Spectrum-polarization signal processing method for flow field measurement emulating compound eye morphology

Abstract

The application discloses a spectrum-polarization signal processing method for flow field measurement and simulation of compound eye shape, and particularly aims at flow field measurement and detection of flow field tracer particles. Through bionic simulation of spectrum-polarization light signal sensitivity of compound eye shape of dragonflies, bionic visual detection processes of light signal-vision nerve signal conversion and vision nerve signal processing are established, a complete bionic model of target detection of compound eyes of dragonflies is established, and a fusion problem of modeling of spatial characteristics and apparent characteristics of hidden weak tracer particles in a flow field measurement process is solved. Under the condition of not depending on data driving and model training, the method disclosed by the application can sensitively detect water flow tracer particles, and is universally applicable to engineering business monitoring application requirements of large-scale particle imaging velocity measurement.

Description

Technical Field

[0001] This invention relates to a spectral-polarization signal processing method for flow field measurement based on the morphology of compound eyes. Specifically, it refers to a biomimetic visual detection process that mimics the visual morphology of dragonfly compound eyes by using spectral-polarization light signal sensitivity to simulate the morphology of dragonfly compound eyes, converting the light signal into a visual neural signal, and then processing the visual neural signal. This method belongs to the field of flow velocity measurement technology. Background Technology

[0002] my country's river system is complex. Water resource management, hydrological forecasting, and the construction and operation of water projects are crucial to people's livelihoods and national water security. River flow conditions, velocity, and flow rate are fundamental monitoring elements in water conservancy, port shipping, and aquatic ecology, and are also indispensable basic data for efficient water resource utilization, scientific decision-making in flood control and disaster relief, risk prevention and control of water project failures, and water environment protection. Conventionally used instruments and equipment such as rotor current meters, acoustic methods, and radar methods can only measure point velocity, line velocity, or local surface velocity, and their timeliness and suitability for high flood seasons, shallow water areas, and high sediment load flows are severely limited. Large-scale particle imaging velocimetry (LSPIV) is a machine vision-based method for detecting the motion of water flow tracers, which can be applied to online real-time measurement of the entire river surface velocity. However, LSPIV technology suffers from technical problems such as low effective data acquisition rate and poor adaptability to complex and changing field environments, resulting in poor accuracy of observation results (especially when applied to large-scale flow field observations), failing to meet the needs of operational engineering monitoring applications, and is a recognized international challenge.

[0003] Flow field tracer particle detection is one of the key technologies of LSPIV. Due to the complexity of the optical environment at the water surface, flow field tracer particles often exhibit poor survivability, and the problem of low tracer particle detection rate has long plagued the technological development and practical application of flow field measurement. Summary of the Invention

[0004] Purpose of the invention: In order to overcome the key bottlenecks in information acquisition and processing during the detection of flow field tracer particles in the existing LSPIV technology, a spectral-polarization signal processing method with a compound eye-like morphology is proposed. This method can accurately detect flow field tracer particles without relying on data-driven and model-learning conditions, thereby improving the adaptability and accuracy of flow field measurement.

[0005] Technical Solution: A spectral-polarization signal processing method for flow field measurement based on the morphology of compound eyes analyzes the neural activity process of dinosaur compound eyes from the retinal layer to the medulla, lobule layer, mushroom body, and central complex. It biomimetically simulates the spectral-polarization light signal sensitivity of dinosaur compound eyes, the conversion of light signals to visual neural signals, and the processing of visual neural signals, establishing a complete biomimetic model for dinosaur compound eye target detection. Based on spectral-polarization imaging, it acquires spectral intensity images and four-angle orthogonal polarization images, biomimetically extracts spatial position features and apparent polarization features, and obtains the compound eye morphology signal processing results through feature fusion, feature decision, and decision enhancement, completing the entire process of optical imaging, feature extraction, and detection of water flow tracer particles.

[0006] Includes the following steps: S1: Design and construct a spectral-polarization imaging sensor and adaptive imaging technology that mimics the compound eye morphology of dragonflies, forming a spectral-polarization imaging system. The spectral-polarization imaging system is combined with a coated beam splitter and a polarization beam splitter prism to form a four-channel polarization camera, with a front-mounted LCTF electrically adjustable filter and a photometer.

