A laser, microwave data receiving and processing system based on real-time perception

Abstract

The application discloses a laser and microwave data receiving and processing system based on real-time sensing, and particularly relates to the technical field of traffic control, comprising a laser data acquisition module, a microwave data acquisition module, a multi-source data fusion processing module and a real-time decision output module. The laser module dynamically matches the target motion state through an adaptive scanning frequency control unit to generate anti-interference laser point cloud data; the microwave module receives reflected signals and generates microwave imaging data; the fusion module extracts laser motion features by using a space-time alignment convolution layer, analyzes microwave scattering characteristics by using a motion compensation attention mechanism, and generates a unified matrix by fusing features through a cross-attention mechanism; the decision module analyzes the matrix by using an FPGA, and outputs control instructions or early warning signals according to a preset threshold. The system realizes the complementary advantages of laser and microwave data, and significantly improves the sensing accuracy and response speed of dynamic targets in complex environments.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of traffic control, more particularly, the present application relates to a laser and microwave data receiving and processing system based on real-time sensing. BACKGROUND

[0002] The dynamic target sensing system in the current traffic environment, such as vehicle and pedestrian detection device, usually uses independently running laser and microwave sensing technology. The laser scanning system generates point cloud data by emitting pulses at a fixed frequency and receiving reflected signals, which is used for high-precision spatial positioning of traffic targets. The microwave radar system generates two-dimensional imaging data by receiving reflected waves, and uses the scattering characteristics of the target to realize all-weather traffic monitoring. After the data of the two systems are independently processed, the target information is generated by simple data splicing or weighted fusion, and finally the central processor outputs traffic control instructions or warning signals based on the preset rules; this scheme has basic sensing ability in static or low-speed target scenarios.

[0003] However, when it is actually used, there are still some shortcomings, such as poor motion adaptability, fixed scanning frequency leading to missed detection or point cloud distortion of high-speed moving vehicles, and inability to dynamically match the acceleration changes of traffic targets such as vehicles; weak data complementarity, simple fusion strategy ignores the depth correlation of laser spatial details and microwave physical properties in traffic target identification, and the sensing accuracy drops sharply in complex traffic environment; insufficient real-time performance, serial processing flow and general processor computing power limit make the system response time delay exceed the standard, which cannot meet the millisecond-level decision-making needs of the traffic system; anti-interference defect, laser data is easily polluted by traffic environment noise, microwave data resolution is insufficient, independent processing amplifies error accumulation, and affects the reliability of traffic control. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the present application provides a laser and microwave data receiving and processing system based on real-time sensing, which solves the problems raised in the background art by the following scheme.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical scheme: a laser and microwave data receiving and processing system based on real-time sensing, comprising:

[0006] A laser data acquisition module receives laser reflection signals of a traffic target area in real time through an adaptive scanning frequency control unit, wherein the scanning frequency is dynamically adjusted according to the traffic target motion speed, and the adjustment rate changes synchronously with the target motion state; pulse code modulation technology is used to generate anti-interference laser point cloud data;

[0007] A microwave data acquisition module receives microwave reflection signals of a traffic target area and generates microwave imaging data;

[0008] The multi-source data fusion processing module is in communication connection with the laser and microwave data acquisition module and comprises a multi-modal feature correlation network.

[0009] The anti-interference laser point cloud data is subjected to real-time feature extraction through a space-time alignment convolution layer to generate a first target feature vector.

[0010] The microwave imaging data is subjected to scattering analysis through a motion compensation attention mechanism to generate a second target feature vector.

[0011] The first and second target feature vectors are input into the multi-modal feature correlation network, feature-level complementary fusion is realized based on a cross-attention mechanism, and a fusion feature matrix is generated.

[0012] The real-time decision output module analyzes the fusion feature matrix and generates a traffic control instruction or a warning signal according to a preset traffic decision rule.

[0013] Preferably, the adaptive scanning frequency control unit comprises a laser reflection signal receiving device that receives laser reflection signals of a traffic target area in real time and an algorithm component based on PID control; the algorithm component dynamically adjusts the scanning frequency according to the motion speed of the traffic target, and the adjustment rate of the scanning frequency is synchronously changed with the motion state of the target characterized by acceleration and speed.

