A device and method for predicting the trajectory of a moving target based on single-pixel imaging

Abstract

This application discloses a moving target trajectory prediction device and method based on single-pixel imaging, relating to the field of motion trajectory prediction. In the device, an optical collection component collects reflected light from the moving target into a spatial light modulator. The spatial light modulator performs spatial coding modulation on the received light and reflects the modulated light to a first detection optical path component and a second detection optical path component, respectively, achieving single-pixel detection to obtain corresponding first and second light intensity signals. The trajectory prediction component performs differential processing on the first and second light intensity signals, and uses the Fourier single-pixel positioning method to locate the moving target and obtain its trajectory position based on the differentially processed light intensity signals. The trajectory position is then input into a preset trajectory prediction model for trajectory prediction to obtain the future trajectory position of the moving target. This application can use single-pixel positioning for moving targets, achieving accurate trajectory prediction.

Description

Technical Field

[0001] This application relates to the field of motion trajectory prediction, and in particular to a device and method for predicting the trajectory of a moving target based on single-pixel imaging. Background Technology

[0002] Single-pixel imaging is a novel imaging method that reconstructs images by measuring fluctuations in the intensity and phase of a light field and utilizing the correlation between the detected and modulated light fields. It boasts advantages such as anti-interference, ultra-sensitivity, and super-diffraction limit. The unique imaging method of single-pixel imaging means that its research focuses primarily on static targets, with an emphasis on improving the imaging quality and efficiency of stationary targets. However, single-pixel detection, localization, and trajectory prediction schemes for moving objects remain lacking, yet many practical applications rely on the detection of moving objects. The relative motion between the object and the imaging system inevitably affects the detection results. Localizing and predicting the trajectory of moving targets is a crucial component for the engineering and practical application of single-pixel imaging technology, with significant potential applications in lidar, space remote sensing, security monitoring, and autonomous driving. How to perform single-pixel localization and trajectory prediction for moving targets is a key problem that needs to be solved. Summary of the Invention

[0003] The purpose of this application is to provide a moving target trajectory prediction device and method based on single-pixel imaging, which can locate moving targets using single pixels and achieve accurate trajectory prediction.

[0004] To achieve the above objectives, this application provides the following solution:

[0005] In a first aspect, this application provides a moving target trajectory prediction device based on single-pixel imaging, including an optical collection component, a first detection optical path component, a second detection optical path component, a spatial light modulator, and a trajectory prediction component;

[0006] The optical collecting component is used to collect the reflected light from the moving target into the spatial light modulator;

[0007] The spatial light modulator is used to perform spatial coding modulation on the received light and reflect the modulated light to the first detection optical path component and the second detection optical path component, respectively.

[0008] Both the first detection optical path component and the second detection optical path component are used to realize single-pixel detection in order to obtain the corresponding first light intensity signal and second light intensity signal;

[0009] The trajectory prediction component is used to: perform differential processing on the first light intensity signal and the second light intensity signal; use the Fourier single-pixel positioning method to locate the moving target based on the differentially processed light intensity signal to obtain the trajectory position; input the trajectory position into a preset trajectory prediction model for trajectory prediction to obtain the future trajectory position of the moving target; the preset trajectory prediction model is obtained by training a long short-term memory network.

[0010] Secondly, this application provides a method for predicting the trajectory of a moving target based on single-pixel imaging, including:

[0011] A motion target trajectory prediction device based on single-pixel imaging was constructed, and the motion target was set;

[0012] The reflected light from the moving target is collected by an optical collecting component and directed to a spatial light modulator;

[0013] The received light is spatially encoded and modulated by the spatial light modulator, and the modulated light is reflected to the first detection optical path component and the second detection optical path component, respectively.

[0014] Single-pixel detection is achieved through the first detection optical path component and the second detection optical path component to obtain the corresponding first light intensity signal and second light intensity signal;

[0015] The trajectory prediction component performs differential processing on the first light intensity signal and the second light intensity signal, and then uses the Fourier single-pixel positioning method to locate the moving target based on the differentially processed light intensity signal, so as to obtain the position of the motion trajectory.

[0016] The trajectory prediction component inputs the position of the motion trajectory into a preset trajectory prediction model for trajectory prediction, so as to obtain the future trajectory position of the moving target; the preset trajectory prediction model is obtained by training a long short-term memory network.

