Insect compound eye detection method, system and equipment and storage medium

Abstract

The invention discloses an insect compound eye detection method, system and device and a storage medium, and relates to the field of computer vision and biological image processing, and the method comprises the steps: carrying out the preprocessing of an insect compound eye CT image; carrying out intensity thresholding processing on voxels in the CT image stack by utilizing a dynamic region self-adaptive thresholding strategy and a GPU (Graphic Processing Unit) multi-thread parallel computing method to obtain screened voxel data; decomposing a CT image stack into a plurality of sub-regions, and performing block processing on the screened voxel data by adopting a GPU multi-thread parallel computing architecture to construct three-dimensional point cloud data; performing coordinate transformation by utilizing a GPU parallel computing method to generate cross slice data of different depths; a CPU multi-core parallel computing method is used for conducting compound eye geometric structure analysis on the cross slice data, and a detection result of the insect compound eye structure is obtained; according to the insect compound eye detection method, system and device and the storage medium, efficient detection of large-scale insect compound eye CT images is achieved.

Description

Technical Field

[0001] The present invention relates to the fields of computer vision and biological image processing, and in particular to an insect compound eye detection method, system, device and storage medium. Background Art

[0002] Insect compound eyes are the main visual organs of many insect species. They are composed of hundreds or thousands of ommatidia, each of which has an independent light-sensing function. The structure of insect compound eyes is complex, and the arrangement of each ommatidia is crucial to the insect's visual system. Therefore, studying the structure of insect compound eyes can help us deeply understand the insect's visual mechanism and provide important references for biological research and the design of bionic visual systems.

[0003] Traditionally, researchers analyze CT images of insect compound eyes manually or semi-automatically, which is not only time-consuming but also susceptible to human error. With the advancement of computer technology, automated detection algorithms have gradually become mainstream, among which the most famous automated ommatidia detection algorithm is ODA (Ommatidia Detection Algorithm). However, the existing ODA algorithm has problems of low efficiency and poor robustness when processing large-scale CT image data, especially when facing data noise or complex structural changes, the performance is significantly reduced. Summary of the invention

[0004] Based on the technical problems existing in the background technology, the present invention proposes an insect compound eye detection method, system, equipment and storage medium, which realizes the efficient detection of large-scale insect compound eye CT images.

[0005] The insect compound eye detection method, system, device and storage medium proposed in the present invention include: Preprocess the insect compound eye CT images; Using dynamic region adaptive thresholding strategy and GPU multi-threaded parallel computing method, the voxels in the CT image stack are subjected to intensity thresholding to obtain the filtered voxel data. Decompose the CT image stack into multiple sub-regions, and use the GPU multi-threaded parallel computing architecture to process the filtered voxel data in blocks. Each GPU thread block processes the voxel data in a sub-region to construct 3D point cloud data. The GPU parallel computing method is used to perform coordinate conversion, converting the voxels of the 3D point cloud data from the Cartesian coordinate system to the spherical coordinate system, and the multi-core CPU parallel computing method is used to perform 2D projection on the voxel data in the spherical coordinate system to generate cross-slice data at different depths; The CPU multi-core parallel computing method is used to analyze the geometric structure of the compound eye of the cross-slice data and obtain the detection results of the insect compound eye structure.

[0006] Furthermore, before the voxel intensity thresholding, the CT image is adaptively binarized using a GPU parallel computing method, specifically including: Calculate the global statistical characteristics of the CT image stack and preliminarily determine the binarization threshold range; Using CUDA parallel kernel functions, local contrast calculation is performed on each pixel of the CT image, and the threshold is dynamically adjusted based on the local window: GPU parallel thread blocks are used to perform binarization on all pixels in parallel to generate binary data, which are then converted into 0 or 1 using a set threshold to distinguish between the target area and the background area.

[0007] Furthermore, in the dynamic adjustment of the threshold based on the local window, the threshold The formula is as follows: ; in, are the coordinates of the voxel in the local window, and are the mean and standard deviation in the local window, is the adaptive adjustment factor.

[0008] Furthermore, the filtered voxel data is processed in blocks using the GPU multi-threaded parallel computing architecture, and each GPU thread block processes the voxel data in a sub-region to construct the three-dimensional point cloud data, specifically: Using a hash index structure, each filtered voxel is stored in a 3D point cloud structure, and outliers are removed through neighborhood statistical filtering to obtain the spatial coordinates and intensity value of each voxel; The spatial coordinates and intensity value of each voxel are stored in the GPU global memory as three-dimensional point cloud data, and each GPU thread processes the three-dimensional point cloud data in parallel.

[0009] Furthermore, the coordinate conversion is performed by using the GPU parallel computing method to convert the voxels of the three-dimensional point cloud data from the Cartesian coordinate system to the spherical coordinate system, specifically including: Each sub-region of the original CT image stack is processed in parallel through GPU multithreading, and the spatial coordinates and intensity values ​​of each voxel are extracted from it; GPU multi-threaded parallel computing is used to transform the voxels in the sub-region from the Cartesian coordinate system to the spherical coordinate system.

[0010] Furthermore, the compound eye geometric structure analysis of the cross-slice data was performed using the CPU multi-core parallel computing method to obtain the detection results of the insect compound eye structure, specifically: When the voxel data is expressed in a spherical coordinate system, a spherical space is formed, and the entire spherical space is divided into multiple spherical windows, each of which corresponds to a local area on the spherical space; A 2D morphological detection algorithm based on morphological detection and geometric feature analysis was used to identify the ommatidium center of the cross-slice data in each spherical window; The ommatidia clusters were classified by calculating the Euclidean distance between the center points of each ommatidia. The geometric characteristics of each ommatidium cluster and the angular relationship between adjacent ommatidia are calculated to obtain the spatial arrangement and geometric morphology of the compound eyes.

[0011] Insect compound eye detection system, including image acquisition module, voxel enhancement threshold module, 3D point cloud construction module, cross-slice data construction module and geometric structure analysis module; The image acquisition module is used to preprocess the insect compound eye CT image; The voxel enhancement threshold module is used to perform intensity thresholding on the voxels in the CT image stack using a dynamic region adaptive thresholding strategy and a GPU multi-threaded parallel computing method to obtain filtered voxel data; The 3D point cloud construction module is used to decompose the CT image stack into multiple sub-regions and use the GPU multi-threaded parallel computing architecture to process the filtered voxel data in blocks. Each GPU thread block processes the voxel data in a sub-region to construct 3D point cloud data. The cross-slice data construction module is used to perform coordinate conversion using GPU parallel computing methods, converting the voxels of the three-dimensional point cloud data from the Cartesian coordinate system to the spherical coordinate system, and using multi-core CPU parallel computing methods to perform two-dimensional projection on the voxel data in the spherical coordinate system to generate cross-slice data at different depths; The geometric structure analysis module is used to analyze the geometric structure of compound eyes on cross-slice data using the CPU multi-core parallel computing method to obtain the detection results of the insect compound eye structure.