[0014] Preferably, the algorithm component based on PID control comprises a proportional coefficient adjusting unit, an integral coefficient adjusting unit, a differential coefficient adjusting unit, and a target speed reference value setting unit; the proportional coefficient adjusting unit adjusts the proportional coefficient in response to changes in the motion speed of the target, the integral coefficient adjusting unit adjusts the integral coefficient based on the speed deviation, the differential coefficient adjusting unit predicts the motion trend of the target, and when the target has an abrupt deceleration or acceleration mutation, the differential coefficient adjusting unit limits the excessive adjustment of the scanning frequency through a damping effect, and the target speed reference value setting unit sets the target speed reference value.

[0015] Preferably, the anti-interference laser point cloud data comprises a sampling frequency determining unit that determines the sampling frequency according to the Nyquist sampling theorem, a signal quantization unit that quantizes the sampled signals according to the quantization level range and the quantization level to determine the quantization interval, an encoding unit that encodes the quantized signals according to the binary encoding rule, and a three-dimensional matrix generating unit that generates the anti-interference laser point cloud data in the form of a three-dimensional matrix, wherein two dimensions represent the spatial plane coordinate positions of the traffic scene such as a road or an intersection, and the third dimension represents the time dimension, and the matrix elements store the characteristic values of the laser reflection intensity corresponding to the spatial positions and time points.

[0016] Preferably, the microwave data acquisition module comprises: a microwave reflection signal receiving antenna with high gain and narrow beam width; a low-noise amplifier to amplify the microwave reflection signal to enhance the signal strength; an analog-to-digital converter to convert the amplified signal into a digital signal; a radar cross section calculation unit to calculate the radar cross section of the vehicle and other traffic targets based on the radar cross section principle by measuring the received power and known parameters of the transmitted power, antenna gain, wavelength and target distance; a microwave imaging data generation unit to generate microwave imaging data in the form of a two-dimensional matrix through signal processing algorithms, two dimensions representing the imaging plane coordinate position, and matrix elements representing the microwave reflection intensity or related quantitative values at the corresponding position.

[0017] Preferably, the matrix generation of the microwave imaging data comprises: a target displacement calculation unit to calculate the target displacement according to the target motion speed and the microwave signal propagation time; a spatial translation correction unit to perform spatial translation correction on the microwave imaging data according to the target displacement; a two-dimensional image generation unit to generate a two-dimensional image using the back projection algorithm on the spatially translated data; and a discretization unit to discretize the two-dimensional image into a microwave imaging data matrix.

[0018] Preferably, the first target feature vector comprises: a vector generated by real-time feature extraction on the anti-interference laser point cloud data in the form of a three-dimensional matrix through a spatio-temporal alignment convolution layer, wherein the spatio-temporal alignment convolution layer binds a timestamp from the laser scanning system clock synchronization signal to each spatial coordinate point, calculates the motion vector of the target within a unit time through the spatial position difference of two consecutive frames of point cloud data, uses a deformable convolution kernel, and extracts the instantaneous motion feature, motion trend and spatial global feature of the target using a multi-scale convolution kernel, and obtains the features after fusion and pooling operation.

[0019] Preferably, the second target feature vector comprises: a vector generated by scattering analysis on the microwave imaging data through a motion compensation attention mechanism, wherein the motion compensation attention mechanism first calculates the motion distance according to the target motion speed and the microwave signal propagation time, performs spatial translation compensation on the microwave imaging data, then expands the compensated microwave imaging data matrix into a vector in a specific way, calculates an attention weight matrix and obtains the scattering analysis result through matrix operation, and further generates a vector carrying the quantitative information related to the microwave reflection characteristics of the target.

[0020] Preferably, the fusion feature matrix comprises: inputting the first target feature vector and the second target feature vector into a multi-modal feature association network to generate a fusion feature matrix through feature-level complementary fusion based on a cross-attention mechanism, wherein the cross-attention mechanism calculates a cross-attention matrix and generates it through matrix operation with the second target feature vector as the multiplicand.

