Quantum positioning assisted laser radar close-range detection blind area compensation method

Abstract

The invention provides a laser radar close-range detection blind area compensation method assisted by quantum positioning, and aims to solve the problem of a laser radar close-range detection blind area. According to the method, a semiconductor laser with the wavelength of 405 nm is adopted to generate pump light, the pump light is modulated by a wave plate to excite a periodically poled potassium titanyl phosphate crystal, and 810 nm entangled two-photon pairs are generated through spontaneous parametric down-conversion; signal light and reference light are separated by a polarization beam splitter and irradiate a target after being adjusted by a wave plate and a reflector, and a reflected signal is received by a single-photon detector; through coincidence counting measurement of the three quantum positioning devices, the specific position of a close-range to-be-positioned target is calculated and determined based on the positioning principle of arrival time; meanwhile, long-distance target point cloud data is acquired by using the laser radar, and after features are extracted through the convolutional neural network, a final positioning result is output by fusing quantum positioning data, so that the blind area problem of the laser radar in short-distance detection is effectively solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of target positioning, and in particular relates to the application of quantum positioning technology in target detection, and specifically is a quantum positioning-assisted laser radar close-range detection blind area compensation method. Background Art

[0002] As an active optical remote sensing technology, LiDAR has been widely used in many fields such as autonomous driving, robotics, and surveying and mapping due to its high precision, high resolution, and strong anti-interference ability. However, despite its significant advantages in long-distance detection, LiDAR still faces certain limitations in practical applications, especially in close-range detection. LiDAR has obvious blind spot problems, which limits its effective application and performance improvement in some scenarios.

[0003] In order to solve the problem of LiDAR's blind spot in close-range detection, researchers have proposed a variety of solutions, such as using multi-line LiDAR, fusing other sensor data, etc. However, these methods are often accompanied by high costs, complex system design, and reliability challenges, making it difficult to achieve optimization in performance and practical applications.

[0004] In recent years, quantum positioning technology has demonstrated extremely high positioning accuracy and anti-interference capabilities thanks to its characteristics such as quantum entanglement and quantum compression. It performs particularly well in high-precision, low-noise environments. It is particularly suitable for accurate detection of close-range targets and maintains good stability and adaptability in complex environments.

[0005] In view of this, the use of quantum positioning technology to assist LiDAR provides a new approach to solving the problem of LiDAR's blind spots in close-range detection. Compensating for the blind spots of LiDAR in close-range detection with quantum positioning technology can significantly improve the performance and application scope of LiDAR in the overall perception system, especially in the fields of autonomous driving and robotics, to achieve more comprehensive and accurate environmental perception.

[0006] To achieve this goal, the present invention proposes a quantum positioning-assisted laser radar short-range detection blind spot compensation method. This method obtains high-precision quantum position information of close-range targets in real time and effectively integrates it with the point cloud data of long-range targets detected by the laser radar, thereby solving the problem of laser radar short-range detection blind spots and further improving the overall performance, reliability and application scope of the laser radar system. Summary of the invention

[0007] The purpose of the present invention is to provide a quantum positioning-assisted laser radar short-range detection blind spot compensation method. Compared with traditional laser radar detection methods and other sensor fusion technologies, the present invention can effectively solve the blind spot problem of laser radar in short-range detection by combining quantum positioning technology, and improve the overall performance and accuracy of the system.

[0008] In order to achieve the above object, the technical solution adopted by the present invention is: a quantum positioning-assisted laser radar short-range detection blind area compensation method, which specifically includes the following steps:

[0009] Step 1: Install the laser radar firmly on the quantum positioning system and adjust the light beam emission direction of the quantum positioning system so as to realize short-range high-precision positioning using quantum positioning technology (compensate for the short-range detection blind spot of the laser radar) and long-range wide-area detection by the laser radar;

[0010] Step 2: Use a semiconductor laser with a wavelength of 405nm as the light source to generate stable and high-quality pump light. Then, with the help of the synergy of the half-wave plate (HWP) and the quarter-wave plate (QWP), the polarization state of the laser is precisely modulated to convert it into linearly polarized light.