[0012] Furthermore, the three-dimensional point cloud construction module specifically includes a filtering module and a storage module; The filtering module is used to store each filtered voxel into a 3D point cloud structure using a hash index structure, remove outliers through neighborhood statistical filtering, and obtain the spatial coordinates and intensity value of each voxel; The storage module is used to store the spatial coordinates and intensity value of each voxel as three-dimensional point cloud data in the GPU global memory, and each GPU thread processes the three-dimensional point cloud data in parallel.

[0013] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the insect compound eye detection method as described above when executing the computer program.

[0014] A computer-readable storage medium stores a plurality of classification programs, wherein the plurality of classification programs are used to be called by a processor and execute the insect compound eye detection method as described above.

[0015] A person of ordinary skill in the art can understand that all or part of the steps of implementing the above-mentioned method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above-mentioned method embodiment; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk, etc., various media that can store program codes.

[0016] The advantages of the insect compound eye detection method, system, device and storage medium provided by the present invention are: the insect compound eye detection method, system, device and storage medium provided in the structure of the present invention, by combining computer vision technology and parallel computing technology for the first time and using them for insect compound eye detection, through image preprocessing, feature extraction, cluster analysis and other means, integrate GPU parallel computing and multi-core CPU parallel computing to solve the computing bottleneck and accuracy problems encountered by the prior art in processing insect compound eye images. In addition, this embodiment proposes an efficient multi-stage image processing framework based on parallel computing, which can realize automatic detection and analysis of the ommatidia structure in insect compound eyes, especially under the interference of complex background and noise, and can still maintain a high detection accuracy; thereby realizing efficient detection of large-scale insect compound eye CT images. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a structural schematic diagram of the present invention; Figure 2 Schematic diagram of the effect of insect ommatidia detection algorithm, (a) schematic diagram of the projection result of the three-dimensional point cloud data of the insect compound eye in the spherical coordinate system, (b) schematic diagram of the precise distribution of the extracted ommatidia center points after completing the ommatidia detection. DETAILED DESCRIPTION

[0018] Below, the technical solution of the present invention is described in detail through specific embodiments. Many specific details are set forth in the following description to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific implementation disclosed below.

[0019] like Figures 1 to 2As shown, the insect compound eye detection method proposed in the present invention pre-processes the insect compound eye CT image, and then uses the insect compound eye detection algorithm integrating GPU parallel computing and multi-core CPU parallel computing to process the insect compound eye so as to efficiently and accurately extract the insect compound eye structure; the detection method specifically includes steps 1 to 5: Step 1: Collect and preprocess insect compound eye CT images to construct an experimental dataset; Specifically, this embodiment first scans the insect compound eye through a high-resolution CT device to obtain the original CT image data of the insect compound eye. These CT images come from multiple samples of different insect species, and the image data covers a variety of resolutions and scales, reflecting the actual situation of different structural complexities and noise interference. Subsequently, the collected CT image data is preprocessed, such as normalization processing such as cropping, to construct a complete experimental data set for the detection and analysis experiment of the insect compound eye structure. Among them, the experimental data set includes the original CT image data and the corresponding image parameter information, such as voxel size, scanning angle and resolution, etc., to provide a data basis for the subsequent insect compound eye detection algorithm.

[0020] Step 2: Using the dynamic region adaptive thresholding strategy and GPU multi-threaded parallel computing method, the voxels in the CT image stack are subjected to intensity thresholding to obtain the screened voxel data; First, GPU multi-threaded parallel computing is used to process voxel data block by block, adaptive thresholding is achieved, and isolated noise points are removed through morphological optimization. Subsequently, multi-core CPU parallel optimization further analyzes the global voxel distribution and adjusts the local threshold to improve detection accuracy and connectivity. Finally, the filtered valid voxel data is stored in the GPU shared memory, providing efficient input for subsequent 3D point cloud construction.

[0021] Step 3: Decompose the CT image stack into multiple sub-regions, and use the GPU thread-based parallel execution model to process the filtered voxel data in blocks. Each GPU thread block processes the voxel data in a sub-region to construct 3D point cloud data. The GPU multi-threaded parallel computing architecture is used to accelerate data processing, while combined with CPU multi-core parallel computing to further optimize data organization and analysis.

[0022] First, the GPU multi-stream parallel architecture divides the computing task into multiple independent data streams through Streams in the CUDA kernel function, realizing the simultaneous processing of multiple sub-areas. Then, in each data stream, the GPU multi-threaded parallel computing further accelerates the voxel data processing. Each GPU thread block is responsible for the construction of a 3D point cloud in a sub-area, and uses local neighborhood statistics to remove outliers, improving the stability and accuracy of the 3D point cloud data.

[0023] In this embodiment, when using local neighborhood statistics to remove outliers, the GPU multi-stream parallel architecture is combined to accelerate the calculation, which has the advantages of: cleaning up 3D point cloud data and removing isolated noise points or misdetected voxels generated during CT scanning. Improving the stability of the point cloud, ensuring the accuracy of the final extracted ommatidium center point position, and avoiding isolated points from affecting structural analysis. Combined with the GPU multi-stream parallel architecture, local neighborhood calculations are performed in parallel in CUDA calculations, enhancing computing efficiency and reducing computing overhead.

[0024] After the GPU calculation is completed, the CPU multi-core parallel computing is responsible for global optimization of the 3D point cloud data, including point cloud sparsity analysis, morphological filtering, and outlier removal to ensure data quality while reducing GPU computing pressure. Ultimately, the optimized high-quality 3D point cloud data is used as input for subsequent spatial coordinate transformation, providing more accurate geometric information for subsequent insect compound eye detection.

