Air-ground integrated multi-domain fusion autonomous vehicle transition scene positioning method

Abstract

The invention discloses a space-ground integrated multi-domain fusion autonomous vehicle transition scene positioning method, which comprises the following steps: acquiring an image in a driving process, and judging that a vehicle arrives at a transition scene area through visual image feature matching; a combined positioning model based on robust self-adaption is constructed, the combined positioning model comprises a GNSS / LEO / IMU combined positioning model, and vehicle positioning is carried out on an outdoor area; vehicle positioning is carried out on an indoor area according to a LiDAR / IMU combined positioning model; and performing smooth switching on different outdoor and indoor positioning modes in the transition scene area by adopting an interactive multi-model algorithm. Aiming at the defects of a traditional positioning method during scene switching, an innovative solution is provided, the problem that the positioning precision is poor under the current scene switching condition can be effectively solved, the coherence and stability of the track during driving of the automatic driving automobile are guaranteed, and the market potential is huge.

Description

Technical Field

[0001] The present invention belongs to the technical field of autonomous vehicle positioning, and particularly relates to a positioning method for an autonomous vehicle in a transition scenario with integrated multi-domain space-earth integration. Background Art

[0002] The rapid development of technology has made autonomous vehicles play a crucial role in modern transportation. It can not only enhance driving safety, but also optimize traffic flow and reduce environmental pollution. The positioning technology of autonomous vehicles is one of its core elements and is also the key to building a modern intelligent transportation system, providing support for autonomous driving, path planning, vehicle monitoring, etc. This technology has gradually developed from the initial single technology application to multi-technology integration and is now moving towards the stage of intelligent application. With the continuous upgrade of technology, the positioning technology of autonomous vehicles will play an increasingly crucial role in autonomous driving, traffic control, smart city construction, etc.

[0003] Accurate positioning is the key to autonomous vehicle path planning, and positioning deviation will directly interfere with the accuracy of planning and control algorithms. At present, significant progress has been made in the positioning technology of autonomous vehicles. Currently, the Global Navigation Satellite System (GNSS) is the mainstream vehicle positioning method. Its principle is simple and the cost is low. Most devices use single-point GNSS positioning, and the accuracy is usually at the meter level. In recent years, with the rise of Precise Point Positioning (PPP) and Real-Time Kinematic (RTK) technologies, the positioning accuracy has been improved to the centimeter level. However, the canyon effect limits the scenarios where GNSS can be used alone. To make up for the shortcomings of single technologies, in outdoor scenarios, a combined positioning technology based on GNSS and Inertial Navigation System (INS) has emerged. Through data fusion, the positioning accuracy and reliability have been significantly improved. At the same time, with the development of sensor technology, Light Detection and Ranging (LiDAR), cameras, ultrasonic sensors, etc. have been widely used in vehicle positioning systems, enhancing the vehicle's perception of the surrounding environment and further improving the accuracy and stability of positioning. In addition to the positioning technologies and strategies on the vehicle side itself, excellent performance has been demonstrated in both indoor and outdoor scenarios.

[0004] However, in the indoor-outdoor transition scenario, the research on positioning technology is relatively scarce. GNSS signals are usually interfered with before moving from an outdoor open area to indoor places, tunnels, underground parking lots, under viaducts, etc., which greatly limits the positioning accuracy in these transition scenarios and cannot meet the requirements of autonomous vehicles for high-precision positioning. Summary of the Invention

[0005] To address the deficiencies in the existing technology, this application proposes a positioning method for autonomous driving vehicles in the transitional scenario of the integration of space and terrestrial multi-domains. Combining LEO satellites, which have many advantages such as low operating orbits, high moving speeds, high power, strong penetration, relatively concentrated positioning error distributions, and relatively stable positioning accuracies, it can have more excellent robustness performance without relying on ground infrastructure in complex terrain environments (such as mountains, canyons, forests, etc.). The present invention adopts multi-sensor and multi-domain fusion, and a comprehensive three-dimensional collaborative positioning mode, successfully achieving high-precision positioning of autonomous driving vehicles in transitional scenarios and seamless connection of indoor and outdoor positioning methods to solve the problems in existing autonomous driving vehicle positioning technologies where GNSS signals are vulnerable to environmental restrictions, resulting in a decline in positioning accuracy and interruption of positioning information.

[0006] The technical solution adopted by the present invention is as follows:

[0007] A positioning method for autonomous driving vehicles in the transitional scenario of the integration of space and terrestrial multi-domains, comprising the following steps:

[0008] Step 1: Obtain images during driving, and determine whether the vehicle reaches the transitional scenario area through visual image feature matching.