[0011] Step 3: Expose the modulated pump light to the periodically polarized potassium titanyl phosphate (PPKTP) crystal to stimulate the spontaneous parametric down-conversion (SPDC) process inside the crystal, thereby generating an entangled two-photon pair with a wavelength of 810nm. The quantum state of the two-photon entangled state can be described by its wave function |ψ>:

[0012]

[0013] Among them, x i and x s is the output plane of signal and idler photons, the function It is used to characterize the spatial distribution characteristics of the entangled photon pair. The quantum state |0,0> represents the initial state of the entangled two-photon pair. The operator and The generation operators corresponding to signal photons and idle photons respectively;

[0014] Step 4: Separate the entangled light with a wavelength of 810 nm through a polarization beam splitter (PBS) to obtain a signal light and a reference light with the same light intensity I. These two beams, the signal light and the reference light, can be expressed as:

[0015] I∝e- △kd d 2 sinc 2 (△kd / 2) (2)

[0016] Among them, “∝” represents positive correlation, c≈3×10 8 m / s is the speed of light, d is the length of the PPKTP crystal, and △k is the phase matching amount. △k satisfies:

[0017] △k=k p -k i -k s -2π / Λ (3)

[0018] Among them, k p , k i and k s represent the wave vectors of pump light, reference light and signal light respectively, 2π / Λ is the grating wave vector, and Λ is the polarization period;

[0019] Step 5: After the signal beam passes through the reflector and wave plate for beam adjustment, it is irradiated to the close-range target object to be located 1 meter away from the polarization beam splitter. After the signal beam is reflected on the target object, the stray light of other wavelengths is removed by a filter with a central wavelength of 810nm, thereby obtaining a high-purity 810nm entangled signal beam. Subsequently, the beam is coupled into the single-mode optical fiber through a fiber collimator and is finally collected by the single-photon detector 1;

[0020] Step 6: The reference beam (idle beam) is adjusted by a reflector and a wave plate in turn, and then converged by a lens. The converged beam is filtered out by a filter with a central wavelength of 810nm to remove stray light of other wavelengths, thereby obtaining a high-purity 810nm entangled reference beam. The beam then enters the fiber collimator, is transmitted to the single-mode optical fiber and collected by the single-photon detector 2;

[0021] Step 7: Perform coincidence counting analysis on the signal photons and reference photons collected by single photon detector 1 and single photon detector 2. The coincidence counting result can be described by the second-order correlation function of the light field, so as to analyze the correlation characteristics between photons. The second-order correlation function of the light field is expressed as:

[0022]

[0023] Where n(τ) is the coincidence count value, T is the total sampling time, g τTo meet the gate width, R1 and γ1 are the photon counting rate and dark counting rate of single photon detector 1, R2 and γ2 are the photon counting rate and dark counting rate of single photon detector 2, and then the distance information of the close-range target to be located can be obtained according to the coincidence counting measurement results;

[0024] Step 8. On the other two independent quantum positioning systems, follow the operation procedures of steps 1 to 7 to complete the optical path adjustment, pump light generation and modulation, entangled two-photon pair generation, signal light and reference light separation, signal light transmission and detection, and reference light transmission and detection. Based on the positioning principle of arrival time (TOA), combined with the spatial geometric relationship of the three sets of data, the specific position of the close-range target to be positioned in three-dimensional space can be calculated and determined;

[0025] Step 9: Start the laser radar and put it into working state, and according to the speed of light c≈3×10 8 m / s and time delay △t, calculate the distance between the long-distance target object to be located and the laser radar. The formula is:

[0026]

[0027] Where D is the distance between the long-distance target object to be located and the laser radar, and △t is the round-trip time of the laser signal. Then the calculated distance is combined with the angle information scanned by the laser radar to generate the three-dimensional point cloud data of the long-distance target object to be located;

[0028] Step 10: Use Convolutional Neural Network (CNN) to extract the spatial features of the point cloud data generated in step 9;

[0029] Step 11: Flatten the point cloud spatial features extracted in step 10 and convert them into a one-dimensional vector. Perform a full connection (Dense) operation on the position information of the close-range target to be located obtained in step 8 and map it to the same dimension as the feature map output by CNN. Then concatenate these two features together to form a new feature vector, preparing for subsequent feature fusion and processing;