[0025] Step 4: Use GPU parallel computing method to perform coordinate conversion, convert the voxels of the three-dimensional point cloud data from the Cartesian coordinate system to the spherical coordinate system, use multi-core CPU parallel computing method to perform two-dimensional projection on the voxel data in the spherical coordinate system, and generate cross-slice data of different depths; First, GPU multi-threaded parallel computing is responsible for converting the 3D point cloud data from the Cartesian coordinate system to the spherical coordinate system, ensuring efficient execution of the coordinate transformation process and adapting to the spherical structure characteristics of insect compound eyes. Subsequently, CPU multi-core parallel computing further optimizes the converted data, reconstructs the data through batch data mapping and adaptive sampling strategies, and improves data uniformity and computing stability.

[0026] The batch data mapping described in this embodiment is to perform efficient mapping and storage optimization between different data structures, while combining GPU and CPU acceleration processing, specifically including: Coordinate system conversion batch mapping: Mapping process from Cartesian coordinate system to spherical coordinate system. The GPU performs this conversion in parallel, but the CPU needs to further optimize the data to make it more structured.

[0027] Point cloud data optimization: The converted spherical coordinate data may be unevenly distributed, so the CPU processes this data in batches to make the data structure more conducive to subsequent calculations (such as sorting and indexing).

[0028] Data storage format adjustment: Due to the different processing methods of GPU and CPU, after the GPU calculation is completed, the data may need to be converted in batches for more efficient reading, indexing or storage on the CPU.

[0029] This embodiment combines the adaptive sampling strategy with the GPU and CPU to achieve an efficient combination of GPU parallel coordinate transformation + CPU adaptive sampling optimization, thereby accelerating the optimization of insect compound eye point clouds. The adaptive sampling strategy can adopt uniform resampling, K-nearest neighbor sampling, error-based adaptive sampling, octree / quadtree adaptive sampling, etc.

[0030] After the coordinate transformation is completed, cross-slice data is generated in conjunction with CPU parallel computing to ensure that each slice is evenly distributed and the resolution is adapted to subsequent detection tasks. Dynamic adjustments are made to large-scale data to optimize the number of slices and hierarchical information, ensuring that the spatial structure of the insect compound eye can be accurately captured.

[0031] Among them, dynamic adjustment is carried out for large-scale data. The core of dynamic adjustment lies in: Adaptive Layer Partitioning: Dynamically adjust the interval of slices according to the density distribution of the compound eye structure. If the density of ommatidia in a certain area is high, increase the number of slices in that area to improve the resolution; for areas with low ommatidia density, reduce the number of slices to optimize the allocation of computing resources.

[0032] Geometry-Aware Slicing: Calculate the curvature distribution of the insect compound eye, increase the number of slices in areas with high curvature (dramatic morphological changes), and reduce the number of slices in areas with low curvature (gentle morphology) to ensure that the geometric features of the detection area can be fully captured.

[0033] Ommatidial Distribution-Based Slicing: After performing ommatid center detection, the slice distribution is adaptively adjusted according to the spatial distribution pattern of the ommatids. For example, if the ommatids are evenly arranged, the slices are sliced ​​at a fixed interval; if the ommatids in some areas are dense (such as the center of the field of view of some insect compound eyes), the slice interval is reduced to improve local detection accuracy; if the ommatids in some areas are sparse, the slice interval is increased to reduce computational redundancy.

[0034] Resolution Adaptation: Adjust the spatial resolution of the slice based on the requirements of the target detection task: if the detection task has high requirements for the compound eye structure, increase the slice resolution (increase pixel density). If the detection task focuses on the overall trend, reduce the resolution to improve the computational efficiency.

[0035] Dynamic hierarchical merging: After the initial slicing is completed, the multi-core CPU is used to parallelly analyze the similarity of the slices and merge the slices with large information redundancy to reduce storage and computing overhead while retaining the necessary geometric information.

[0036] The key to dynamically adjusting the cross-slice data in this embodiment is to adjust the number of slice layers based on the density of the ommatidia to improve the detection accuracy of high-density areas. Optimize the slice interval based on geometric curvature to adapt to the three-dimensional structural changes of the compound eye. Adaptive optimization based on resolution requirements to ensure the reasonable allocation of computing resources. Use CPU parallel computing to perform hierarchical merging to reduce redundant data and improve storage and computing efficiency.

[0037] Dynamic adjustment of cross-slice data ensures that the spatial structure of insect compound eyes can be accurately captured and improves the accuracy of subsequent ommatidia detection and geometric analysis.

[0038] Step 5: Use the CPU multi-core parallel computing method to analyze the compound eye geometric structure of the cross-section data to obtain the detection result of the insect compound eye structure; The cross-section data is used to reconstruct the structure of the insect compound eye, feature extraction is performed through GPU parallel computing, and the final structure analysis and ommatidia detection are completed using CPU multi-core parallel computing.

[0039] First, GPU parallel computing is responsible for morphological filtering and preliminary edge detection of cross-slice data, enhancing ommatidia area features and removing background noise. Then, CPU multi-core parallel computing performs morphological analysis and geometric clustering to accurately detect ommatidia centers and determine ommatidia cluster structure based on neighborhood search.

[0040] On this basis, CPU parallel computing is used to further calculate the geometric characteristics of the ommatidia, including lens diameter, lens area and axial vector information, and the Euclidean distance and spherical angle between the ommatidia are calculated using the optimized KD-Tree neighborhood search algorithm, ultimately constructing a complete three-dimensional spatial model of the insect compound eye.

[0041] Multi-core CPUs are used to parallelly calculate the geometric relationship of the ommatidia units of insect compound eyes, and GPU computing is used to optimize spatial arrangement analysis. First, GPU parallel computing is used to perform preliminary screening and neighborhood search of ommatidia centers, and the topological structure of ommatidia clusters is determined by accelerated nearest neighbor calculation. Subsequently, CPU multi-core parallel computing is used to calculate ommatidia geometric features, including lens diameter, area, and axial vector, and the Euclidean distance between the centers of each ommatidia is calculated using an optimized KD-Tree structure.

[0042] On this basis, CPU parallel computing is used to further calculate the spherical angle between the centers of the ommatidia, obtain the spatial arrangement relationship of the ommatidia units in the insect compound eye, and combine global geometric analysis to identify special structural features, ultimately completing the accurate detection and analysis of the overall structure of the insect compound eye.

[0043] According to steps one to five, existing studies have explored the application of parallel computing in image processing, but have not yet effectively applied it to the automated processing of large-scale insect compound eye CT images. Therefore, this embodiment combines computer vision technology and parallel computing technology for the first time and uses them for insect compound eye detection. By means of image preprocessing, feature extraction, cluster analysis, etc., it integrates GPU parallel computing and multi-core CPU parallel computing to solve the computational bottlenecks and precision problems encountered by the prior art in processing insect compound eye images. In addition, this embodiment proposes an efficient multi-stage image processing framework based on parallel computing, which can automatically detect and analyze the ommatidia structure in insect compound eyes, especially under the interference of complex backgrounds and noise, and can still maintain a high detection accuracy; it significantly improves the processing speed and robustness of traditional insect compound eye detection methods, and can efficiently process large-scale insect compound eye CT image data.