[0007] The scene (flow field) optical information is first spectrally filtered through an LCTF electrically tunable filter. The LCTF is connected to a photometer. When the photometer detects a light intensity > 500 lx, the LCTF is selected in the 360 ​​nm to 520 nm band. When the photometer detects a light intensity ≤ 500 lx, the LCTF is selected in the 760 nm to 1050 nm band. The spectrally filtered light information is then split into two optical information paths by a coated beam splitter: the first optical information is split into 0° polarization information and 45° polarization information by a polarizing beam splitter, forming 0° polarization images and 45° polarization images on the imaging sensor, respectively; the second optical information is split into 90° polarization information and 135° polarization information by a polarizing beam splitter, forming 90° polarization images and 135° polarization images on the imaging sensor, respectively.

[0008] This study biomimetically simulates the adaptability of the retinal layer in the visual field of tadpoles using adaptive imaging technology. The LCTF electrically tunable filter is controlled by a photometer, and this control process simulates the spectral sensitivity of the tadpole compound eye. The adaptive spectral sensitivity bands for imaging are selected as [360 nm–520 nm] and [760 nm–1050 nm], resulting in the following imaging spectrum:

[0009] Where L represents the photometer reading. floor ( ) is an integer operator.

[0010] S2: The biomimetic visual detection process based on visual neural signal processing is a flow field tracing particle detection method that mimics the morphological signal processing of a compound eye, including the following: A1 analyzes and biomimetic models the neural activity processes in the medulla and lobules of the compound eye of fugu. First, calculate the changes in the spectral intensity image at consecutive time points. (1) (2) in, for t The result of SIFT registration of the spectral intensity image at time -1. for t Spectral intensity image at time of time for t Spectral intensity image at time -1 The SIFT transformation matrix is... For convolution; Subsequently, spatial contrast information was obtained by simulating the transverse inhibition mechanism in the medulla oblongata, and the transverse inhibition intensity was modeled as a weighted sum of the inhibitory signals of adjacent neurons: (3) in, for t time Inhibitory signals of neurons at the location, for t The intensity change of the spectral image at time -1. q , s Constrain the size of the neighborhood space. i , j This represents the positional offset of the neighboring neurons. The lateral suppression weights are used to achieve lateral suppression responses in four directions within a two-dimensional space by adjusting these weights. (4) in, , , and These are the suppression signals in the left, right, up, and down directions, respectively. For example, in formula (3), the suppression signals are set as follows: The matrix can be calculated by setting the right half of the elements to 0. , This is for half-wave rectification calculations.

[0011] Simulating the neural activity process of the lobule layer, accumulating t Lateral suppression response in four directions at time: (5) Based on the time delay mechanism of neural response, accumulation t -1 and t The lateral suppression response at time t is used to obtain spatial location features: (6) Simulating the opposition-inhibition sensing mechanism in the medulla and lobule layers to perform opposition calculations on four-angle orthogonal polarization images: (7) (8) in, The contrasting information between 0° and 90° orthogonal polarization images. The contrasting information from 45° and 135° orthogonal polarization images. This is a 0° polarization image. This is a 90° polarized image. This is a 45° polarized image. This is a 135° polarized image. , , and The weights are for the polarization angle responses of 0°, 45°, 90° and 135°, respectively, and their values ​​are proportional to the information entropy of the corresponding polarization image.

[0012] Based on the ON-OFF channel mechanism, the changes in polarization opposition information at consecutive time points are first calculated: (9) (10) The intensity increment of the ON visual pathway for the two sets of polarization-opposing information is calculated as follows: (11) (12) The intensity reduction of the two sets of polarization-opposing information OFF visual pathways is calculated as follows: (13) (14) The inhibitory intensity in each interneuron of the ON visual pathway is modeled as a weighted sum of the inhibitory signals from adjacent neurons: (15) (16) in, and for t ON path Inhibitory signals of neurons at the location, rConstrain the size of the neighborhood space. i , j This represents the positional offset of the neighboring neurons. The weights for the suppression intensity.