[0021] Preferably, the real-time decision output module comprises: a high-performance field programmable gate array chip with high-speed parallel processing capability; a component for analyzing the fused feature matrix, a unit for presetting decision rules and parameter thresholds, and a unit for generating traffic control instructions such as traffic signal control, gate control, or early warning signals for triggering warning devices according to conditions met by the feature values; and a high-speed data transmission bus is used to connect the multi-source data fusion processing module to optimize the analysis and decision algorithms.

[0022] Technical effects and advantages of the present application:

[0023] Dynamic motion adaptability improvement: improve the dynamic adaptability of the system to the motion state changes of traffic targets such as vehicles, effectively avoid missing detection of high-speed vehicles, and ensure smooth and safe traffic flow;

[0024] Multi-source data deep complementary fusion: through a deep feature fusion mechanism, the advantages of laser and microwave data in traffic target perception are complementary, and the perception accuracy in complex traffic environments is improved;

[0025] Millisecond-level real-time decision-making capability: using a hardware parallel processing architecture, the system ensures millisecond-level real-time decision-making requirements for traffic control such as intersection signal control and danger warning;

[0026] Overall anti-interference performance is strengthened: the anti-interference performance of the system is strengthened to ensure the reliability of the perception data in dense traffic, environment clutter interference and other variable traffic environments, and to improve the accuracy of the control instructions. BRIEF DESCRIPTION OF DRAWINGS

[0027] Fig. 1 The figure is a schematic diagram of the overall structure of the present application.

[0028] Fig. 2 The figure is a schematic diagram of the multi-modal feature fusion structure of the present application.

[0029] Fig. 3 The figure is a schematic diagram of the real-time decision logic structure of the present application. DETAILED DESCRIPTION

[0030] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0031] Reference Figs. 1-3The laser, microwave data receiving and processing system based on real-time sensing includes a laser data acquisition module, a microwave data acquisition module, a multi-source data fusion processing module, and a real-time decision output module, wherein the connection mode is that the laser data acquisition module and the microwave data acquisition module are connected with the multi-source data fusion processing module, and the multi-source data fusion processing module is connected with the real-time decision output module.

[0032] The laser data acquisition module receives the laser reflection signal of the traffic target area in real time through an adaptive scanning frequency control unit, wherein the scanning frequency is dynamically adjusted according to the traffic target movement speed, and the adjustment rate changes synchronously with the target movement state; and pulse coding modulation technology is used to generate anti-interference laser point cloud data.

[0033] As a preferred feasible example, the adaptive scanning frequency control unit of the laser data acquisition module adopts a high-sensitivity laser reflection signal receiving device to ensure that the laser reflection signal of the traffic target area can be received in real time and accurately;

[0034] The scanning frequency is dynamically adjusted according to the traffic target movement speed, and a PID control-based algorithm is used to achieve this; the target movement speed is , the scanning frequency is , the initial scanning frequency is , the proportional coefficient is , the integral coefficient is , the differential coefficient is , and the scanning frequency adjustment amount is calculated according to the following formula:

[0035] , wherein is the set target speed reference value, is the time;

[0036] The real-time scanning frequency ; in this way, the scanning frequency can be adjusted in a timely and accurate manner according to the change of the traffic target movement speed;

[0037] The target movement state is comprehensively represented by acceleration and speed ;

[0038] The synchronous change of the adjustment rate and the target movement state is realized according to the following formula: , wherein is the microwave signal propagation time, and are coefficients obtained through experimental calibration according to the system hardware performance and actual application requirements; when the target object accelerates, the acceleration According to the formula, the adjustment rate of the scanning frequency also increases accordingly, ensuring that the movement of the target object can be tracked in time;

[0039] When pulse coding modulation is performed on the laser reflection signal, the sampling frequency is first determined According to the Nyquist sampling theorem, the sampling frequency must satisfy , where is the highest frequency component in the laser reflection signal; in practical applications, a is generally selected to reserve a safety margin to ensure the accuracy of sampling;

[0040] After sampling, the signal is quantized, with the quantization level range set to and the quantization level set to , and the quantization interval is . The quantized signal is encoded according to the binary coding rule, thereby generating anti-interference laser point cloud data.