[0044] (A) In one embodiment, between step 1 and step 2, a GPU-accelerated CT image binarization process is provided, specifically: calculating the global statistical characteristics of the CT image stack and preliminarily determining the binarization threshold range; using the CUDA parallel kernel function to perform local contrast calculation on each pixel of the CT image, and dynamically adjusting the threshold based on the local window; using the GPU parallel thread block to perform binarization on all pixels in parallel to generate binarized data, and using the set threshold to convert the binarized data into 0 or 1 to distinguish the target area from the background area.

[0045] In order to improve the accuracy of compound eye detection, this embodiment adopts a binarization strategy with different threshold values, and dynamically adjusts the threshold according to different image characteristics. This embodiment uses the dynamic region-based adaptive thresholding (Dynamic Region-based Adaptive Thresholding), which calculates the statistical information of different regions (such as the local mean) and standard deviation ), dynamically adjust the threshold The formula is as follows: ; in, are the coordinates of the voxel in the local window, and are the mean and standard deviation in the local window, respectively, calculated based on the statistical characteristics of the local voxels. is the adaptive adjustment factor.

[0046] The binarization process of each CT image is processed in parallel on the GPU, and the parallel computing power of the GPU is used to significantly improve the processing speed. The goal of the binarization process is to extract effective information from the CT image, remove background noise, and highlight the ommatidia area of ​​the compound eye to ensure that there is no interference from irrelevant areas in subsequent steps. The binarized CT image is then used to perform operations such as voxel intensity thresholding of the CT image stack, construction of 3D point cloud data, and construction of cross-slice data.

[0047] (B) In one embodiment, in step 2, using a dynamic region adaptive thresholding strategy and a GPU multi-threaded parallel computing method, intensity thresholding is performed on voxels in the CT image stack to obtain filtered voxel data, specifically: This embodiment proposes an ultra-high-speed parallel voxel intensity thresholding method, which combines the dynamic adaptive thresholding method with GPU parallel computing acceleration, greatly improving the processing efficiency of complex high-resolution 3D image data. Compared with traditional methods, the method of this embodiment can achieve millisecond-level computing performance on millions of voxel data while ensuring accurate screening of tiny regions of interest (ROI).

[0048] In this embodiment, the dynamic region adaptive thresholding (DRA-T) method adaptively adjusts the intensity threshold of voxels according to different tissue density, material and noise distribution. The method includes the following key points: (b1) Dynamic adjustment of multi-level thresholds: Calculation of local mean based on local statistical analysis (LSA) and variance , dynamically set threshold : ; in, is the spatial coordinate of the voxel in the Cartesian coordinate system, It is an adaptive adjustment factor to ensure that the target area can be accurately extracted under different imaging conditions.

[0049] It should be noted that the dynamic adjustment threshold set during binarization in (A) and the dynamic adjustment threshold set in the multi-level threshold dynamic adjustment here are both threshold processing of voxel data, because they both calculate an adaptive threshold based on local statistics and then perform binary classification. The only difference is that during binarization, the threshold adjustment is performed on the voxels in the local window, while in the multi-level threshold dynamic adjustment, the threshold adjustment is performed on the voxels in the Cartesian coordinate system. Therefore, this embodiment uses the same parameter representation for the formula of the dynamic adjustment thresholds in the two places.

[0050] It is understandable that the binarization process and the multi-level threshold dynamic adjustment process are similar in adjustment methods, but they are different in purpose and implementation details and cannot be completely equivalent. Binarization is a pre-step to multi-level threshold dynamic adjustment (i.e., voxel thresholding), which is used to preliminarily screen the target area, while voxel thresholding is a more fine-grained screening in three-dimensional space to remove low-confidence voxels and optimize the reconstruction of the compound eye structure. In addition, although the calculation formulas of binarization and multi-level threshold dynamic adjustment are similar, binarization focuses on pixel-level 2D preprocessing, while multi-level threshold dynamic adjustment focuses on 3D voxel-level screening, and the two are in a progressive relationship.

[0051] (b2) GPU multi-threaded parallel acceleration: CUDA kernels are used to parallelize the calculation of each voxel, so that the threshold calculation of a large number of voxels can be completed in a very short time.

[0052] (b3) Local Morphological Enhancement (LME): After voxel thresholding, morphological optimization is performed on the edge region to remove isolated noise points while retaining small target structures to improve the accuracy of subsequent ommatidia detection.

[0053] (C) In one embodiment, step 3, decomposing the CT image stack into multiple sub-regions, and using a GPU multi-threaded parallel computing architecture to process the filtered voxel data in blocks, each GPU thread block processes the voxel data in a sub-region, and constructs three-dimensional point cloud data, specifically: After completing the voxel screening, this embodiment constructs high-precision three-dimensional point cloud data through the following steps: (c1) Voxel coordinate mapping: Using an efficient spatial hashing structure, each filtered voxel Stored in a 3D point cloud structure to avoid redundant calculations.

[0054] (c2) Boundary refinement and denoising: Through Neighbor Statistical Filtering, outliers are removed to obtain the spatial coordinates and intensity values ​​of each voxel: ; in, For neighboring points, ensure that the density of the point cloud is uniform and the outline is clear. is the coordinate of a point in the point cloud after filtering, that is, the required three-dimensional point cloud data, is the coordinate of the current point (i.e. the center point). It is the reference point for filtering and is used to calculate the relative positions of other points in its neighborhood. is an index variable, which is used to traverse all neighboring points in the neighborhood of the current point, indicating the neighboring points, is the number of neighboring points in the neighborhood of the current point.

[0055] (c3) To speed up the whole process, the present invention adopts GPU multi-threaded parallel computing architecture (GPU thread-basedparallel execution model) for accelerated processing; Block parallel computing: Decompose CT image data into multiple sub-regions, and each GPU thread block independently processes a sub-region to improve parallel computing efficiency.

[0056] Shared memory optimization: GPU shared memory is used to store temporary calculation results, reduce memory access latency, and achieve low-overhead parallel processing.