[0013] The inhibitory intensity in each interneuron of the OFF visual pathway is modeled as a weighted sum of the inhibitory signals from adjacent neurons: (17) (18) in, and for t OFF Path Inhibitory signals of neurons at the location, r Constrain the size of the neighborhood space. i , j This represents the positional offset of the neighboring neurons. The weights for the suppression intensity.

[0014] The lateral inhibition response in each interneuron of the ON visual pathway is calculated as follows: (19) (20) The lateral inhibition response in each interneuron of the OFF visual pathway is calculated as follows: (twenty one) (twenty two) in, , The influence weights of neighboring neurons on the central neuron; By fusing the lateral inhibition responses of the ON and OFF visual pathways, the apparent polarization characteristics of the elusive target are obtained: (twenty three) in, The modulation weights are used for the ON-OFF path fusion.

[0015] Furthermore, the feature fusion process of the leaflet layer is simulated to determine the reliability of spatial location features and apparent polarization features: (twenty four) in, For spatial location features, For apparent polarization characteristics, n The length of the time series, when Spatial location features provide key decision-making information when There is a conflict between spatial location features and apparent polarization features, so further feature decisions are made within the mushroom body.

[0016] A2 analyzes and biomimetic models the neural activity processes in the compound-eyed mushroom body of fusiform dinosaurs, and the modeling calculations are as follows: (25) in, , These are the convolution matrices for the spatial location feature map and the apparent polarization feature map of the hidden target, respectively. The response output of the mushroom body is represented as a two-dimensional feature map, and the output is the feature decision for the detection of hidden targets.

[0017] A3 analyzes and biomimetic models the neural activity process in the central complex of the compound eye of dinosaurs, and the modeling calculations are as follows: (26) in, Output for mushroom body i Point response, Output for mushroom body j Point response, For neighborhood spatial pixels, Output for mushroom body i The spatial location of a point Output for mushroom body j Spatial location of a point Output for mushroom body i Point and j Difference in strength at points Output the spatial distance between points i and j for the mushroom shape; The neural oscillation process in the central complex can be modeled and calculated as follows: (27) in, and They are respectively Stages and Phase-based central complex response, For a normalized diagonal matrix, As a unit array, , , (28) in NB ( i Definition of the first i Adjacent groups of each group For modulation weights, , .

[0018] Ultimately, when When the central complex stops oscillating, in which The threshold is used for output. , The spatial locations of weak tracer particles are marked to enable tracer particle detection.

[0019] Based on tracer particle detection, A4 uses a target tracking algorithm to track multiple tracer particles, mark their trajectories, and realize flow field measurement.

[0020] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the spectral-polarization signal processing method for flow field measurement with a compound eye-like morphology as described above.

[0021] A computer-readable storage medium storing a computer program that performs the spectral-polarization signal processing method for flow field measurement with a compound eye-like morphology as described above.

[0022] Beneficial effects: Compared with the prior art, the present invention can accurately detect flow field tracer particles without relying on data-driven and model learning conditions, improve the adaptability and accuracy of flow field measurement, and provide key basic technical support for LSPIV. Attached Figure Description

[0023] Figure 1 This is a general block diagram of the modeling of the compound eye morphology of dragonflies according to an embodiment of the present invention; Figure 2 This is the biomimetic spectral-polarization imaging optical path of an embodiment of the present invention; Figure 3 These are the flow field tracing particle detection results of an embodiment of the present invention. Detailed Implementation

[0024] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. After reading the present invention, any modifications of the present invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.

[0025] Dragonfly vision possesses a natural mechanistic advantage, enabling it to keenly detect and identify water surface targets in all weather conditions. This shares a profound mechanistic similarity with flow field tracer particle detection tasks. Therefore, this paper proposes a spectral-polarization signal processing method based on the compound eye morphology of dragonflies. This method biomimeticizes the spectral-polarization light signal sensitivity of dragonfly compound eyes, converts the light signal into a visual neural signal, and processes the visual neural signal, forming a complete set of compound eye-like water surface tracer particle target detection methods. This supports the implementation of LSPIV (Limited-Input Spectroscopic Visual Detection), providing a stable and reliable technical means for water surface flow field measurement.