[0017] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides a moving target trajectory prediction device and method based on single-pixel imaging. By setting up a first detection optical path component, a second detection optical path component, and a trajectory prediction component, single-pixel detection is achieved, and the two detected light intensity signals are differentially processed to obtain the light intensity signal to be used subsequently. Then, the moving target in the scene is located using the Fourier single-pixel positioning method, and a motion preset trajectory prediction model is built using an LSTM network to predict the trajectory, obtaining accurate trajectory prediction coordinates, thus achieving accurate trajectory prediction. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of a moving target trajectory prediction device based on single-pixel imaging, provided in an embodiment of this application.

[0020] Figure 2 This is a schematic diagram of trajectory prediction results provided in an embodiment of this application.

[0021] Figure 3 This is a schematic diagram of trajectory prediction error results provided in an embodiment of this application.

[0022] Reference numerals: 1-Moving target, 2-Optical collecting component, 3-Optical converging component, 4-Single pixel detector, 5-Spatial light modulator, 6-Trajectory prediction component. Detailed Implementation

[0023] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] To make the objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0025] In one exemplary embodiment, such as Figure 1 As shown, a moving target trajectory prediction device based on single-pixel imaging is provided, including an optical collection component 2, a first detection optical path assembly, a second detection optical path assembly, a spatial light modulator 5, and a trajectory prediction component 6. The first and second detection optical path assemblies each include an optical converging component 3 and a single-pixel detector 4 arranged sequentially.

[0026] The optical collecting component 2 is used to collect the reflected light from the moving target 1 onto the spatial light modulator 5, such as onto the micromirror surface of a digital micromirror device (DMD).

[0027] The spatial light modulator 5 is used to spatially code and modulate the received light, and reflect the modulated light to the first detection optical path component and the second detection optical path component respectively; the first detection optical path component and the second detection optical path component are both used to realize single-pixel detection to obtain the corresponding first light intensity signal and second light intensity signal; specifically, the optical converging component 3 converges and collects the light modulated by the spatial light modulator 5 to the single-pixel detector 4 to detect the light intensity signal, and then uses it for subsequent trajectory prediction.

[0028] The trajectory prediction component 6 is used to: perform differential processing on the first light intensity signal and the second light intensity signal; use the Fourier single-pixel positioning method to locate the moving target based on the differentially processed light intensity signal to obtain the position of the moving trajectory; input the position of the moving trajectory into a preset trajectory prediction model for trajectory prediction to obtain the future trajectory position of the moving target; the preset trajectory prediction model is obtained by training a long short-term memory network.

[0029] In one application example, the trajectory prediction component 6 is further configured to: after obtaining the position of the motion trajectory, perform denoising processing on the position of the motion trajectory using a Kalman filter algorithm, and then input it into a preset trajectory prediction model for trajectory prediction. That is, this application uses the trajectory prediction component 6 to perform localization, denoising, and trajectory prediction on the moving target in the scene sequentially based on the differentially processed light intensity signal.

[0030] In one application example, the moving target 1 can be a moving target display.

[0031] In one application example, the spatial light modulator 5 is a digital micromirror array, which performs spatial coding modulation using a Fourier array. During the spatial coding modulation process, the spatial light modulator 5 reflects the modulated light along the ±R direction to the first detection optical path component and the second detection optical path component, respectively; where R represents the angle with the optical axis direction of the spatial light modulator 5, which can be 24°.

[0032] Based on the same inventive concept, this application also provides a method for predicting the trajectory of a moving target based on single-pixel imaging. The solution provided by this method is similar to the solution described in the above-described apparatus; therefore, the specific limitations in one or more method embodiments provided below can be found in the limitations of the apparatus described above, and will not be repeated here. The method for predicting the trajectory of a moving target based on single-pixel imaging includes the following steps 100-600.

[0033] Step 100: Construct a moving target trajectory prediction device based on single-pixel imaging, and set the moving target, such as... Figure 1 As shown.

[0034] Step 200: The reflected light from the moving target is collected to the spatial light modulator by an optical collection component; specifically, the light containing the moving target information can be focused onto the digital micromirror array by an imaging lens.