[0030] Step 12: In order to prevent the network from overfitting during training, a Dropout layer with a probability of 0.4 is introduced;

[0031] Step 13: Send the feature vector after Dropout processing to the fully connected layer to further aggregate and extract high-level feature information. The number of nodes in the first fully connected layer is set to 120, and the number of nodes in the second fully connected layer is set to 80;

[0032] Step 14: Input the feature vector generated by the fully connected layer into the Softmax classification function, and output the positioning result after data fusion;

[0033] The step seven includes the following steps:

[0034] Step 7 (1) Considering the existence of scattering particles in the environment, photons may collide with these particles during transmission, causing path changes, resulting in a time error τ of coincidence counting noise If there is no collision, the arrival time difference of the entangled photons in the two channels is τ 0 , and the actual measured arrival time difference of the two-channel entangled photons is τ k , then the time error τ caused by the scattering particles noise It can be defined as:

[0035] τ noise =τ k -τ 0 (6)

[0036] Step 7 (2), during the transmission process of photons from the light source to the target object, the actual transmission distance L can be expressed as:

[0037] L=L real +n d (7)

[0038] Among them, L real Represents the ideal photon transmission distance, that is, the distance without the influence of scattering particles. d Indicates the ranging error caused by the collision of photons and particles in the scattering environment. According to the time delay τ corresponding to the peak value of the light source k The actual transmission distance L can be calculated using the formula:

[0039]

[0040] The step eight includes the following steps:

[0041] Step 8 (1), assuming that the physical coordinates of the three quantum positioning systems are (x1, y1), (x2, y2), (x3, y3), the position of the close-range target to be positioned is represented by (x, y), and the propagation time of the signal light to the close-range target to be positioned is represented by t1, t2, t3, respectively, then:

[0042]

[0043] Expand and simplify formula (9):

[0044]

[0045] Step 8 (2): The two-dimensional coordinates (x, y) of the target to be located can be accurately calculated using the least squares algorithm.

[0046] The step eleven includes the following steps:

[0047] Step 11 (1): Map the point cloud data into a 32×32×32 three-dimensional grid (voxel grid). Each voxel represents a small unit in the three-dimensional space. A binary value of 1 indicates that there is point cloud data in the voxel, and a binary value of 0 indicates that there is no point cloud data. At this point, the three-dimensional point cloud data is converted into a three-dimensional array for CNN processing;

[0048] Step 11 (2): Input the preprocessed voxel data into the constructed CNN, set the number of convolution kernels of convolution module 1 to 32, the convolution kernel size to 3×3×3, the step size to 1, and use the same padding method to keep the size of the input data unchanged;

[0049] Step 11 (3): Perform batch normalization (BN) on the result of step 11 (2), and then apply maximum pooling with a step size of 2 (pooling window is 2×2×2) and rectified linear unit (ReLU) activation operation to further extract shallow features and obtain a shallow feature map of size 16×16×16.

[0050] The present invention has the following advantages: First, it uses quantum positioning technology to provide high-precision short-range target positioning information, solves the blind spot problem of laser radar in short-range detection, and significantly improves the detection ability of laser radar in complex environments. Secondly, the present invention uses quantum positioning technology to assist laser radar, making full use of the advantages of both, and provides more comprehensive target detection information. The advantages of quantum positioning in short-range high-precision positioning and the capabilities of laser radar in wide-area detection complement each other, ensuring that target positioning can be achieved efficiently and accurately under various complex conditions.

[0051] In addition, the present invention uses data fusion technology to effectively integrate the detection information of quantum positioning and lidar, avoiding the limitations of a single sensor and further improving the robustness and reliability of the system. The fused positioning data not only improves the positioning accuracy, but also provides stable information in bad weather, complex terrain and obstructed environments, significantly improving the adaptability and fault tolerance of the system.