[0026] A spectral-polarization signal processing method based on the compound eye morphology of tadpoles is proposed for flow field measurement. This method designs and constructs a spectral-polarization imaging sensor and adaptive imaging technology that mimics the compound eye morphology of tadpoles to acquire spectral-polarization images of the flow field. It analyzes and biomimetously models the neural activity processes in the medulla and lobules of tadpoles' compound eyes, extracting the spatial position and apparent polarization features of flow field tracer particles. It also analyzes and biomimetously models the neural activity processes in the mushroom-shaped structure of tadpoles' compound eyes, enabling feature-based decision-making based on spatial position and apparent polarization features. Finally, it analyzes and biomimetously models the neural activity processes in the central complex of tadpoles' compound eyes, achieving decision enhancement. Based on tracer particle detection, a target tracking algorithm is used to track multiple tracer particles, marking their trajectories to achieve flow field measurement, such as... Figure 3 As shown.

[0027] The process of establishing a biomimetic model for target detection of dragonfly compound eyes includes: analyzing the neural activity process of dragonfly compound eyes from the retinal layer to the medulla, lobule layer, mushroom body, and central complex; biomimetic visual detection process that simulates the spectral-polarized light signal sensitivity of dragonfly compound eye morphology, from light signal to visual neural signal conversion, to visual neural signal processing; acquiring spectral intensity images and four-angle orthogonal polarization images based on spectral-polarization imaging, biomimetic extraction of spatial position features and apparent polarization features, and obtaining the compound eye morphology signal processing results through feature fusion, feature decision, and decision enhancement, thus completing the entire process of optical imaging of water flow tracer particles, feature extraction of water flow tracer particles, and recognition and detection of water flow tracer particles.

[0028] A biomimetic model for detecting compound eyes of dragonflies was established. By inputting spectral intensity images and four-angle orthogonal polarization images, the signal processing results of the compound eye morphology were obtained.

[0029] Specifically, the steps include the following: S1: Design and construct a spectral-polarization imaging sensor and adaptive imaging technology that mimics the compound eye morphology of dragonflies, forming a spectral-polarization imaging system. The spectral-polarization imaging optical path is as follows: Figure 2 As shown, the spectral-polarization imaging system combines a coated beam splitter and a polarizing beam splitter to form a four-channel polarization camera, with a front-mounted LCTF electrically adjustable filter and a photometer. The scene (flow field) optical information is first spectrally filtered through an LCTF electrically tunable filter. The LCTF is connected to a photometer. When the photometer detects a light intensity > 500 lx, the LCTF is selected in the 360 ​​nm to 520 nm band. When the photometer detects a light intensity ≤ 500 lx, the LCTF is selected in the 760 nm to 1050 nm band. The spectrally filtered light information is then split into two optical information paths by a coated beam splitter: the first optical information is split into 0° polarization information and 45° polarization information by a polarizing beam splitter, forming 0° polarization images and 45° polarization images on the imaging sensor, respectively; the second optical information is split into 90° polarization information and 135° polarization information by a polarizing beam splitter, forming 90° polarization images and 135° polarization images on the imaging sensor, respectively.

[0030] This study biomimetically simulates the adaptability of the retinal layer in the visual field of tadpoles, employing adaptive imaging technology. An LCTF electrically tunable filter is controlled by a photometer, simulating the spectral sensitivity of the tadpole compound eye during this control process. The adaptive spectral sensitivity bands for imaging are selected as [360 nm–520 nm] and [760 nm–1050 nm], resulting in the following imaging spectrum:

[0031] S2: A flow field tracing particle detection method using compound eye-like signal processing A1 analyzes and biomimetic models the neural activity processes in the medulla and lobules of the compound eye of dinosaurs. First, calculate the changes in the spectral intensity image at consecutive time points. (1) (2) in, for t The result of SIFT registration of the spectral intensity image at time -1. for t Spectral intensity image at time of time for t Spectral intensity image at time -1 The SIFT transformation matrix is... For convolution; Subsequently, spatial contrast information was obtained by simulating the transverse inhibition mechanism in the medulla oblongata, and the transverse inhibition intensity was modeled as a weighted sum of the inhibitory signals of adjacent neurons: (3) in, for t time Inhibitory signals of neurons at the location, for tThe intensity change of the spectral image at time -1. q , s Constrain the size of the neighborhood space. i , j This represents the positional offset of the neighboring neurons. The lateral suppression weights are used to achieve lateral suppression responses in four directions within a two-dimensional space by adjusting these weights. (4) in, , , and These are the suppression signals in the left, right, up, and down directions, respectively. For example, in formula (3), the suppression signals are set as follows: The matrix can be calculated by setting the right half of the elements to 0. , This is for half-wave rectification calculations.

[0032] Simulating the neural activity process of the lobule layer, accumulating t Lateral suppression response in four directions at time: (5) Based on the time delay mechanism of neural response, accumulation t -1 and t The lateral suppression response at time t is used to obtain spatial location features: (6) Simulating the opposition-inhibition sensing mechanism in the medulla and lobule layers to perform opposition calculations on four-angle orthogonal polarization images: (7) (8) in, The contrasting information between 0° and 90° orthogonal polarization images. The contrasting information from 45° and 135° orthogonal polarization images. , , and The weights are for the polarization angle responses of 0°, 45°, 90° and 135°, respectively, and their values ​​are proportional to the information entropy of the corresponding polarization image.

[0033] Based on the ON-OFF channel mechanism, the changes in polarization opposition information at consecutive time points are first calculated: (9) (10) The intensity increment of the ON visual pathway for the two sets of polarization-opposing information is calculated as follows: (11) (12) The intensity reduction of the two sets of polarization-opposing information OFF visual pathways is calculated as follows: (13) (14) The inhibitory intensity in each interneuron of the ON visual pathway is modeled as a weighted sum of the inhibitory signals from adjacent neurons: (15) (16) in, and for t ON path Inhibitory signals of neurons at the location, r Constrain the size of the neighborhood space. i , j This represents the positional offset of the neighboring neurons. The weights for the suppression intensity.

[0034] The inhibitory intensity in each interneuron of the OFF visual pathway is modeled as a weighted sum of the inhibitory signals from adjacent neurons: (17) (18) in, and for t OFF Path Inhibitory signals of neurons at the location, r Constrain the size of the neighborhood space. i , j This represents the positional offset of the neighboring neurons. The weights for the suppression intensity.

[0035] The lateral inhibition response in each interneuron of the ON visual pathway is calculated as follows: (19) (20) The lateral inhibition response in each interneuron of the OFF visual pathway is calculated as follows: (twenty one) (twenty two) in, , The influence weights of neighboring neurons on the central neuron; By fusing the lateral inhibition responses of the ON and OFF visual pathways, the apparent polarization characteristics of the elusive target are obtained: (twenty three) in, The modulation weights are used for the ON-OFF path fusion.

[0036] Furthermore, the feature fusion process of the leaflet layer is simulated to determine the reliability of spatial location features and apparent polarization features: (twenty four) when Spatial location features provide key decision-making information when There is a conflict between spatial location features and apparent polarization features, so further feature decisions are made within the mushroom body.

[0037] A2 analyzes and biomimetic models the neural activity processes in the compound-eyed mushroom body of fusiform dinosaurs, and the modeling calculations are as follows: (25) in, , These are the convolution matrices for the spatial location feature map and the apparent polarization feature map of the hidden target, respectively. The response output of the mushroom body is represented as a two-dimensional feature map, and the output is the feature decision for the detection of hidden targets.

[0038] A3 analyzes and biomimetic models the neural activity process in the central complex of the compound eye of dinosaurs, and the modeling calculations are as follows: (26) in, Output for mushroom body i Point response, Output for mushroom body j Point response, For neighborhood spatial pixels, Output for mushroom body i The spatial location of a point Output for mushroom body j Spatial location of a point Output for mushroom body i Dot and j Difference in strength at points Output the spatial distance between points i and j for the mushroom shape; The neural oscillation process in the central complex can be modeled and calculated as follows: (27) in, and They are respectively Stages and Phase-based central complex response, For a normalized diagonal matrix, As a unit array, , , (28) in NB ( i Definition of the first i Adjacent groups of each group For modulation weights, , .