[0009] Step 2: Construct a combined positioning model based on robust adaptive filtering. The combined positioning model includes a GNSS / LEO / IMU combined positioning model for vehicle positioning in outdoor areas, and a LiDAR / IMU combined positioning model for vehicle positioning in indoor areas.

[0010] Step 3: Adopt an interactive multi-model algorithm to smoothly switch different outdoor and indoor positioning methods in the transitional scenario area.

[0011] Further, the transitional scenario determination method is as follows:

[0012] Step 1-1: Obtain images during the driving of the autonomous driving vehicle.

[0013] Step 1-2: Segment the image into multiple sub-image blocks, calculate the number of main color pixels in each sub-image, extract the main color and its pixel number in the sub-image block to construct the main color feature vector of the entire image.

[0014] Step 1-3: Take pictures of the road scene by setting data sampling points for the transitional scenario, and form a transitional scenario database with the sampling point images according to the main color feature vector descriptors.

[0015] Step 1-4: During the driving of the autonomous driving vehicle, match the collected image information with the transitional scenario image database to determine whether the vehicle reaches the transitional scenario area.

[0016] Further, in step 1-2, the K-means clustering algorithm is used to calculate the number of dominant color pixels in each sub-image block. Specifically, all pixel points in the sub-image are regarded as data points, and the RGB value of each pixel point is used as its feature vector.

[0017] Further, in step 1-4, the one-dimensional dynamic programming matching results of each group of two dominant color feature vectors are calculated iteratively. Let Γ au and Λ bv be the dominant color feature vectors of two images. The dynamic programming matching technology is used to determine the matching distance between the image to be matched and the images in the transition scene database. When the matching distance is greater than the set threshold ζ, it is determined that the vehicle has reached the transition scene area.

[0018] Further, the GNSS / LEO / IMU combined positioning model is constructed based on GNSS, LEO, and IMU. The positioning information vector x G output by the GNSS positioning module or the LEO positioning module, and the positioning information vector x I output by the IMU positioning module are used for positioning information fusion, which is expressed as: And the robust filtering algorithm is used to process the fused positioning information.

[0019] Further, the basis for switching between the GNSS positioning module and the LEO positioning module is as follows: when the number of stably visible satellites is greater than or equal to 4, the positioning information vector x G output by the GNSS positioning module is used; when the number of stably visible satellites is less than 4, the positioning information vector x G output by LEO is used.

[0020] Further, the LiDAR / IMU combined positioning model is constructed based on lidar and IMU. The positioning information vector x L solved based on lidar and the positioning information vector x I solved based on IMU are used for positioning information fusion, which is expressed as: And the robust filtering algorithm is used to process the fused positioning information.

[0021] Further, the interactive multiple model algorithm is used to construct the indoor and outdoor seamless positioning solution as follows:

[0022] Step 3-1: Interact the outdoor positioning obtained from the GNSS / LEO / IMU combined positioning model and the indoor positioning result x i (k-1) and the transition probability matrix Π to obtain the state and variance q i (k-1) of the hybrid system respectively;

[0023] Step 3-2: Estimate and update the system state variables and state covariance using the robust filtering algorithm, and incorporate the obtained results into the interactive session of the final output.

[0024] Step 3-3: In the transition scenario, update the model probability according to the likelihood function of the indoor-outdoor combined positioning model.

[0025] Step 3-4: According to the updated model probability, weight and fuse the robust filtering results of each sub-model in the hybrid model set, output the optimal estimated state, and determine the positioning information of the transition scenario.

[0026] Furthermore, the dynamic programming matching technique determines the matching distance between the two as:

[0027]

[0028] where Γ au represents the main color feature vector of the u-th currently acquired image, Λ bv represents the main color feature vector of the v-th image in the transition scenario image database, d(Γ au , Λ bv ) represents the element distance between Γ au and Λ bv , Γ[a, u] represents the value of Γ au , Λ[b, v] represents the value of Λ bv , u = 1, 2, 3..., z, v = 1, 2, 3..., z, and z is the number of images.

[0029] Advantages of the present invention:

[0030] (1) Aiming at the problems of decreased positioning accuracy and interrupted positioning information faced by traditional positioning methods in the transition scenario, the present invention proposes a space-earth integrated multi-domain fusion high-level autonomous driving vehicle transition scenario positioning method applicable to the transition scenario, which is different from the current mainstream methods such as multi-sensor fusion and vehicle-road collaborative positioning. It effectively solves the problem of positioning performance attenuation in the transition scenario and improves the positioning accuracy in this area.

[0031] (2) The present invention innovatively smooths the transition from the outdoor GNSS / LEO / IMU combined positioning method to the indoor lidar / IMU combined positioning method in the transition scenario area, realizing the positioning of autonomous driving vehicles in the full scenario.