[0057] According to (c1) to (c3), the GPU multi-threaded parallel computing architecture is used to construct three-dimensional point cloud data, which improves the computing efficiency and realizes efficient detection of large-scale insect compound eye CT images.

[0058] (D) In ​​one embodiment, step 4, using a GPU parallel computing method to perform coordinate conversion, converting the voxels of the three-dimensional point cloud data from a Cartesian coordinate system to a spherical coordinate system, and using a multi-core CPU parallel computing method to perform two-dimensional projection on the voxel data in the spherical coordinate system to generate cross-slice data of different depths, specifically: In this stage of data processing, this embodiment converts the 3D voxel data in the original image into cross-slice data that is convenient for subsequent analysis by means of parallel computing. The specific steps are as follows: (d1) Data loading and voxel extraction: First, load the original CT image stack and extract the spatial coordinates of each voxel from it. This process divides the image stack into blocks through parallel computing, and each GPU is responsible for processing the data of a sub-area, thereby achieving fast data reading and storage.

[0059] Through GPU parallel computing, the CT image stack is processed in blocks to improve data loading speed and reduce I / O bottlenecks; using a block-based approach, each GPU processes a sub-area, making data storage more efficient and reducing storage overhead; only valid voxels are stored to avoid loading the entire CT image data, thereby optimizing computing resource utilization.

[0060] (d2) Coordinates are transformed into spherical coordinates: For each voxel coordinate , convert it from the Cartesian coordinate system to the spherical coordinate system, and use the spherical coordinate system to describe the position of the ommatidia, making the subsequent spatial analysis easier and more accurate. In particular, when analyzing the structure of the compound eye, the spherical coordinate system can provide a more intuitive geometric relationship. The specific conversion formula is as follows: , , ; in, represents the radius of the voxel from the origin, is the azimuth, is the polar angle. Through this transformation, the position of each point has a clearer geometric meaning in the spherical coordinate system, which is more suitable for spatial analysis of compound eye ommatidia.

[0061] The spherical coordinate system is more consistent with the geometric characteristics of insect compound eyes, making the spatial relationship of small eye clusters more intuitive and reducing the complexity of subsequent analysis. Using GPU multi-threaded computing, each thread independently completes the voxel coordinate transformation to achieve efficient coordinate conversion.

[0062] In addition, the spherical coordinate system enables compound eye structures of different sizes and shapes to be analyzed in a unified coordinate system, which is suitable for compound eye structures of different scales.

[0063] (d3) Parallel processing and acceleration: During the coordinate transformation process, each thread on the GPU processes different point sets independently. During this process, each GPU thread calculates the coordinate transformation of voxels in a sub-region. This parallel processing method ensures the efficiency of data transformation and greatly shortens the calculation time.

[0064] GPU parallel computing optimization: Each GPU thread independently processes different point sets, greatly improving the throughput of data conversion, which can be accelerated by dozens of times compared to CPU serial computing. At the same time, block processing is combined with streaming computing to minimize synchronization overhead and improve overall efficiency. It is suitable for large-scale CT image data sets and can efficiently process millions of voxel data to ensure the scalability of the algorithm.

[0065] (d4) Data storage and structuring: The transformed spherical coordinate data is stored as a new data structure, where the spherical coordinates of each voxel are They are saved in a corresponding array to facilitate subsequent compound eye ommatidia detection and geometric analysis.

[0066] The spherical coordinate data is stored in an efficient index structure, which is convenient for subsequent quick query and processing, and improves computing efficiency. Avoid storing invalid information, store only key data, reduce redundant data, optimize storage usage, and reduce computing resource consumption; data structured storage enables subsequent analysis to quickly access required data, improves data access efficiency and real-time computing.

[0067] (D5) Generate cross-slice data: Generate cross-slice data at different depths based on the converted spherical coordinate system data to more clearly analyze the location and distribution of each ommatidium cluster. These slices will be dynamically adjusted according to the ommatidium distribution to ensure that the most accurate compound eye structure information can be obtained.

[0068] According to steps (d1) to (d5), by dynamically generating cross-slice data of different depths, the spatial arrangement relationship of ommatidia clusters becomes clearer, thereby enhancing the ability to analyze ommatidia distribution. Slice data of different depths provide more fine-grained information, making compound eye detection more accurate and reducing the possibility of missed detection and false detection. Slice generation is dynamically adjusted according to the ommatidia distribution to ensure that the analysis of each area has the highest accuracy and is adapted to compound eye structures of different morphologies.

[0069] (E) In one embodiment, step 5, using a CPU multi-core parallel computing method to perform compound eye geometric structure analysis on the cross-slice data to obtain the detection result of the insect compound eye structure, specifically includes: (e1) When the voxel data is expressed in a spherical coordinate system, a spherical space is formed, and the entire spherical space is divided into multiple spherical windows, each of which corresponds to a local area on the spherical space; In the insect compound eye detection algorithm, the CT image is first converted into 3D point cloud data through threshold processing and voxel screening. Next, these 3D point cloud data are converted to a spherical coordinate system. The purpose of this step is to approximate the insect compound eye as a spherical structure for analysis, where: Represents the distance from each voxel to the center of the sphere; and Represents the spatial orientation of the voxel on the sphere.

[0070] When the voxel data is represented in a spherical coordinate system, the entire spherical space is divided into many smaller local areas or "windows" for the convenience of detection and analysis. Each window corresponds to a local area on the spherical space. This locally divided area is called a "spherical window."

[0071] In each spherical window, two-dimensional morphological detection is performed independently to detect the ommatidium center in the window area. The advantages of this division method are: the voxel data in each spherical window is easier to process, reducing the amount of data and computational complexity; parallelization is effectively achieved, and each spherical window can be independently processed in parallel by different computing units (such as GPU thread blocks or CPU cores); it ensures independent detection and analysis of structural differences in different regions, improving the accuracy and robustness of detection.

[0072] (e2) Using a 2D morphological detection algorithm based on morphological detection and geometric feature analysis, the ommatidium center of the cross-slice data in each spherical window is identified; Identification of ommatidium center and local morphological operations: For each spherical window, this embodiment first performs morphological operations, such as dilation and erosion. These operations help highlight the boundaries of ommatidium and demarcate its core area. Each GPU processes different areas of the image independently. This parallel processing not only improves the processing speed, but also ensures that local features can be accurately identified.