[0041] The anti-interference laser point cloud data is ultimately represented in the form of a three-dimensional matrix , where and represent the coordinate positions in the spatial plane, corresponding to the sampling point positions of the laser scanning in the plane; represents the time dimension, recording the anti-interference laser point cloud data information at different time points; each matrix element stores the laser reflection intensity and other related characteristic values at the corresponding spatial position and time point .

[0042] It should be further noted that the proportional coefficient mainly ensures the rapid response capability of the system to target speed changes. When the target suddenly accelerates and the target speed error increases, the proportional term dominates the rapid adjustment of the scanning frequency, reducing tracking lag, and avoiding overshoot oscillation caused by too rapid response, balancing sensitivity and stability; the integral coefficient is used to eliminate long-term small-amplitude speed deviations, and a smaller value can effectively suppress the cumulative amplification of environmental noise in the integration process, preventing integral saturation and causing system instability; the core role of the differential coefficient is to predict the target motion trend and suppress overshoot. When the target suddenly decelerates or the acceleration suddenly changes, the differential term limits the excessive adjustment of the scanning frequency through damping effect, avoiding system oscillation, and the value is half of , which can effectively smooth the response curve and reduce the sensitivity to high-frequency noise, complementing the hardware filtering measures.

[0043] It needs to be further explained that the target speed reference value is because the value is in the middle of the common motion speed range of the dynamic target aimed at by the laser data acquisition module , which can cover the regular motion state of most low-speed dynamic targets, match the performance constraints of the laser scanning hardware, avoid overloading due to excessive speed or frequent invalid adjustments due to low speed, facilitate parameter calibration, reduce integral item cumulative error, maintain reasonable sampling density at the initial scanning frequency, balance data acquisition efficiency and redundancy, and have wide applicability and practical value.

[0044] The microwave data acquisition module receives the microwave reflection signal of the traffic target area and generates microwave imaging data;

[0045] As a preferred example of feasibility, the microwave data acquisition module is equipped with a microwave reflection signal receiving antenna with high gain and narrow beam width to improve the reception efficiency and accuracy of the microwave reflection signal of the traffic target area;

[0046] The microwave reflection signal is first amplified by a low-noise amplifier to enhance the signal strength and facilitate subsequent processing; the amplified signal is converted into a digital signal by an analog-to-digital converter, and the sampling accuracy is set to bits. When generating microwave imaging data, the radar scattering cross-section principle of microwaves is used; the microwave transmission power is , the antenna gain is , the microwave wavelength is , the target distance is , and the received power is , and the calculation formula of the radar scattering cross-section is: By measuring and known , , and , etc., the radar scattering cross-section of the target object is calculated, and then microwave imaging data is generated according to a series of signal processing algorithms;

[0047] The microwave imaging data is represented in the form of a two-dimensional matrix , where and represent the coordinate positions on the imaging plane; each matrix element represents the microwave reflection intensity or the quantized value related to the microwave scattering characteristics at position on the imaging plane, reflecting the microwave reflection characteristics of the traffic target area at that position.

[0048] It needs to be further explained that the signal processing algorithm generates microwave imaging data:

[0049] Based on the speed of the traffic target and microwave signal propagation time Calculate the target displacement: Spatial translation correction is performed on the microwave imaging data based on the target displacement; a two-dimensional image is generated from the motion-compensated data using a back projection algorithm; and the two-dimensional image is discretized into a microwave imaging data matrix. .

[0050] The multi-source data fusion processing module, which is communicatively connected to the laser and microwave data acquisition modules, includes a multimodal feature association network.

[0051] The anti-interference laser point cloud data is extracted in real time by a spatiotemporally aligned convolutional layer to generate a first target feature vector.

[0052] A second target feature vector is generated by performing scattering analysis on microwave imaging data using a motion-compensated attention mechanism.

[0053] The first and second target feature vectors are input into a multimodal feature association network, and feature-level complementary fusion is achieved based on the cross-attention mechanism to generate a fused feature matrix.

[0054] As a preferred example of feasibility, spatiotemporally aligned convolutional layers are used with a three-dimensional matrix. The anti-interference laser point cloud data, represented in a formal format, is processed as follows:

[0055] The convolutional layer uses multiple convolutional kernels of different sizes. , ,in The number of convolution kernels, These represent the convolution kernels at... Dimensions in three dimensions;

[0056] The formula for convolution is:

[0057] ;

[0058] in These represent the dimensions of the convolution kernel in the three dimensions. This represents the three-dimensional coordinates on the output feature map after convolution. It is the traversal index of the convolution kernel along a certain dimension. For the first After convolution by several kernels, The feature maps generated by multiple convolutional kernels are fused and then pooled to finally generate the first target feature vector. .