[0032] (3) The present invention adopts a diversified comprehensive positioning method of multi-sensor fusion and space-earth integrated multi-domain fusion to achieve higher-precision positioning and seamless switching in the transition scenario, enabling large-scale commercial applications. Compared with traditional positioning technologies, it effectively improves the continuity and smoothness of the trajectory during the driving of autonomous driving vehicles and has broad market prospects. Description of the Drawings

[0033] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained according to the provided accompanying drawings.

[0034] Figure 1 It is the technical flow chart of the present invention.

[0035] Figure 2 It is the schematic diagram of the indoor and outdoor combined positioning interactive multi-model. Specific embodiments

[0036] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the following further details the present invention in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0037] In this embodiment, the positioning experiment of the present invention is completed when entering the underground garage entrance. A positioning method for the transition scene of a sky-earth integrated multi-domain fusion high-level autonomous driving vehicle provided by the present invention, the specific technical flow chart is as Figure 1 shown, and the specific implementation steps are as follows:

[0038] Step 1: Determine that the vehicle reaches the transition scene area through visual image feature matching.

[0039] Step 1-1: Use the on-vehicle image acquisition device to obtain images during the driving of the autonomous driving vehicle;

[0040] Step 1-2: Divide the currently acquired frame image into m×n sub-image blocks. In this embodiment, m = n = 5 is taken. Use the K-means clustering algorithm to calculate the number of main color pixels in each sub-image block. Specifically, all pixel points in the sub-image are regarded as data points, and the RGB value of each pixel point is used as its feature vector. Select an appropriate number of clustering centers K, take K = 15, and run the K-means clustering algorithm until convergence. Finally, each clustering center represents a main color, and the number of pixels belonging to this clustering center is the number of pixels of this main color. In this way, the main color and its pixel number in the sub-image block are accurately extracted. Then, the occurrence frequencies of each main color in each sub-block are integrated to construct the main color feature vector of the entire image. Thus, the color content of the current frame image can be represented by a main color feature vector of size m×n×K = 375.

[0041] Step 1-3: In the present invention, data sampling points are set for the transition scene to take pictures of the road scene, and the sampled point images are formed into a transition scene database according to the main color feature vector descriptors.

[0042] Step 1-4: The self-driving vehicle matches the collected image information with the transition scene image database during driving. Specifically, the one-dimensional dynamic programming (DP) matching results of each group of two main color feature vectors are calculated in a loop. Let Γ au and Λ bv be the main color feature vectors of two images. The matching distance between them is determined by means of the dynamic programming (DP) matching technique as follows:

[0043]

[0044] where Γ au represents the main color feature vector of the u-th currently collected image, Λ bv represents the main color feature vector of the v-th image in the transition scene image database, d(Γ au , Λ bv ) represents the element distance between Γ au and Λ bv , Γ[a, u] represents the value of Γ au , Λ[b, v] represents the value of Λ bv , u = 1, 2, 3..., z, v = 1, 2, 3..., z, and z is the number of images.

[0045] Equation (1) can cover the matrix of all potential element distances between Γ a and Λ b . In the matrix, d(Γ au , Λ bv ) is simplified to d(u, v), and the specific description is as follows:

[0046]

[0047] To further calculate the nearest feature vector matching distance between Γ a and Λ b , let the weights of d(u - 1, v), d(u, v - 1), and d(u - 1, v - 1) to d(u, v) be ω1, ω2, and ω3 respectively. Among them, ω3 = τ·d(u, v), where τ is an empirical coefficient, and the specific setting is as follows:

[0048]

[0049] In the interval of u - v ∈ [-R, R], each of the above d(u, v) is connected according to the corresponding weight to define a directed graph D. Its starting point is d(1, 1), and the ending point is d(n, n). The shortest path length d between d(1, 1) and d(n, n) is solved by the Dijkstra algorithm. min , then the matching distance of the feature vectors between the current image to be matched and the images in the transition scene database is:

[0050]

[0051] Then, the image to be matched is compared with the images in the transition scene database one by one. When the matching distance is greater than the set threshold ζ, it is further determined that the vehicle has reached the transition scene area.

[0052] Step 2: Construct a combined positioning model based on robust adaptation.

[0053] Step 2 - 1: Respectively construct an outdoor scene GNSS / LEO / IMU combined positioning model based on GNSS, LEO, and IMU to obtain accurate vehicle positioning information in the area where satellite signals can be effectively utilized outdoors.

