feature region extraction module, a shooting parameter matching module and a 3D imaging analysis module; The data acquisition module is used to acquire multi-frame multi-angle images within the target area; The associated feature region extraction module is used to extract feature pixel points of multi-frame multi-angle images, locally aggregate the feature pixel points, generate feature regions, perform similarity clustering and interference elimination on the feature regions, and obtain associated feature regions in multi-frame multi-angle grayscale images; The process of performing similarity clustering and interference elimination on the feature regions and obtaining the associated feature regions in the multi-frame multi-angle grayscale images includes: The texture structure of each feature area in the multi-frame multi-angle grayscale image is obtained by using the gray-level co-occurrence matrix technology, and the size of each feature area is obtained at the same time. The texture structure and size of each feature area in the first frame multi-angle grayscale image in the multi-frame multi-angle grayscale image are used as the standard matching feature set of the second frame multi-angle grayscale image. The texture structure and size of each feature area in the second frame multi-angle grayscale image are feature matched with the standard matching feature set to obtain the texture structure similarity and size difference value of each feature area in the second frame multi-angle grayscale image and each feature area in the standard matching feature set; A texture structure similarity threshold and a size difference threshold are preset, and if the texture structure similarity between a feature region in the sub-frame multi-angle grayscale image and a feature region in the standard matching feature set is greater than or equal to the texture structure similarity threshold and the size difference value is less than or equal to the size difference threshold, the feature region is marked as an associated feature region; If the texture structure similarity between the feature region in the sub-frame multi-angle grayscale image and each feature region in the standard matching feature set is less than the texture structure similarity threshold or the size difference value is greater than the size difference threshold, the feature region is marked as an interference region and the interference region is removed; The texture structure and size of the associated feature area in the multi-angle grayscale image of the next frame are used as the standard matching feature set of the multi-angle grayscale image of the next frame of the next frame, and the associated feature area in the multi-angle grayscale image of the next frame of the next frame is obtained. If there is an interference area in the multi-angle grayscale image of the next frame of the next frame, the interference area is eliminated, and so on, until the associated feature area in the multi-angle grayscale image of the last frame is obtained, and the associated feature area in the multi-angle grayscale image of the last frame is used as the associated feature area in the multi-frame multi-angle grayscale images; The shooting parameter matching module is used to dynamically adjust the shooting parameters of the data acquisition module according to the associated feature areas in the multi-frame multi-angle grayscale images, including: According to the size of the associated feature area in the multi-frame multi-angle grayscale image, the size of the fly corresponding to the associated feature area is obtained, the geometric center of the associated feature area is marked as the associated feature point, and the Euclidean distance between the associated feature point of each frame of the multi-angle grayscale image and the associated feature point of the previous frame of the multi-angle grayscale image is obtained, and the average movement speed of the flies corresponding to the associated feature area is obtained according to the Euclidean distance; Pre-constructing a focal length comparison table and a shooting frame rate comparison table, obtaining a multi-angle focal length according to the fly size in the multi-frame multi-angle grayscale image and the focal length comparison table, obtaining a multi-angle shooting frame rate according to the average movement speed of the flies in the multi-frame multi-angle grayscale image and the shooting frame rate comparison table, and generating adjusted shooting parameters according to the multi-angle focal length and the multi-angle shooting frame rate; The 3D imaging analysis module is used to construct a 3D imaging model of flies in the target area, and extract a vector data set of feature parts to be identified of the 3D imaging model of flies for feature vector matching, and determine whether the identification is successful based on the matching result. If the identification is not successful, a dynamic resolution update operation is performed.

2. A fly identification system based on 3D imaging according to claim 1, characterized in that: The process of extracting characteristic pixel points of multi-frame multi-angle images by the associated characteristic region extraction module includes: The multi-frame multi-angle images of the target area are denoised and grayscaled to obtain multi-frame multi-angle grayscale images, and a multi-angle background grayscale image in the target area is set. The image alignment technology is used to align the pixels of each frame of the multi-angle grayscale image with the multi-angle background grayscale image respectively, and the grayscale values ​​of the pixels at the corresponding positions of each frame of the multi-angle grayscale image and the multi-angle background grayscale image are subtracted to obtain the grayscale difference value of the pixel points. A grayscale difference threshold is preset, and the pixel points whose grayscale difference values ​​are greater than the grayscale difference threshold are marked as feature pixels.