[0073] Combination of multi-scale feature extraction and attention mechanism: In order to detect ommatidia of different sizes, this embodiment combines multi-scale feature extraction and attention mechanism. First, the image is subjected to convolution operations at multiple scales to extract features, which contain different details of the ommatidia in the image. Subsequently, the key areas are weighted through the attention mechanism, so that the system can focus more on the detection area of ​​the ommatidia, improving the detection ability of ommatidia in complex backgrounds. The feature maps of each scale are calculated in parallel by the GPU, which not only speeds up the calculation speed, but also improves the efficiency of multi-scale analysis.

[0074] Parallel processing accelerates morphological operations: Each GPU independently processes a sub-region in the CT image stack and performs morphological processing of the image. This process includes closing and opening operations, which help remove background noise and further refine the edges of the ommatidia. The parallel design allows each GPU to process morphological operations in different regions simultaneously, which significantly speeds up the entire process, especially when processing large-scale data.

[0075] (e3) Classifying ommatidia clusters by calculating the Euclidean distance between the center points of each ommatidia; In three-dimensional space, a single ommatidium usually does not exist in isolation, but together with the surrounding adjacent ommatidia form a local structural unit. This local aggregation structure is called an "ommatidium cluster". Each ommatidium cluster contains multiple adjacent ommatidia units, which have a close arrangement relationship in space and can be regarded as a subunit or substructure with a specific local geometric relationship in the overall structure of the compound eye. That is, an ommatidium cluster is a collection of multiple adjacent ommatidia, reflecting the local spatial structure of the compound eye; the angular relationship between ommatidia refers to the angle between the axes of adjacent ommatidia within the cluster and between clusters. These angles directly determine the overall spatial arrangement and visual performance of the compound eye.

[0076] Detection of ommatidia clusters and estimation of lens diameter: In order to identify and classify ommatidia clusters, this embodiment adopts a clustering method based on the distance between center points. By calculating the Euclidean distance between the center points of each ommatidia, the ommatidia can be assigned to a suitable cluster.

[0077] In this embodiment, a clustering method based on the distance between center points is used, and the following rules (1) to (5) are set in combination with the geometric features of insect compound eyes to assign ommatidia to appropriate clusters: (1) Nearest neighbor distance threshold; Calculate the Euclidean distance from the center of each ommatidium to its nearest neighbor ommatidium. If the distance is less than the set threshold, , they are considered to belong to the same cluster. It is set according to the average ommatidia distance of insect compound eyes to ensure that the division of clusters conforms to the biological characteristics of the compound eye structure.

[0078] (2) Cluster density constraints; If the number of ommatidia in a cluster is less than a minimum threshold , then the cluster may be noise or isolated points and need further analysis or elimination. If the number of ommatidia in the cluster is too small, it may be due to misdetection or segmentation error and should be corrected.

[0079] (3) Maximum cluster expansion range; Set the maximum radius of the cluster If the maximum distance from the geometric center of a cluster to the external ommatidia exceeds this threshold, it needs to be split into multiple sub-clusters. This constraint prevents over-merging and ensures that ommatidia in different regions are not misclassified into the same cluster.

[0080] (4) morphological constraints; The spatial distribution of clusters was calculated using morphological analysis. If the ommatidia arrangement direction changed suddenly (for example, the compound eyes of some insects have irregular shapes), the clusters were re-divided. The normal vector distribution of the ommatidia was calculated to ensure that the clusters had reasonable geometric consistency.

[0081] (5) Local optimization and adjustment; Combined with K-Means or DBSCAN and other clustering algorithms, the preliminary clusters are optimized and the boundary points are adjusted. If the distance between the small eyes between adjacent clusters meets the threshold, And the distribution direction is consistent, the clusters can be merged.

[0082] The lens diameter was then estimated using the average distance between ommatidium centers. , which is calculated based on the following formula: ; in, represents the average distance, Indicates A small eye center point, Indicates A small eye center point.

[0083] In order to cope with images with different complex backgrounds, this embodiment designs an adaptive threshold strategy. Based on the image features of the local area, the system dynamically adjusts the threshold to adaptively process background noise and ommatidia. In this process, each GPU independently optimizes the processing parameters according to the characteristics of the local image area to ensure that the system can automatically adapt to different environmental conditions when processing ommatidia with complex morphology.

[0084] Therefore, this embodiment first uses multi-level threshold dynamic adjustment (b1) to set the initial global threshold, and then further executes an adaptive threshold strategy inside the GPU, so that different regions can be optimized more fine-grainedly for local ommatidia morphology, thereby improving the accuracy and robustness of detection.

[0085] (e4) Calculate the geometric characteristics of each ommatidium cluster and the angular relationship between adjacent ommatidia to obtain the spatial arrangement and geometric morphology of the compound eye.

[0086] Calculation of geometric features: For each detected ommatidium cluster, its geometric features such as lens diameter, area, and vector direction are calculated. This process uses a neighborhood search method and a KD-Tree structure to optimize the distance between adjacent ommatidia. This method can efficiently calculate the geometric parameters of each ommatidium cluster, and further accelerate the process through parallel computing.

[0087] The geometric features of each ommatidium cluster are the basis for subsequent analysis, including information such as the size, distribution and arrangement of ommatidia. These geometric features are of great significance for understanding the visual mechanism of insect compound eyes.

[0088] In order to further describe the spatial arrangement and geometric morphology of insect compound eyes, the angular relationship between adjacent ommatidia is calculated. Specifically, the spherical angle between the centers of adjacent ommatidia is calculated using the angle formula in the spherical coordinate system. , the formula is as follows: ; in, is the center of the compound eye, expressed as the spherical fitting center. This process can also be accelerated by parallel computing to ensure the efficiency of multiple angle calculations. Calculating the angular relationship between adjacent ommatidia can help us better understand the geometric arrangement of compound eyes.

[0089] It should be noted that the differences between the multi-level threshold dynamic adjustment and the adaptive threshold strategy in this embodiment are as follows: Multi-level threshold dynamic adjustment (b1) sets the threshold through statistical analysis (mean + variance), which is suitable for threshold optimization in the global image range to ensure a good segmentation effect in the entire image range.

[0090] The adaptive threshold strategy is more inclined to local adaptive optimization, that is, the processing parameters are dynamically optimized in a fine-grained manner within each GPU computing unit, allowing the processing thresholds of different regions to be dynamically adjusted to adapt to the complex morphology of the ommatidia structure, and combining regional feature extraction (such as local gradient, texture, contrast, etc.) for local optimization, so that different regions can independently adapt to the complex morphology of the ommatidia structure. The adaptive threshold strategy of this embodiment is similar to the existing local adaptive threshold method, but this embodiment combines the GPU parallel computing method for dynamic adjustment, dynamically adjusts the processing parameters of different regions, and has special optimization in ommatidia structure analysis.