[0035] Step 300: The received light is spatially encoded and modulated by the spatial light modulator, and the modulated light is reflected to the first detection optical path component and the second detection optical path component respectively; Specifically, during the spatial encoding and modulation process, the digital micromirror array reflects the modulated light signal along ±24°, which is collected by the optical convergence module and collected by the single pixel detector, and the light intensity is recorded.

[0036] Step 400: Single-pixel detection is achieved through the first detection optical path component and the second detection optical path component to obtain the corresponding first light intensity signal and second light intensity signal.

[0037] Step 500: The trajectory prediction component performs differential processing on the first light intensity signal and the second light intensity signal, and then uses the Fourier single-pixel positioning method to locate the moving target based on the differentially processed light intensity signal, so as to obtain the position of the motion trajectory.

[0038] Fourier single-pixel imaging was initially proposed for image acquisition. Its key feature is the use of Fourier basis (i.e., sinusoidal intensity) modes for spatial light modulation, encoding the spatial information of an object into a one-dimensional temporal light signal. Then, by measuring the intensity of the generated light signal using a single-pixel detector, Fourier single-pixel imaging can reconstruct the Fourier spectrum of the target image. Furthermore, the target image can be obtained by performing an inverse Fourier transform on the reconstructed Fourier spectrum. Each Fourier basis map P(x,y) is composed of its spatial frequency pairs (f... x ,f y ) and initial phase The representation can be expressed as the following function:

[0039]

[0040] Where (x, y) are the two-dimensional coordinates in the spatial domain, A is the average intensity of the pattern, and B is the contrast. The light intensity D obtained by modulating the illumination or detection light field using Fourier basis modes is equal to the inner product of the target object image I and the Fourier basis P, and can be characterized by the following function:

[0041]

[0042] The integral in the equation indicates that the Fourier transform is a global-to-point transform. Specifically, each coefficient in the Fourier domain is contributed by all points in the spatial domain. Therefore, any change in the spatial domain will affect all coefficients in the Fourier domain. Utilizing this property, the presence or motion of objects in a scene can be detected by monitoring changes in one or more Fourier coefficients in the Fourier domain. This can be accomplished using Fourier single-pixel imaging, as Fourier single-pixel imaging obtains the Fourier coefficients through spatial light modulation using Fourier-based speckle.

[0043] By modulating the illumination field with Fourier basis speckle, the changes in single-pixel measurements can be observed when an object enters the scene or begins to move within it. This application uses two Fourier coefficients for target localization. Since only two Fourier coefficients need to be measured instead of the complete Fourier spectrum, this application is highly efficient in data acquisition and beneficial for detecting fast-moving targets. During moving target tracking, the linear phase shift characteristic of the Fourier transform is utilized. An image with a displacement of (x0, y0) in the spatial domain has a phase shift of (-2πf) in the Fourier domain. x x0,-2πf y y0):

[0044]

[0045] Among them, (f x ,f y ) represents the spatial frequency coordinates in the Fourier domain. F represents the Fourier spectrum of the moving target image I(x,y). -1 This represents the inverse Fourier transform operation. To obtain the displacement of the target object, this application uses background subtraction, which can be achieved by using... replace To complete. The coefficients are obtained from the background frame, which is a frame obtained before the moving target enters the scene or begins to move.

[0046] The displacement of an object can be determined by the phase term. The result is obtained. Since there are two unknowns (x0 and y0) in the formula, a system of equations consisting of two equations can be established to solve for these two unknowns. This linear system of equations requires two Fourier coefficients. The Fourier coefficients can be obtained by referring to the n-step phase-shifting method. Specifically, this application chose the three-step phase-shifting method because this method has high data acquisition efficiency and strong robustness to noise and illumination fluctuations. To obtain the Fourier coefficients using the three-step phase-shifting method... Three Fourier speckle patterns are required. The three Fourier speckle patterns have the same spatial frequency pairs (f...). x ,f y However, the initial phases are different, i.e. The Fourier coefficients are obtained from three corresponding single-pixel probe values.

[0047]

[0048] In this method, one frame is defined as one localization of a moving target object. Since target localization requires the use of two Fourier coefficients to locate the coordinates in two directions, and each coefficient uses three modes, only six Fourier-based speckles are needed per frame. Target localization can be performed using the corresponding six single-pixel detection values. However, Fourier-based modes are grayscale, while digital micromirror devices (DMDs) can only generate binary modes. Therefore, this application uses binarized Fourier-based speckles for spatial light modulation and employs the Floyd-Steinberg dithering method to binarize the localization speckles. The Fourier localization speckles after Floyd-Steinberg dithering are shown in the figure. The dithered localization speckles consist of 0s and 1s, corresponding to the DMD states "0" and "1" respectively.