[0052] Ultimately, the present invention can achieve accurate target detection and positioning in various complex application scenarios (such as autonomous driving, robot navigation, etc.), making up for the defects of traditional lidar methods in close-range detection blind spots, and has broad practical application value and commercial prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a diagram showing the placement of the entangled photon platform and the laser radar platform of the present invention;

[0054] Figure 2 This is the optical path diagram of the entangled light quantum positioning system of the present invention;

[0055] Figure 3 The TOA-based positioning principle diagram adopted by the present invention;

[0056] Figure 4 This is a neural network architecture diagram for the fusion of quantum positioning and laser radar data in the present invention; Specific implementation plan

[0057] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0058] Step 1: Install the LiDAR firmly on the quantum positioning system and adjust the light beam emission direction of the quantum positioning system to achieve close-range high-precision positioning using quantum positioning technology (compensate for the close-range detection blind spot of the LiDAR) and long-range wide-area detection by the LiDAR. Figure 1 As shown;

[0059] Step 2: Use a semiconductor laser with a wavelength of 405nm as the light source to generate stable and high-quality continuous pump light. Then, with the help of the synergy of the half-wave plate (HWP) and the quarter-wave plate (QWP), the polarization state of the laser is precisely modulated to convert it into linearly polarized light;

[0060] Step 3: Expose the modulated pump light to the periodically polarized potassium titanyl phosphate (PPKTP) crystal to stimulate the spontaneous parametric down-conversion (SPDC) process inside the crystal, thereby generating an entangled two-photon pair with a wavelength of 810nm. The quantum state of the two-photon entangled state can be described by its wave function |ψ>:

[0061]

[0062] Among them, x i and x s is the output plane of signal and idler photons, the function It is used to characterize the spatial distribution characteristics of the entangled photon pair. The quantum state |0,0> represents the initial state of the entangled two-photon pair. The operator and The generation operators corresponding to signal photons and idle photons respectively;

[0063] Step 4: Separate the entangled light with a wavelength of 810 nm through a polarization beam splitter (PBS) to obtain a signal light and a reference light with the same light intensity I. These two beams, the signal light and the reference light, can be expressed as:

[0064] I∝e -△kd d 2 sinc 2 (△kd / 2) (12)

[0065] Among them, “∝” represents positive correlation, c≈3×10 8 m / s is the speed of light, d is the length of the PPKTP crystal, and △k is the phase matching amount. △k satisfies:

[0066] △k=k p -k i -k s -2π / Λ (13)

[0067] Among them, k p , k i and k s represent the wave vectors of pump light, reference light and signal light respectively, 2π / Λ is the grating wave vector, and Λ is the polarization period;

[0068] Step 5: After the signal beam is adjusted by a reflector and a wave plate in turn, it is irradiated to the target object to be located at a close distance of 1 meter from the polarization beam splitter. After the signal beam is reflected on the target object, the stray light of other wavelengths is removed by a filter with a central wavelength of 810nm, thereby obtaining a high-purity 810nm entangled signal beam. Subsequently, the beam is coupled into a single-mode optical fiber through a fiber collimator and is finally collected by the single-photon detector 1.

[0069] Step 6: After the reference beam (idle beam) is adjusted by the reflector and wave plate in turn, it is converged by the lens. The converged beam is filtered out by a filter with a central wavelength of 810nm to remove stray light of other wavelengths, thereby obtaining a high-purity 810nm entangled reference beam. The beam then enters the fiber collimator, is transmitted to the single-mode fiber and collected by the single-photon detector 2. The specific experimental optical path of the quantum positioning system can be referred to Figure 2 ;

[0070] Step 7: Perform coincidence counting analysis on the signal photons and reference photons collected by single photon detector 1 and single photon detector 2. The coincidence counting result can be described by the second-order correlation function of the light field, so as to analyze the correlation characteristics between photons. The second-order correlation function of the light field is expressed as:

[0071]

[0072] Where n(τ) is the coincidence count value, T is the total sampling time, g τ To meet the gate width, R1 and γ1 are the photon counting rate and dark counting rate of single photon detector 1, R2 and γ2 are the photon counting rate and dark counting rate of single photon detector 2, and then the distance information of the close-range target to be located can be obtained according to the coincidence counting measurement results;