[0091] In the analysis of the ommatidia structure of insect compound eyes, the key to dynamically adjusting the processing parameters of different regions lies in: (1) Based on the large changes in the density and morphology of ommatidia in different regions, it is necessary to locally optimize the processing parameters; (2) Based on the fact that different thread blocks in GPU parallel computing process different regions, it is necessary to adaptively adjust the parameters of each computing unit; (3) Based on the fact that the distribution structure of ommatidia may be affected by background noise, contrast changes, etc., it is necessary to dynamically optimize the threshold.

[0092] In order to optimize the ommatidium detection effect, this embodiment uses a dynamic threshold adjustment strategy based on local area features and combines GPU parallel computing to perform regional level optimization. The specific method is as follows: ① Use multi-scale regional analysis to adaptively optimize local thresholds: In the traditional Otsu or Sauvola threshold methods, a global threshold is usually calculated based on the entire image, but the distribution density, contrast, and noise level of ommatidia in different areas of the insect compound eye structure are different, so local optimization thresholds are required. Otsu or Sauvola is an existing local adaptive threshold method. The specific steps are as follows: (1-1) Calculate local features; On the GPU, the entire compound eye area is divided into multiple sub-areas (each GPU thread block is responsible for one sub-area).

[0093] Each GPU thread block calculates: local mean , local variance , local contrast , local ommatidium density , the local noise level (calculated using Laplace transform), where The pixel intensity of the local area refers to the absorption coefficient of the voxel (Hounsfield Unit, HU value), which represents the X-ray attenuation coefficient of the voxel, usually in the range of [-1000, 1000] or larger; is a very small positive number (usually close to zero) used for numerical stability (to avoid the denominator If the brightness is too small, a zero division error will occur) and normalization correction (in extreme cases (such as when the brightness of a certain area is almost uniform), It can prevent the calculation results from being too large or too small and improve the robustness of numerical calculations). Essentially a regularization parameter, it is often used in computer vision and image processing algorithms to stabilize contrast calculations. and Represents the maximum pixel intensity and minimum pixel intensity in the local area, which is used to calculate the local contrast of the area to measure the brightness difference between the ommatidium area and the background. This indicates that there are obvious edges or structures in the local area.

[0094] (1-2) Setting adaptive threshold; Through local mean + adaptive adjustment factor Dynamically set thresholds, where It is set dynamically, and the setting method is: for high contrast areas (edges, small eyes dense areas), in order to place transition filtering, A smaller setting will not improve robustness for low-contrast areas (background, noise areas). Larger settings.

[0095] ② Dynamically optimize the processing parameters of each computing unit in GPU parallel computing: Under the CUDA computing framework, a local parameter adjustment strategy based on GPU thread blocks is adopted; (2-1) Different GPU thread blocks independently calculate regional features: Each GPU thread block is responsible for a sub-region and calculates the region, dynamically adjusts the threshold based on the local mean and local variance, determines the sharpness of the ommatidium edge based on local contrast and gradient information, and controls the filtering parameters based on the local ommatidium density.

[0096] (2-2) Adjust processing parameters online. GPU thread blocks exchange information through shared memory, and adjacent areas can be smoothly adjusted. Set adaptive adjustment rules: if the feature differences between adjacent areas are large, perform local parameter interpolation and smoothing; if the small eye density in a certain area is too high or too low, dynamically adjust the processing weight to avoid over-enhancement or loss of small targets. (2-3) Ommatidium structure optimization: adaptive strategy guided by local geometric features; Ommatidia spacing guided optimization: During the detection phase, the parameters are adjusted by calculating the neighborhood distribution density of each ommatidia center; in high-density areas (where ommatidia are densely distributed), the threshold should be stricter to avoid false detections, and in low-density areas (edge ​​areas), a looser threshold is allowed to ensure detection integrity.

[0097] ③ Spatial smoothing optimization; Local spatial smoothing is performed through GPU inter-thread communication. For high-density areas, a smaller calculation window is used to reduce errors; for low-density areas, the calculation window is expanded to enhance detection stability.

[0098] ④Adaptive feature clustering; Parallel computing on the CPU side for optimization based on geometric distribution: Use KNN (k-nearest neighbor) or DBSCAN for clustering. If some ommatidia are detected in isolation, adjust the detection parameters to merge them into the correct cluster. DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is a representative density-based clustering algorithm.

[0099] In summary, the present invention greatly improves the efficiency of insect compound eye ommatidia detection through a parallel computing framework, especially when processing high-resolution CT images, through the joint acceleration of GPU and multi-core CPU, a faster and more efficient target detection method is achieved. Each step can be executed independently in parallel, ensuring the maximum utilization of computing resources during the processing. Ultimately, the present invention can effectively identify and analyze ommatidia structures in complex data sets, providing strong technical support for large-scale biological research and computational biology.

[0100] The compound eye detection method of the above insect compound eye detection method is used to detect the compound eye, and the Figure 2 The result diagram is as follows: Figure 2 (a) shows the projection result of the three-dimensional point cloud data of the insect compound eye in the spherical coordinate system. The corresponding specific steps are: (f1) Cartesian coordinates are converted to spherical coordinates: All voxel data are projected into the spherical coordinate system , making the entire compound eye structure appear in a more intuitive way.

[0101] (f2) Generation of 2D projection slices; Through GPU parallel computing, each voxel The spherical coordinates are mapped to generate cross-slice data, which contain rich geometric information and can help with subsequent ommatidia detection.

[0102] (f3) Grid structures of different colors; The center point of each ommatidium has been preliminarily detected and assigned different colors according to its position. The color coding represents the distribution of ommatidia in different areas, which is helpful for the subsequent analysis of their geometric arrangement.

[0103] Visualization characteristics: The ommatidia are arranged in a regular honeycomb pattern, which is consistent with the geometric characteristics of the compound eye. Due to the spherical structure of the projection, the edge area may have a large deformation, but the point cloud distribution in the central area is relatively uniform.

[0104] Figure 2 (b) shows the precise distribution of the extracted ommatidium center points after ommatidium detection. At this time, unnecessary background noise has been removed and the geometric information of the ommatidium structure has been optimized. The corresponding specific steps are: (g1) Accurate detection of the center of the ommatidium; For the detected ommatidia, the geometric features (such as diameter and area) of each ommatidia are calculated, and the axial vector of the ommatidia is calculated through parallel computing by multi-core CPU, and the noise points are corrected.