[0049] Assuming no moving object is initially present, the theoretical single-pixel measurement should be constant, but in reality, it varies within a very small range. This variation may be caused by noise and ambient light. Once a moving object enters the scene or begins to move, it will cause a significant change in the single-pixel measurement. Once a moving object is detected, its location is determined. Based on this, a Fourier single-pixel localization method is used to locate the moving target based on the differentially processed light intensity signal to obtain the motion trajectory position. This includes: calculating the Fourier coefficients using the following formula based on the differentially processed light intensity signal, and calculating the displacement using the Fourier coefficients:

[0050]

[0051] Among them, f x When it is 0, the Fourier coefficients are obtained. f y When it is 0, the Fourier coefficients are obtained. D represents the light intensity signal after differential processing. The subscript of D indicates the initial phase of the Fourier basis, which are 0, 2π / 3, and 4π / 3, respectively. j represents the imaginary sign. (x0, y0) represents the displacement of the moving target image in the spatial domain. arg{} is used to calculate the argument of the complex number.

[0052] In one application example, after step 500, the motion target trajectory prediction method based on single-pixel imaging further includes: using the Kalman filter algorithm to denoise the motion trajectory position through the trajectory prediction component, and then performing the subsequent step 600.

[0053] Kalman filtering is essentially a recursive method for minimizing variance. It analyzes the state variables of the system to understand their changing patterns. Then, by constructing a mathematical model for the Kalman filter algorithm, it processes the observed data from the current moment with the state update value from the previous moment to obtain the current state update value. The Kalman filter algorithm's processing flow consists of two parts: prediction and estimation. Through multiple iterative iterations, it continuously corrects the error and finally obtains the optimal estimation result. The basic equations of the Kalman filter are shown below:

[0054] 1. Time Update:

[0055] One-step prediction of the calculated state:

[0056] One-step prediction mean square error:

[0057] 2. Measurement Update:

[0058] Filter gain:

[0059] State estimation:

[0060] The estimated mean square error is: P k =(IK k H k )P k / k-1 .

[0061] in: Predict the state in one step; For state estimation; P k / k-1 K is the variance matrix of the prediction error in one step; k P is the filter gain matrix; k To estimate the error variance matrix. Given initial values... When P0, the five basic equations of the Kalman filter are used, according to t k Measurement of time z k Then the state estimate at time k can be calculated recursively. When performing state estimation, the measurement z is required. k Make corrections, K k This represents the filter gain matrix that indicates the corrected gain. And determining K... k The process metric is to minimize the variance of the estimate; therefore, Kalman filtering is a linear minimum variance estimate.

[0062] Considering the noise impact in the centroid determination process of the single-pixel centroid detection method, this application uses the Kalman filter algorithm to denoise the obtained centroid coordinates, thereby improving the accuracy of centroid positioning and thus improving the accuracy of subsequent trajectory prediction.

[0063] Step 600: The trajectory position is input into a preset trajectory prediction model through the trajectory prediction component to predict the future trajectory position of the moving target; the preset trajectory prediction model is obtained by training a long short-term memory network.

[0064] In one application example, the moving target trajectory prediction method based on single-pixel imaging further includes:

[0065] When the reflected light from the moving target within a preset time period is collected to the spatial light modulator by the optical collection component, the trajectory prediction component obtains multiple corresponding motion trajectory positions (e.g., one motion trajectory position for each moment within the preset time period); the trajectory prediction component inputs the multiple motion trajectory positions into a preset trajectory prediction model for trajectory prediction to obtain the future trajectory position of the moving target; the future trajectory position is one future moment trajectory position or multiple future moment trajectory positions.

[0066] LSTM (Long Short-Term Memory) is a deep learning model used to process sequential data. Its key feature is its ability to control the flow of information through its internal logic gates, thus enabling it to better handle long sequences of data. LSTM introduces three types of logic gates: input gates, output gates, and forget gates. These gates can autonomously select which information needs to be passed and which needs to be forgotten, thereby effectively processing long sequences of data.