[0059] For a two-dimensional matrix Microwave imaging data, represented in formal form, is first subjected to motion compensation. This is done if the velocity of the target object during microwave signal transmission and reception is known. and the propagation time of microwave signals Then the target's movement distance Spatial translation compensation is applied to the microwave imaging data based on the motion distance; then, scattering analysis is performed using an attention mechanism; the attention weight matrix is... Its elements Calculated using the following formula: ,in and It is a microwave imaging data matrix A vector expanded in a specific way. It is an eigenvector transpose, Let the dimension be the vector. The number of vectors, For index variables, For index variables, For the circular index variable, It is a microwave imaging data matrix A vector expanded in a specific way;

[0060] Through matrix operations The results obtained after scattering analysis are then used to generate the second target feature vector. Here is the microwave imaging data matrix. Each element is ,in and Determine the position on the imaging plane, where the element represents the quantized value related to the microwave reflection characteristics at the corresponding position;

[0061] The first target feature vector Second target feature vector Input a multimodal feature association network for feature-level complementary fusion;

[0062] Calculate the cross-attention matrix Its elements The calculation formula is: ,in, Indicates the first Individual and Related feature vectors Indicates the first Individual and Related feature vectors The dimension of the feature vector;

[0063] Through matrix operations Generate a fusion feature matrix This enables feature-level complementary fusion of anti-interference laser point cloud data and microwave imaging data.

[0064] It should be further explained that the spatiotemporally aligned convolutional layer: binds precise timestamp information to each spatial coordinate point in the 3D matrix of the laser point cloud data. The timestamps are derived from the clock synchronization signal of the laser scanning system; by using the spatial position difference between two consecutive frames of point cloud data, it calculates the target's motion vector (including horizontal displacement) per unit time. Vertical displacement and time interval ), forming motion vectors Design deformable convolution kernels, whose sampling points have original coordinates. Dynamically adjusted according to the formula ,in As a time decay factor, when the target suddenly accelerates, Increase to 1 for rapid compensation of position offset; when the target moves at a constant speed, Reduced to 0.8 to avoid overcorrection; multi-scale convolutional kernels work together, including short-term tracking kernel, trajectory prediction kernel and global alignment kernel, to extract the instantaneous motion features, motion trends and spatial global features of the target respectively.

[0065] It should be further explained that in the multimodal feature fusion process, the second target feature vector is chosen as the multiplicand instead of the first target feature vector. The reason for this is the physical adaptability of the technical architecture and the synergistic optimization of the computational logic: microwave features carry the essential physical properties of the target object, have strong anti-interference capabilities and all-weather stability, and provide a reliable foundation of basic attributes for fusion. While laser features have high-precision spatial details, they are susceptible to environmental interference, leading to data fluctuations. Through the cross-attention mechanism, laser features act as a dynamic query vector to actively focus on key areas, while microwave features act as enhanced key-value vectors, receiving weight allocation guided by the laser in matrix operations. This achieves dynamic correction of the resolution of microwave data with laser spatial precision, while preserving the integrity of the physical properties of microwaves. If the operation is reversed, using laser features as the multiplicand, not only will environmental noise contaminate the fusion results, but the high-dimensionality of laser data will also significantly increase the computational load, compromising the system's real-time performance.

[0066] The real-time decision output module parses the fused feature matrix and generates traffic control instructions or early warning signals according to preset traffic decision rules.

[0067] As a preferred feasible example, the real-time decision output module uses a high-performance field-programmable gate array chip, which has high-speed parallel processing capabilities and can meet real-time requirements;

[0068] Chip analysis and fusion feature matrix , preset a series of decision rules; set decision parameter threshold ; if a certain feature value in the fusion feature matrix , maintain the current state and continue to monitor; if a certain feature value in the fusion feature matrix satisfies and (where and are specific position indexes in the feature matrix), a control instruction is generated, such as adjusting the operation of the related equipment; if ( is another specific position index), a warning signal is generated, and an alarm is issued through a sound and light alarm device.