[0073] Step 7 (1) Considering the existence of scattering particles in the environment, photons may collide with these particles during transmission, causing path changes, resulting in a time error τ of coincidence counting noise If there is no collision, the arrival time difference of the entangled photons in the two channels is τ 0 , and the actual measured arrival time difference of the two-channel entangled photons is τ k , then the time error τ caused by the scattering particles noise It can be defined as:

[0074] τ noise =τ k -τ 0 (15)

[0075] Step 7 (2), during the transmission process of photons from the light source to the target object, the actual transmission distance L can be expressed as:

[0076] L=L real +n d (16)

[0077] Among them, L real Represents the ideal photon transmission distance, that is, the distance without the influence of scattering particles. d Indicates the ranging error caused by the collision of photons and particles in the scattering environment. According to the time delay τ corresponding to the peak value of the light source k The actual transmission distance L can be calculated using the formula:

[0078]

[0079] Step 8. On the other two independent quantum positioning systems, follow the operation procedures of steps 1 to 7 to complete the optical path adjustment, pump light generation and modulation, entangled two-photon pair generation, signal light and reference light separation, signal light transmission and detection, and reference light transmission and detection. Based on the positioning principle of arrival time (TOA), combined with the spatial geometric relationship of the three sets of data, the specific position of the close-range target to be positioned in three-dimensional space can be calculated and determined;

[0080] Step 8 (1), assuming that the physical coordinates of the three quantum positioning systems are (x1, y1), (x2, y2), (x3, y3), the position of the close-range target to be positioned is represented by (x, y), and the propagation time of the signal light to the close-range target to be positioned is represented by t1, t2, t3, respectively, then:

[0081]

[0082] Expand and simplify formula (18):

[0083]

[0084] Step 8 (2): The two-dimensional coordinates (x, y) of the target to be located can be accurately calculated using the least squares algorithm.

[0085] Step 9: Start the laser radar and put it into working state to receive the light signal reflected from the surface of the target object to be located at a long distance. First, the received light signal is converted into an electrical signal through a highly sensitive photoelectric detector. Then, the collected electrical signal is preliminarily filtered to remove high-frequency noise and background light interference, thereby improving the signal-to-noise ratio. Then, a high-precision time measurement unit is used to accurately measure the time delay from the emission to the reception of the laser pulse. According to the speed of light c≈3×10 8 m / s and time delay △t, calculate the distance between the long-distance target object to be located and the laser radar. The formula is:

[0086]

[0087] Where D is the distance between the long-distance target object to be located and the laser radar, and △t is the round-trip time of the laser signal. Then, the calculated distance is combined with the angle information scanned by the laser radar to generate the three-dimensional point cloud data of the long-distance target object to be located;

[0088] Step 10: Figure 4 As shown in the point cloud data feature extraction module, a convolutional neural network (CNN) is used to extract the spatial features of the point cloud data generated in step nine;

[0089] Step 11: Flatten the point cloud spatial features extracted in step 10 and convert them into a one-dimensional vector. Perform a full connection (Dense) operation on the position information of the close-range target to be located obtained in step 8 and map it to the same dimension as the feature map output by CNN. Then concatenate these two features together to form a new feature vector, preparing for subsequent feature fusion and processing;

[0090] Step 11 (1) Map the point cloud data to a 32×32×32 three-dimensional grid (voxel grid). Each voxel represents a small unit in the three-dimensional space. A binary value of 1 indicates that there is point cloud data in the voxel, and a binary value of 0 indicates that there is no point cloud data. At this point, the three-dimensional point cloud data is converted into a three-dimensional array for CNN processing;

[0091] Step 11 (2), input the preprocessed voxel data into the constructed CNN, set the number of convolution kernels of convolution module 1 to 32, the convolution kernel size to 3×3×3, the step size to 1, and use the same padding method to keep the size of the input data unchanged;

[0092] Step 11 (3), perform batch normalization (BN) on the result of step 11 (2), and then apply maximum pooling with a step size of 2 (pooling window is 2×2×2) and rectified linear unit (REctified Linear Unit, ReLU) activation operation to further extract shallow features and obtain a shallow feature map with a size of 16×16×16;

[0093] Step 12: In order to prevent the network from overfitting during training, a Dropout layer with a probability of 0.4 is introduced;

[0094] Step 13: Send the feature vector after Dropout processing to the fully connected layer to further aggregate and extract high-level feature information. The number of nodes in the first fully connected layer is set to 120, and the number of nodes in the second fully connected layer is set to 80;

[0095] Step 14: Input the feature vector generated by the fully connected layer into the Softmax classification function, and output the positioning result after data fusion.