[0067] Specifically, each time step in an LSTM has a hidden state h. t and a cell state c t LSTM can remove or add "cell state" information through "gate" structures. Through a sigmoid layer, the output is a probability value between 0 and 1, describing how much of each part can pass through. 0 means "disallow the variable to pass," and 1 means "allow the variable to pass." The core of LSTM is the cell state, represented by a horizontal line passing through the cell. This line runs throughout the entire cell, ensuring that information remains constant as it flows within it. LSTM is controlled by three gates: the forget gate, the input gate, and the output gate.

[0068] The forget gate is processed by a sigmoid unit, which checks the hidden state h from the previous time step. t-1 and the input information x at the current moment t Output a value between 0 and 1, indicating whether to keep or discard the cell state c. t-1 Which information is in the forget gate? 0 means discard all, 1 means retain all. The formula for the forget gate is as follows: f t =σ(W f ·[h t-1,x t ]+b f ), where f t Indicates the output of the forget gate, h t-1 Let x represent the hidden state at the previous time step. t This represents the input information at the current moment, and σ(···) represents the sigmoid unit.

[0069] The input gate determines what new information is added to the cell state. First, h t-1 and x t Information is updated via the sigmoid unit. Then, h t-1 and x t A new candidate cell state is created using the tanh unit for subsequent cell information updates. The output gate formula is as follows: i t =σ(W i ·[h t-1 ,x t ]+b i );

[0070]

[0071] Then, the cell information is updated. The update rule is as follows: first, a forgetting gate selects a portion of past cell information for forgetting; then, an input gate selects a portion of candidate cell information to obtain new cell information. The formula is as follows:

[0072] After the cell state is updated, h is needed t-1 and x t To determine the characteristics of the hidden state, the input is passed through the sigmoid unit of the output gate to obtain the decision condition. Then, the cell state is passed through the tanh unit to obtain a value between -1 and 1. This value is multiplied by the decision condition of the output gate to obtain the output hidden state. The output gate formula is as follows: σ t =σ(W o ·[h t-1 ,x t ]+b o );h t =o t ·tanh(C t ).

[0073] In summary, this application performs differential processing on the light intensity signals detected by two single-pixel detectors, and then uses the Fourier single-pixel localization method to locate moving targets in the scene. Considering the noise and binarized speckle error in the single-pixel centroid detection method, a Kalman filter algorithm is used to denoise the obtained centroid coordinates, improving the accuracy of centroid localization and thus improving the accuracy of subsequent trajectory prediction. Compared to the hidden units of RNNs, the hidden units of LSTMs have a more complex internal structure. As information flows along the network, intervention allows LSTMs to selectively add or remove information. LSTM networks can solve long-term dependency problems and have long-term memory capabilities for historical information, making them more suitable for applications in moving target trajectory prediction. Therefore, this application uses an LSTM network to build a moving target trajectory prediction model, uses the denoised historical coordinates obtained through Kalman filtering for trajectory prediction, and outputs more accurate trajectory prediction coordinates by combining the Kalman filtering prediction results. Figure 2 This is a schematic diagram of the trajectory prediction results. The blue curve represents the historical trajectory position (i.e., the positioning result obtained through single-pixel positioning), and the orange coordinate represents the future trajectory position (i.e., the trajectory prediction result). Figure 3 This is a schematic diagram of the trajectory prediction error results, where the horizontal axis represents the number of positioning frames and the vertical axis represents the number of trajectory error pixels.