[0039] Ultimately, when When the central complex stops oscillating, in which The threshold is used for output. , The spatial location of the weak tracer particles is marked to achieve tracer particle detection.

[0040] Based on tracer particle detection, A4 uses a target tracking algorithm to track multiple tracer particles, mark their trajectories, and realize flow field measurement.

[0041] It is obvious to those skilled in the art that the steps of the spectral-polarization signal processing method for flow field measurement with a compound eye-like morphology described in the above embodiments of the present invention can be implemented using general-purpose computing devices. These steps can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, which can then be stored in a storage device for execution by the computing device. In some cases, the steps shown or described can be performed in a different order than those described herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the embodiments of the present invention are not limited to any particular hardware and software combination.

Claims

1. A spectral-polarization signal processing method for flow field measurement based on a compound eye-like morphology, characterized in that, This study analyzes the neural activity processes of dinosaur compound eyes, from the retinal layer to the medulla, lobule layer, mushroom body, and central complex. It biomimetically simulates the spectral-polarized light signal sensitivity and optical signal-visual neural signal conversion of dinosaur compound eye morphology, performing visual neural signal processing to establish a complete biomimetic model for dinosaur compound eye target detection. Using a spectral-polarized imaging system, spectral intensity images and four-angle orthogonal polarization images are acquired. Spatial location features and apparent polarization features are biomimetically extracted. Through feature fusion, feature decision-making, and decision enhancement, the signal processing results of the simulated compound eye morphology are obtained, completing the entire process of optical imaging, feature extraction, and detection of water flow tracer particles. A complete biomimetic model for target detection using dragonfly compound eyes was established, including the following steps: S1: Design and construct a spectral-polarization imaging sensor and adaptive imaging technology that mimics the compound eye morphology of dragonflies, forming a spectral-polarization imaging system. The spectral-polarization imaging system combines a coated beam splitter and a polarization beam splitter prism to form a four-channel polarization camera, with a front-mounted LCTF electrically adjustable filter and a photometer; it biomimetically simulates the adaptability of the retinal layer of dragonflies' vision, using adaptive imaging technology, in which the LCTF electrically adjustable filter is controlled by the photometer; S2: The biomimetic visual detection process based on visual neural signal processing is a flow field tracing particle detection method that mimics the morphological signal processing of a compound eye, including the following: A1 analyzes and biomimetic models the neural activity processes in the medulla and lobular layers of the compound eye of fumaroles; A2 analyzes and biomimetic models the neural activity processes in the compound-eyed mushroom body of fusiform dinosaurs; A3 analyzes and biomimetic models the neural activity process in the central complex of the compound eye of dinosaurs; Based on tracer particle detection, A4 uses a target tracking algorithm to track multiple tracer particles, mark their trajectories, and realize flow field measurement.

2. The spectral-polarization signal processing method for flow field measurement based on the compound eye morphology according to claim 1, characterized in that, Analyzing and biomimetic modeling the neural activity processes in the medulla and lobules of the compound eye of dinosaurs, including: First, calculate the changes in the spectral intensity image at consecutive time points; Subsequently, spatial contrast information was obtained by simulating the transverse inhibition mechanism in the medulla. The transverse inhibition intensity was modeled as a weighted sum of the inhibitory signals of adjacent nerves. By adjusting the transverse inhibition weights, lateral inhibition responses in the left, right, up, and down directions in two-dimensional space were achieved. Simulating the neural activity process of the lobule layer, accumulating t Lateral inhibition response in the four directions of left, right, up, and down at any given moment; Based on the time delay mechanism of neural response, accumulation t -1 and t The spatial location features are obtained from the lateral inhibition response results at each time step. The opposition-inhibition sensing mechanism in the medulla and lobule was simulated by performing opposition calculations on orthogonal polarization images at four polarization angles of 0°, 45°, 90° and 135°. Based on the ON-OFF channel mechanism, the intensity increment of the ON visual path and the intensity decrease of the OFF visual path are calculated based on the change of polarization opposition information at continuous time. The inhibitory intensity in each interneuron of the OFF visual pathway is modeled as a weighted sum of the inhibitory signals of adjacent neurons; The lateral inhibition response in each interneuron of the ON visual pathway and the lateral inhibition response in each interneuron of the OFF visual pathway are calculated; the lateral inhibition responses of the ON and OFF visual pathways are fused to obtain the apparent polarization characteristics of the hidden target. The feature fusion process of the simulated leaflet layer is used to determine the reliability of spatial location features and apparent polarization features. Spatial location features provide decision-making information when There is a conflict between spatial location features and apparent polarization features. In this case, further feature decisions are made in the mushroom body, where sim represents reliability.