Claims

1. A quantum positioning-assisted laser radar short-range detection blind area compensation method, characterized in that The following steps are involved: Step 1: Install the laser radar firmly on the quantum positioning system and adjust the light beam emission direction of the quantum positioning system so as to realize short-range high-precision positioning using quantum positioning technology (compensate for the short-range detection blind spot of the laser radar) and long-range wide-area detection by the laser radar; Step 2: Use a semiconductor laser with a wavelength of 405nm as the light source to generate stable and high-quality pump light. Then, with the help of the synergy of the half-wave plate (HWP) and the quarter-wave plate (QWP), the polarization state of the laser is precisely modulated to convert it into linearly polarized light. Step 3: Expose the modulated pump light to the periodically polarized potassium titanyl phosphate (PPKTP) crystal to stimulate the spontaneous parametric down-conversion (SPDC) process inside the crystal, thereby generating an entangled two-photon pair with a wavelength of 810nm. The quantum state of the two-photon entangled state can be described by its wave function |ψ>: Among them, x i and x s is the output plane of signal and idler photons, the function It is used to characterize the spatial distribution characteristics of the entangled photon pair. The quantum state |0,0> represents the initial state of the entangled two-photon pair. The operator and The generation operators corresponding to signal photons and idle photons respectively; Step 4: Separate the entangled light with a wavelength of 810 nm through a polarization beam splitter (PBS) to obtain a signal light and a reference light with the same light intensity I. These two beams, the signal light and the reference light, can be expressed as: I∝e -△kd d 2 from c 2 (△kd / 2) (2) Among them, "∝" represents positive correlation, c≈3×10 8 m / s is the speed of light, d is the length of the PPKTP crystal, and △k is the phase matching amount. △k satisfies: △k=k p -k i -k s -2p / L (3) Among them, k p , k i and k s represent the wave vectors of pump light, reference light and signal light respectively, 2π / Λ is the grating wave vector, and Λ is the polarization period; Step 5: After the signal beam is adjusted by a reflector and a wave plate in turn, it is irradiated to the target object to be located at a close distance of 1 meter from the polarization beam splitter. After the signal beam is reflected on the target object, the stray light of other wavelengths is removed by a filter with a central wavelength of 810nm, thereby obtaining a high-purity 810nm entangled signal beam. Subsequently, the beam is coupled into a single-mode optical fiber through a fiber collimator and is finally collected by the single-photon detector 1. Step 6: The reference beam (idle beam) is adjusted by a reflector and a wave plate in turn, and then converged by a lens. The converged beam is filtered out by a filter with a central wavelength of 810nm to remove stray light of other wavelengths, thereby obtaining a high-purity 810nm entangled reference beam. The beam then enters the fiber collimator, is transmitted to the single-mode optical fiber and collected by the single-photon detector 2; Step 7: Perform coincidence counting analysis on the signal photons and reference photons collected by single photon detector 1 and single photon detector 2. The coincidence counting result can be described by the second-order correlation function of the light field, so as to analyze the correlation characteristics between photons. The second-order correlation function of the light field is expressed as: Where n(τ) is the coincidence count value, T is the total sampling time, g τ To meet the gate width, R1 and γ1 are the photon counting rate and dark counting rate of single photon detector 1, R2 and γ2 are the photon counting rate and dark counting rate of single photon detector 2, and then the distance information of the close-range target to be located can be obtained according to the coincidence counting measurement results; Step 8. On the other two independent quantum positioning systems, follow the operation procedures of steps 1 to 7 to complete the optical path adjustment, pump light generation and modulation, entangled two-photon pair generation, signal light and reference light separation, signal light transmission and detection, and reference light transmission and detection. Based on the positioning principle of arrival time (TOA), combined with the spatial geometric relationship of the three sets of data, the specific position of the close-range target to be positioned in three-dimensional space can be calculated and determined; Step 9: Start the laser radar and put it into working state, and according to the speed of light c≈3×10 8 m / s and time delay △t, calculate the distance between the long-distance target object to be located and the laser radar. The formula is: Where D is the distance between the long-distance target object to be located and the laser radar, and △t is the round-trip time of the laser signal. Then the calculated distance is combined with the angle information scanned by the laser radar to generate the three-dimensional point cloud data of the long-distance target object to be located; Step 10: Use Convolutional Neural Network (CNN) to extract the spatial features of the point cloud data generated in step 9; Step 11: Flatten the point cloud spatial features extracted in step 10 and convert them into a one-dimensional vector. Perform a full connection (Dense) operation on the position information of the close-range target to be located obtained in step 8 and map it to the same dimension as the feature map output by CNN. Then concatenate these two features together to form a new feature vector, preparing for subsequent feature fusion and processing; Step 12: In order to prevent the network from overfitting during training, a Dropout layer with a probability of 0.4 is introduced; Step 13: Send the feature vector after Dropout processing to the fully connected layer to further aggregate and extract high-level feature information. The number of nodes in the first fully connected layer is set to 120, and the number of nodes in the second fully connected layer is set to 80; Step 14: Input the feature vector generated by the fully connected layer into the Softmax classification function, and output the positioning result after data fusion.