[0074] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0075] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A moving target trajectory prediction device based on single-pixel imaging, characterized in that, The motion target trajectory prediction device based on single-pixel imaging includes an optical collection component, a first detection optical path assembly, a second detection optical path assembly, a spatial light modulator, and a trajectory prediction component. The optical collecting component is used to collect the reflected light from the moving target into the spatial light modulator; The spatial light modulator is used to perform spatial coding modulation on the received light and reflect the modulated light to the first detection optical path component and the second detection optical path component, respectively. Both the first detection optical path component and the second detection optical path component are used to realize single-pixel detection in order to obtain the corresponding first light intensity signal and second light intensity signal; The trajectory prediction component is used to: perform differential processing on the first light intensity signal and the second light intensity signal; The Fourier single-pixel localization method is used to locate the moving target based on the light intensity signal after differential processing, so as to obtain the position of the moving trajectory; the position of the moving trajectory is input into a preset trajectory prediction model to predict the trajectory, so as to obtain the future trajectory position of the moving target; the preset trajectory prediction model is obtained by training a long short-term memory network. The Fourier single-pixel positioning method is used to locate a moving target based on the differentially processed light intensity signal, thereby obtaining the position of the motion trajectory. This includes: calculating the Fourier coefficients using the following formula based on the differentially processed light intensity signal, and then calculating the displacement using the Fourier coefficients: ； ； ； ； in, Image representing a moving target Fourier spectrum, Represents the spatial frequency coordinates in the Fourier domain. f x When it is 0, the Fourier coefficients are obtained. ; f y When it is 0, the Fourier coefficients are obtained. D represents the light intensity signal after differential processing, and the subscript of D indicates the initial phase of the Fourier basis, which is 0, 2, or 3 respectively. / 3、4 / 3; j Represents the imaginary number symbol, ( x 0, y 0) represents the displacement of the moving target image in the spatial domain. To find the argument of a complex number, The coefficients are obtained from the background frame, which is a frame obtained before the moving target enters the scene or begins to move.

2. The moving target trajectory prediction device based on single-pixel imaging according to claim 1, characterized in that, Both the first detection optical path assembly and the second detection optical path assembly include an optical converging component and a single-pixel detector arranged sequentially. During operation, the optical converging component focuses and collects the light modulated by the spatial light modulator to the single-pixel detector in order to detect the light intensity signal.

3. The moving target trajectory prediction device based on single-pixel imaging according to claim 1, characterized in that, The spatial light modulator is a digital micromirror array, which performs spatial coding modulation through a Fourier array.

4. The moving target trajectory prediction device based on single-pixel imaging according to claim 1, characterized in that, During spatial coding modulation, the spatial light modulator reflects the modulated light along the ±R direction to the first detection optical path component and the second detection optical path component, respectively. Where R represents the angle with respect to the optical axis direction of the spatial light modulator.

5. The moving target trajectory prediction device based on single-pixel imaging according to claim 1, characterized in that, The trajectory prediction component is further configured to: after obtaining the position of the motion trajectory, use a Kalman filter algorithm to denoise the position of the motion trajectory, and then input it into a preset trajectory prediction model.

6. A method for predicting the trajectory of a moving target based on single-pixel imaging, characterized in that, The moving target trajectory prediction method based on single-pixel imaging includes: Construct a motion target trajectory prediction device based on single-pixel imaging as described in any one of claims 1-5, and set the motion target; The reflected light from the moving target is collected by an optical collecting component and sent to a spatial light modulator; The received light is spatially encoded and modulated by the spatial light modulator, and the modulated light is reflected to the first detection optical path component and the second detection optical path component, respectively. Single-pixel detection is achieved through the first detection optical path component and the second detection optical path component to obtain the corresponding first light intensity signal and second light intensity signal; The trajectory prediction component performs differential processing on the first light intensity signal and the second light intensity signal, and then uses the Fourier single-pixel positioning method to locate the moving target based on the differentially processed light intensity signal, so as to obtain the position of the motion trajectory. The trajectory prediction component inputs the position of the motion trajectory into a preset trajectory prediction model for trajectory prediction, so as to obtain the future trajectory position of the moving target; the preset trajectory prediction model is obtained by training a long short-term memory network.

7. The method for predicting the trajectory of a moving target based on single-pixel imaging according to claim 6, characterized in that, After obtaining the position of the motion trajectory, the motion target trajectory prediction method based on single-pixel imaging further includes: using the Kalman filter algorithm to denoise the position of the motion trajectory through the trajectory prediction component.

8. The method for predicting the trajectory of a moving target based on single-pixel imaging according to claim 6, characterized in that, The moving target trajectory prediction method based on single-pixel imaging also includes: When the reflected light from the moving target within a preset time period is collected by the optical collection component and sent to the spatial light modulator, the trajectory prediction component obtains the corresponding multiple motion trajectory positions. The trajectory prediction component inputs multiple motion trajectory positions into a preset trajectory prediction model for trajectory prediction to obtain the future trajectory position of the moving target; the future trajectory position is a trajectory position at one future time or multiple trajectory positions at future time.

Images