3. The spectral-polarization signal processing method for flow field measurement based on the compound eye morphology according to claim 1, characterized in that, The neural activity process in the compound eye mushroom body of fusiform dinosaurs was analyzed and modeled using biomimetic techniques. The modeling calculations are as follows: （25） in, , These are the convolution matrices for the spatial location feature map and the apparent polarization feature map of the hidden target, respectively. Indicates spatial location characteristics, The response output of the mushroom body is represented as a two-dimensional feature map, and the output is the feature decision for the detection of hidden targets.

4. The spectral-polarization signal processing method for flow field measurement based on the compound eye morphology according to claim 1, characterized in that, The neural activity process in the central complex of the compound eye of dinosaurs was analyzed and modeled using biomimetic techniques. The modeling calculations are as follows: （26） in, Output for mushroom body i Point response, Output for mushroom body j Point response, For neighborhood spatial pixels, Output for mushroom body i The spatial location of a point Output for mushroom body j The spatial location of a point Output for mushroom body i Dot and j Difference in strength at points Output the spatial distance between points i and j for the mushroom shape; The neural oscillation process in the central complex is modeled and calculated as follows: （27） in, and They are respectively Stages and Phase-based central complex response, For a normalized diagonal matrix, As a unit array, , , （28） in NB ( i Definition of the first i Adjacent groups of each group For modulation weights, , ; Ultimately, when When the central complex stops oscillating, in which The threshold is used for output. , The spatial locations of weak tracer particles are marked to enable tracer particle detection.

5. The spectral-polarization signal processing method for flow field measurement based on the compound eye morphology according to claim 1, characterized in that, This study biomimetically simulates the adaptability of the retinal layer in the visual field of tadpoles using adaptive imaging technology. The LCTF electrically tunable filter is controlled by a photometer, and this control process simulates the spectral sensitivity of the tadpole compound eye. The adaptive spectral sensitivity bands for imaging are selected as [360 nm–520 nm] and [760 nm–1050 nm], resulting in the following imaging spectrum: Where L represents the photometer reading. floor ( ) is an integer operator; The selected filter band for the LCTF electrically adjustable filter, where L is the scene brightness value measured by the photometer. The scaling factor is used; four orthogonal polarization images with polarization angles of 0°, 45°, 90°, and 135° are obtained by passing the coated beam splitter and two polarizing beam splitters, respectively. , , and .

6. The spectral-polarization signal processing method for flow field measurement based on the compound eye morphology according to claim 1, characterized in that, The spectral-polarization imaging system includes a coated beam splitter and a polarizing beam splitter prism forming a four-channel polarization camera, a front-mounted LCTF electrically adjustable filter and a photometer; the LCTF electrically adjustable filter is controlled by the photometer. The scene or flow field optical information is first spectrally filtered through an LCTF electrically tunable filter. The LCTF electrically tunable filter is connected to a photometer. When the photometer detects a light intensity > 500 lx, the LCTF electrically tunable filter band is selected from 360 nm to 520 nm. When the photometer detects a light intensity ≤ 500 lx, the LCTF electrically tunable filter band is selected from 760 nm to 1050 nm. The spectrally filtered light information is then split into two optical information paths by a coated beam splitter: the first optical information is split into 0° polarization information and 45° polarization information by a polarizing beam splitter, forming 0° polarization images and 45° polarization images on the imaging sensor, respectively; the second optical information is split into 90° polarization information and 135° polarization information by a polarizing beam splitter, forming 90° polarization images and 135° polarization images on the imaging sensor, respectively.