2. According to claim 1, a quantum positioning-assisted laser radar short-range detection blind area compensation method is characterized in that The step seven includes the following steps: Step 7 (1) Considering the existence of scattering particles in the environment, photons may collide with these particles during transmission, causing path changes, resulting in a time error τ of coincidence counting noise If there is no collision, the arrival time difference of the entangled photons in the two channels is τ 0 , and the actual measured arrival time difference of the two-channel entangled photons is τ k , then the time error τ caused by the scattering particles noise It can be defined as: t noise =t k -t 0 (6) Step 7 (2), during the transmission of photons from the light source to the target object, the actual transmission distance L can be expressed as: L=L real +n d (7) Among them, L real Represents the ideal photon transmission distance, that is, the distance without the influence of scattering particles. d It indicates the ranging error caused by the collision between photons and particles in the scattering environment. According to the time delay τ corresponding to the peak value of the light source k The actual transmission distance L can be calculated using the formula:

3. The quantum positioning-assisted laser radar short-range detection blind area compensation method according to claim 1 is characterized in that The step eight includes the following steps: Step 8 (1), assuming that the physical coordinates of the three quantum positioning systems are (x1, y1), (x2, y2), (x3, y3), the position of the close-range target to be positioned is represented by (x, y), and the propagation time of the signal light to the close-range target to be positioned is represented by t1, t2, t3, respectively, then: Expand and simplify formula (9): Step 8 (2) uses the least squares algorithm to accurately calculate the two-dimensional coordinates (x, y) of the target to be located at a close distance.

4. According to claim 1, a quantum positioning-assisted laser radar short-range detection blind area compensation method is characterized in that The step eleven includes the following steps: Step 11 (1) Map the point cloud data to a 32×32×32 three-dimensional grid (voxel grid). Each voxel represents a small unit in the three-dimensional space. A binary value of 1 indicates that there is point cloud data in the voxel, and a binary value of 0 indicates that there is no point cloud data. At this point, the three-dimensional point cloud data is converted into a three-dimensional array for CNN processing; Step 11 (2), input the preprocessed voxel data into the constructed CNN, set the number of convolution kernels of convolution module 1 to 32, the convolution kernel size to 3×3×3, the step size to 1, and use the same padding method to keep the size of the input data unchanged; Step 11 (3), perform batch normalization (BN) on the result of step 11 (2), and then apply maximum pooling with a step size of 2 (pooling window is 2×2×2) and rectified linear unit (ReLU) activation operation to further extract shallow features and obtain a shallow feature map of size 16×16×16.

Images