3. A fly identification system based on 3D imaging according to claim 2, characterized in that: The process of locally aggregating feature pixels by the associated feature region extraction module to generate feature regions includes: A feature pixel point in each frame of the multi-angle grayscale image is randomly selected and marked as a target feature pixel point, and it is determined whether the adjacent pixel points of the target feature pixel point are feature pixels. If the adjacent pixel points are feature pixels, the adjacent pixel points are marked as target feature pixels, and it is determined whether the adjacent pixel points of the target feature pixel point are feature pixels, until there are no feature pixels adjacent to the target feature pixel point, a feature region is constructed according to the marked target feature pixel point, and when the construction of the above feature region is completed, any feature pixel point except the feature region in each frame of the multi-angle grayscale image is re-selected as the target feature pixel point, and the feature region is re-constructed, and so on, until all feature pixels in each frame of the multi-angle grayscale image are marked as target feature pixels.

4. A fly identification system based on 3D imaging according to claim 3, characterized in that: The process of the 3D imaging analysis module constructing a 3D imaging model of flies in the target area and extracting a vector data set of feature parts to be identified of the 3D imaging model of flies for feature vector matching includes: Acquire multiple frames and multiple angles of images in the target area acquired by the data acquisition module according to the adjusted shooting parameters, construct a 3D imaging model of flies in the target area according to the multiple frames and multiple angles of images, construct a fly feature extraction model, import the fly feature extraction model according to the fly 3D imaging model to perform feature extraction, obtain a number of vector data of feature parts to be identified, and construct a vector data set of feature parts to be identified according to the number of vector data of feature parts to be identified; A fly feature database is preset, and the fly feature database includes feature part vector data sets corresponding to different fly types. The feature part vector data set to be identified is matched with the feature part vector data set of each fly type one by one, and the comprehensive similarity of each fly type is obtained. The fly type with the highest comprehensive similarity is obtained, and the fly type is marked as a first fly type. A comprehensive similarity threshold is preset, and it is determined whether the comprehensive similarity of the first fly type is greater than the comprehensive similarity threshold. If it is greater, it means that the identification is successful. If not, a dynamic resolution update operation is performed.

5. A fly identification system based on 3D imaging according to claim 4, characterized in that: The 3D imaging analysis module matches the feature vector data set of the feature part to be identified with the feature vector data set of each fly type one by one to obtain the comprehensive similarity of each fly type, including: Obtain the characteristic part types corresponding to several characteristic part vector data in the characteristic part vector data set corresponding to each fly type, and the characteristic part types corresponding to several characteristic part vector data to be identified in the characteristic part vector data set to be identified, perform feature vector comparison between the characteristic part vector data of the same characteristic part type and the characteristic part vector data to be identified, obtain the feature vector similarity corresponding to the characteristic part type, and so on, obtain the feature vector similarity corresponding to each characteristic part type corresponding to each fly type, and obtain the comprehensive similarity of each fly type according to the feature vector similarity corresponding to each characteristic part type corresponding to each fly type.

6. A fly identification system based on 3D imaging according to claim 5, characterized in that: The process of the 3D imaging analysis module performing a dynamic resolution update operation includes: Obtain the similarity of feature vectors corresponding to each feature part type corresponding to the first fly type, obtain the feature part type with the lowest feature vector similarity, obtain the coupling coefficient between the feature part type and each shooting angle, screen out the shooting angle with the largest coupling coefficient with the feature part type, mark the shooting angle as the first shooting angle, determine the resolution according to the similarity of the feature vectors corresponding to the feature part type, and adjust the resolution of the first shooting angle according to the resolution.