7. The spectral-polarization signal processing method for flow field measurement based on the compound eye morphology according to claim 1, characterized in that, A1 analyzes and biomimetic models the neural activity processes in the medulla and lobules of the compound eye of dinosaurs, including: First, calculate the changes in the spectral intensity image at consecutive time points. （1） （2） in, for t The result of SIFT registration of the spectral intensity image at time -1. for t Spectral intensity image at time of time for t Spectral intensity image at time -1 The SIFT transformation matrix is... For convolution; Subsequently, spatial contrast information was obtained by simulating the transverse inhibition mechanism in the medulla oblongata, and the transverse inhibition intensity was modeled as a weighted sum of the inhibitory signals of adjacent neurons: （3） in, for t time Inhibitory signals of neurons at the location, for t The intensity change of the spectral image at time -1. q , s Constrain the size of the neighborhood space. i , j This represents the positional offset of the neighboring neurons. The lateral suppression weights are used to achieve lateral suppression responses in four directions within a two-dimensional space by adjusting these weights. （4） in, , , and The suppression signals are located in the left, right, up, and down directions, respectively. Formula (3) is used to set the suppression signals in these directions. The right half of the matrix is ​​0. , Calculations are performed for half-wave rectification; Simulating the neural activity process of the lobule layer, accumulating t Lateral suppression response in four directions at time: （5） Based on the time delay mechanism of neural response, accumulation t -1 and t The lateral suppression response at time t is used to obtain spatial location features: （6） Simulating the opposition-inhibition sensing mechanism in the medulla and lobule layers to perform opposition calculations on four-angle orthogonal polarization images: （7） （8） in, The contrasting information between 0° and 90° orthogonal polarization images. The contrasting information from 45° and 135° orthogonal polarization images. This is a 0° polarization image. This is a 90° polarized image. This is a 45° polarized image. This is a 135° polarized image. , , and The weights for the polarization angle responses of 0°, 45°, 90° and 135° are respectively, and their values ​​are proportional to the information entropy of the corresponding polarization image; Based on the ON-OFF channel mechanism, the changes in polarization opposition information at consecutive time points are first calculated: （9） （10） The intensity increment of the ON visual pathway for the two sets of polarization-opposing information is calculated as follows: （11） （12） The intensity reduction of the two sets of polarization-opposing information OFF visual pathways is calculated as follows: （13） （14） The inhibitory intensity in each interneuron of the ON visual pathway is modeled as a weighted sum of the inhibitory signals from adjacent neurons: （15） （16） in, and for t ON path Inhibitory signals of neurons at the location, r Constrain the size of the neighborhood space. i , j This represents the positional offset of the neighboring neurons. The weights for the suppression intensity; The inhibitory intensity in each interneuron of the OFF visual pathway is modeled as a weighted sum of the inhibitory signals from adjacent neurons: （17） （18） in, and for t OFF Path Inhibitory signals of neurons at the location, r Constrain the size of the neighborhood space. i , j This represents the positional offset of the neighboring neurons. The weights for the suppression intensity; The lateral inhibition response in each interneuron of the ON visual pathway is calculated as follows: （19） （20） The lateral inhibition response in each interneuron of the OFF visual pathway is calculated as follows: （21） （22） in, , The influence weights of neighboring neurons on the central neuron; By fusing the lateral inhibition responses of the ON and OFF visual pathways, the apparent polarization characteristics of the elusive target are obtained: （23） in, For ON-OFF path fusion modulation weights; The feature fusion process of the simulated leaflet layer is used to determine the reliability of spatial location features and apparent polarization features: （24） in, For spatial location features, For apparent polarization characteristics, n The length of the time series, when Spatial location features provide key decision-making information when There is a conflict between spatial location features and apparent polarization features, so further feature decisions are made within the mushroom body.

8. A computer device, characterized in that: The computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the spectral-polarization signal processing method for flow field measurement with a compound eye-like morphology as described in any one of claims 1-7.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program that performs a spectral-polarization signal processing method for flow field measurement with a compound eye-like morphology as described in any one of claims 1-7.

Images