[0105] (g2) spatial structure optimization; Redundant points were further removed and spatial calibration was performed to make the arrangement of ommatidia consistent with the geometric characteristics of real compound eyes. Through geometric cluster analysis, ommatidia in different clusters were classified to ensure that their spatial arrangement was consistent with biological characteristics.

[0106] (g3) color coding; The color gradient reflects the position distribution of the ommatidia on the spherical surface (such as light reflection or viewing angle changes). Figure 2 (a) Image, this Figure 2 The data of image (b) is more sparse, retaining only the precise center point information.

[0107] Visualization characteristics: The centers of the ommatidia are closely arranged and conform to the regular distribution on the spherical surface. Figure 2 (b) Image compared to Figure 2 The (a) image is more concise and only retains key detection information.

[0108] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for detecting insect compound eyes, characterized in that: include: Preprocess the insect compound eye CT images; Using dynamic region adaptive thresholding strategy and GPU multi-threaded parallel computing method, the voxels in the CT image stack are subjected to intensity thresholding to obtain the filtered voxel data. Decompose the CT image stack into multiple sub-regions, and use the GPU multi-threaded parallel computing architecture to process the filtered voxel data in blocks. Each GPU thread block processes the voxel data in a sub-region to construct 3D point cloud data. The GPU parallel computing method is used to perform coordinate conversion, converting the voxels of the 3D point cloud data from the Cartesian coordinate system to the spherical coordinate system, and the multi-core CPU parallel computing method is used to perform 2D projection on the voxel data in the spherical coordinate system to generate cross-slice data at different depths; The CPU multi-core parallel computing method is used to analyze the geometric structure of the compound eye of the cross-slice data and obtain the detection results of the insect compound eye structure.

2. The method for detecting insect compound eyes according to claim 1, characterized in that: Before voxel intensity thresholding, the GPU parallel computing method is used to perform adaptive binarization on the CT image, including: Calculate the global statistical characteristics of the CT image stack and preliminarily determine the binarization threshold range; Using CUDA parallel kernel functions, local contrast calculation is performed on each pixel of the CT image, and the threshold is dynamically adjusted based on the local window: GPU parallel thread blocks are used to perform binarization on all pixels in parallel to generate binary data, which are then converted into 0 or 1 using a set threshold to distinguish between the target area and the background area.

3. The method for detecting insect compound eyes according to claim 2, characterized in that: In the dynamic adjustment of threshold based on local window, the threshold The formula is as follows: ； in, are the coordinates of the voxel in the local window, and are the mean and standard deviation in the local window, is the adaptive adjustment factor.

4. The method for detecting insect compound eyes according to claim 1, characterized in that: The filtered voxel data is processed in blocks using the GPU multi-threaded parallel computing architecture. Each GPU thread block processes the voxel data in a sub-area to construct the 3D point cloud data. Specifically: Using a hash index structure, each filtered voxel is stored in a 3D point cloud structure, and outliers are removed through neighborhood statistical filtering to obtain the spatial coordinates and intensity value of each voxel; The spatial coordinates and intensity value of each voxel are stored in the GPU global memory as three-dimensional point cloud data, and each GPU thread processes the three-dimensional point cloud data in parallel.

5. The method for detecting insect compound eyes according to claim 4, characterized in that: The coordinate transformation is performed by using the GPU parallel computing method to transform the voxels of the 3D point cloud data from the Cartesian coordinate system to the spherical coordinate system, specifically including: Each sub-region of the original CT image stack is processed in parallel through GPU multithreading, and the spatial coordinates and intensity values ​​of each voxel are extracted from it; GPU multi-threaded parallel computing is used to transform the voxels in the sub-region from the Cartesian coordinate system to the spherical coordinate system.

6. The method for detecting insect compound eyes according to claim 5, characterized in that: The compound eye geometric structure analysis of cross-slice data was performed using the CPU multi-core parallel computing method to obtain the detection results of the insect compound eye structure, specifically: When the voxel data is expressed in a spherical coordinate system, a spherical space is formed, and the entire spherical space is divided into multiple spherical windows, each of which corresponds to a local area on the spherical space; A 2D morphological detection algorithm based on morphological detection and geometric feature analysis was used to identify the ommatidium center of the cross-slice data in each spherical window; The ommatidia clusters were classified by calculating the Euclidean distance between the center points of each ommatidia. The geometric characteristics of each ommatidium cluster and the angular relationship between adjacent ommatidia are calculated to obtain the spatial arrangement and geometric morphology of the compound eyes.

7. Insect compound eye detection system, characterized in that: It includes image acquisition module, voxel enhancement threshold module, 3D point cloud construction module, cross-slice data construction module and geometric structure analysis module; The image acquisition module is used to preprocess the insect compound eye CT image; The voxel enhancement threshold module is used to perform intensity thresholding on the voxels in the CT image stack using a dynamic region adaptive thresholding strategy and a GPU multi-threaded parallel computing method to obtain filtered voxel data; The 3D point cloud construction module is used to decompose the CT image stack into multiple sub-regions and use the GPU multi-threaded parallel computing architecture to process the filtered voxel data in blocks. Each GPU thread block processes the voxel data in a sub-region to construct 3D point cloud data. The cross-slice data construction module is used to perform coordinate conversion using GPU parallel computing methods, converting the voxels of the three-dimensional point cloud data from the Cartesian coordinate system to the spherical coordinate system, and using multi-core CPU parallel computing methods to perform two-dimensional projection on the voxel data in the spherical coordinate system to generate cross-slice data at different depths; The geometric structure analysis module is used to analyze the geometric structure of compound eyes on cross-slice data using the CPU multi-core parallel computing method to obtain the detection results of the insect compound eye structure.

8. The insect compound eye detection system according to claim 7, characterized in that: The three-dimensional point cloud construction module specifically includes a filtering module and a storage module; The filtering module is used to store each filtered voxel into a 3D point cloud structure using a hash index structure, remove outliers through neighborhood statistical filtering, and obtain the spatial coordinates and intensity value of each voxel; The storage module is used to store the spatial coordinates and intensity value of each voxel as three-dimensional point cloud data in the GPU global memory, and each GPU thread processes the three-dimensional point cloud data in parallel.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the insect compound eye detection method according to any one of claims 1 to 6 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a plurality of classification programs, and the plurality of classification programs are used to be called by a processor and execute the insect compound eye detection method according to any one of claims 1 to 6.

Images