[0069] To ensure that the system response time delay meets the real-time perception requirements, in the hardware design, a high-speed data transmission bus is used to connect the multi-source data fusion processing module and the real-time decision output module, so as to reduce the data transmission delay, and the bus bandwidth needs to meet the real-time transmission requirements of the fusion feature matrix; in terms of algorithm optimization, the analysis and decision algorithms are optimized to reduce unnecessary calculation steps and reduce the time complexity of the algorithm.

[0070] It needs to be further explained that , when the target is moving at a constant speed and there is no environmental interference, the fusion feature matrix takes , which can ensure that more than 95% of the normal state does not trigger a false alarm, and is lower than the minimum adjustment threshold of the laser adaptive scanning frequency, avoiding invalid response caused by slight fluctuations; , the spatiotemporal alignment deviation of the laser point cloud exceeds the correction range of the deformable convolution kernel, or the fluctuation of the microwave imaging scattering characteristics exceeds the error tolerance range of the back projection algorithm, at which time the parameters need to be actively adjusted; , if the invalid points of the laser point cloud are more than 30% and the calculation error of the microwave scattering cross section is more than 50%, the fusion feature matrix loses its decision value.

[0071] Secondly: in the drawings of the disclosed embodiments, only the structures related to the disclosed embodiments are involved, other structures can be referred to the usual design, and in the case of no conflict, the same embodiment and different embodiments of the present application can be combined with each other;

[0072] Finally: the above only describes the preferred embodiments of the present application, and is not used to limit the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A real-time perception-based laser, microwave data receiving and processing system, characterized in that, The application relates to a traffic target detection system based on multi-source data fusion, which comprises the following parts: a laser data acquisition module which receives laser reflection signals of a traffic target area in real time through an adaptive scanning frequency control unit, wherein the scanning frequency is dynamically adjusted according to the motion speed of the traffic target, and the adjustment rate changes synchronously with the motion state of the target; and anti-interference laser point cloud data is generated by using pulse coding modulation technology; a microwave data acquisition module which receives microwave reflection signals of the traffic target area and generates microwave imaging data; a multi-source data fusion processing module which is in communication connection with the laser and microwave data acquisition modules and comprises a multi-modal feature correlation network: real-time feature extraction of the anti-interference laser point cloud data is realized through a space-time alignment convolution layer, and a first target feature vector is generated; scattering analysis of the microwave imaging data is realized through a motion compensation attention mechanism, and a second target feature vector is generated; the first and second target feature vectors are input into the multi-modal feature correlation network, feature-level complementary fusion is realized based on a cross-attention mechanism, and a fusion feature matrix is generated; a real-time decision output module which analyzes the fusion feature matrix and generates a traffic control instruction or a traffic warning signal.

2. The real-time perception-based laser and microwave data receiving and processing system according to claim 1, characterized in that, The adaptive scanning frequency control unit comprises a laser reflection signal receiving device which receives laser reflection signals of a traffic target area in real time and an algorithm component based on PID control; the algorithm component dynamically adjusts the scanning frequency according to the motion speed of the traffic target, and makes the adjustment rate of the scanning frequency change synchronously with the motion state of the target which is characterized by acceleration and speed.

3. The real-time perception-based laser, microwave data receiving and processing system according to claim 2, characterized in that, The algorithm component based on PID control comprises a proportional coefficient adjusting unit, an integral coefficient adjusting unit, a differential coefficient adjusting unit, a target speed reference value setting unit; the proportional coefficient adjusting unit adjusts the proportional coefficient in response to the change of the motion speed of the traffic target, the integral coefficient adjusting unit adjusts the integral coefficient based on the speed deviation, the differential coefficient adjusting unit predicts the motion trend of the target, and when the target suddenly decelerates or the acceleration suddenly changes, the differential coefficient adjusting unit limits the excessive adjustment of the scanning frequency through damping effect, and the target speed reference value setting unit sets the target speed reference value.

4. The real-time perception-based laser and microwave data receiving and processing system according to claim 1, characterized in that, The anti-interference laser point cloud data comprises a sampling frequency determining unit which determines the sampling frequency according to the Nyquist sampling theorem; a signal quantization unit which quantizes the sampled signals, determines the quantization interval according to the quantization level range and the quantization level; an encoding unit which encodes the quantized signals according to the binary encoding rule; and a three-dimensional matrix generating unit which generates anti-interference laser point cloud data in the form of a three-dimensional matrix, wherein two dimensions represent the spatial plane coordinate positions of a road or intersection traffic scene, and the third dimension represents the time dimension, and the matrix elements store the characteristic values of the laser reflection intensities of the corresponding spatial positions and time points.

5. The real-time perception-based laser, microwave data receiving and processing system according to claim 1, characterized in that, The microwave data acquisition module comprises: a microwave reflection signal receiving antenna with high gain and narrow beam width; a low noise amplifier for amplifying the microwave reflection signal to enhance the signal strength; an analog-to-digital converter for converting the amplified signal into a digital signal; a radar cross section calculation unit for calculating the radar cross section of the target based on the radar cross section principle by measuring the received power and the known transmission power, antenna gain, wavelength and target distance parameters; and a microwave imaging data generation unit for generating microwave imaging data in the form of a two-dimensional matrix through a signal processing algorithm, wherein two dimensions represent the imaging plane coordinate position, and the matrix elements represent the microwave reflection intensity or related quantitative values at the corresponding positions.

6. The real-time perception-based laser, microwave data receiving and processing system according to claim 1, characterized in that, The matrix generation of the microwave imaging data comprises: a target displacement calculation unit for calculating the target displacement according to the traffic target motion speed and the microwave signal propagation time; a spatial translation correction unit for performing spatial translation correction on the microwave imaging data according to the target displacement; a two-dimensional image generation unit for generating a two-dimensional image by using an inverse projection algorithm on the spatially translated data; and a discretization unit for discretizing the two-dimensional image into a microwave imaging data matrix.

7. The real-time perception-based laser and microwave data receiving and processing system according to claim 1, characterized in that, The first target feature vector comprises: a vector generated by real-time feature extraction on anti-interference laser point cloud data in the form of a three-dimensional matrix through a space-time alignment convolution layer, wherein the space-time alignment convolution layer binds a timestamp from a laser scanning system clock synchronization signal to each spatial coordinate point, calculates the motion vector of the target within a unit time through the spatial position difference of two consecutive frames of point cloud data, uses a deformable convolution kernel, and extracts the instantaneous motion feature, motion trend and spatial global feature of the target by using a multi-scale convolution kernel, and obtains the features by fusing the feature maps generated by the multiple convolution kernels and performing a pooling operation.

8. The real-time perception-based laser and microwave data receiving and processing system according to claim 1, characterized in that, The second target feature vector comprises: a vector generated by performing scattering analysis on the microwave imaging data through a motion compensation attention mechanism, wherein the motion compensation attention mechanism first calculates the motion distance according to the traffic target motion speed and the microwave signal propagation time, performs spatial translation compensation on the microwave imaging data, then expands the compensated microwave imaging data matrix into a vector in a specific manner, calculates an attention weight matrix and obtains the scattering analysis result through matrix operation, and further generates a vector carrying quantitative information related to the microwave reflection characteristics of the target.

9. The real-time perception-based laser, microwave data receiving and processing system according to claim 1, characterized in that, The fusion feature matrix comprises: inputting the first target feature vector and the second target feature vector into a multi-modal feature association network to generate a fusion feature matrix through feature-level complementary fusion based on a cross-attention mechanism, wherein the cross-attention mechanism calculates a cross-attention matrix and generates the fusion feature matrix through matrix operation with the second target feature vector as the multiplicand.

10. The real-time perception-based laser and microwave data receiving and processing system according to claim 1, characterized in that, The real-time decision output module comprises: a high-performance field programmable gate array chip with high-speed parallel processing capability; a component for analyzing the fusion feature matrix, a unit for presetting decision rules and parameter thresholds, and a component for generating traffic control instructions or traffic warning signals according to the conditions met by the feature values; and a high-speed data transmission bus is used to connect the multi-source data fusion processing module to optimize the analysis and decision algorithms.

Images