[0054] The working principles of GNSS, LEO, and IMU are introduced respectively as follows:

[0055] (1) GNSS positioning module: The position information received by the GNSS signal receiver on the vehicle is: [x, y, z] T , and the coordinate information of the i-th visible satellite that can be observed is [x (i) , y (i ), z (i) T , then the actual distance between the receiver installed on the vehicle and the satellite is:

[0056]

[0057] Suppose there are n visible satellites that can provide positioning information, and n pseudo-range observation equations of the visible satellites can be established:

[0058]

[0059] Among them, ρ (i) is the pseudo-range of the receiver relative to the i-th satellite, c represents the signal propagation speed, and δt u is the receiver clock error.

[0060] ​As can be seen from Equation (4), there are four unknowns in the system of equations. Therefore, when the number of stable visible satellites is greater than or equal to 4, the system of equations can be solved to obtain GNSS positioning information. At the same time, the RTK (Real-Time Kinematic) positioning technology is used to correct the GNSS positioning information and output a positioning information vector x with higher accuracy. G 。

[0061] (2) LEO positioning module: If the number of stable visible satellites is less than 4, it is considered that the GNSS satellite signal quality does not meet the requirements. Then, LEO needs to be used for outdoor position positioning and output the positioning information vector x. G 。

[0062] (3) IMU positioning module: The inertial measurement unit (IMU) is used to output the acceleration a and angular velocity ω information of the vehicle. The discrete-time data is integrated to obtain the velocity v and position information l of the vehicle in the world coordinate system. The calculation formulas are as follows:

[0063]

[0064] where v0 and l0 are the initial velocity and position information of the vehicle, a(t) is the acceleration a of the vehicle at time t, and v(t) is the velocity v of the vehicle at time t.

[0065] Furthermore, by integrating the angular velocity, the heading angle of the vehicle is solved. The calculation formula is as follows:

[0066]

[0067] where is the initial heading angle of the vehicle, and ω(t) is the angular velocity of the vehicle at time t.

[0068] Finally, by combining the position information l and heading angle information of the vehicle, the positioning information vector x of the vehicle is calculated. I 。

[0069] (4) Based on the above GNSS, LEO, and IMU positioning modules, an outdoor scene GNSS / LEO / IMU integrated positioning model is constructed. The process of fusing the positioning information of GNSS, LEO, and IMU is as follows:

[0070] Based on the positioning information vector x output by the above GNSS positioning module or LEO positioning module G and the positioning information vector x output by the IMU positioning module I , the positioning information is fused as shown in the following formula:

[0071]

[0072] Combined with the non - linear characteristics of the GNSS / LEO / IMU integrated positioning model, a robust filtering algorithm is used as the filtering algorithm for integrated positioning.

[0073] The integrated positioning system can be described as:

[0074]

[0075] where, x k is the state vector of the integrated positioning system parameters, Φ k-1 is the state transition matrix, Τ is the noise distribution matrix, w k-1 is the process noise vector, z k is the observation vector of the integrated positioning system, H k is the system observation matrix, v k is the observation noise vector.

[0076] The robust estimation first obtains the state error and observation error equations of the integrated positioning system, as shown in the following formula:

[0077]

[0078] where, z k is the observation vector of the integrated positioning system, is the state prediction vector at time k, is the state estimation value at time k.

[0079] In the robust estimation, a robust M - estimation is performed on the observation vector, and the following conditional extremum is constructed:

[0080]

[0081] where, is the equivalent weight matrix of the observation vector z k and is the equivalent weight matrix of the prediction vector .

[0082] The present invention uses the IGGⅢ weight function for solution, and its expression is as follows:

[0083]

[0084] Thus, the equivalent weight function can be obtained as:

[0085]

[0086] In the formula, the value ranges of c0 and c1 are generally [1.5, 2.0] and [3.0, 8.5] respectively. is its standardized residual value, σi is the mean square error of v i , and b i is the weight of the i-th observation value.

[0087] Taking the extreme value of Equation (11) with respect to x k gives the robust solution vector of the system state parameters:

[0088]

[0089] Then the corresponding posterior covariance matrix can be described as:

[0090]

[0091] The filtering gain is:

[0092]

[0093] Furthermore, the recursive solution of

[0094]

[0095] The posterior covariance matrix corresponding to the recursive solution can be approximately expressed as:

[0096]

[0097] In the formula, represents the robust equivalent covariance of represents that the robust equivalent covariance of z k , and

[0098] Thus, by optimizing the appropriate control parameters c0 and c1, the observed data can be classified. When outliers appear in the observed data, the equivalent weight function will assign corresponding weights to them according to the size of the standard residual, thereby reducing the interference of outlier observations on state estimation, exerting the robust function, and further enhancing the stability of the filtering estimation.

[0099] Through the above steps, the fusion of GNSS, LEO, and IMU positioning information can be achieved, and more accurate vehicle positioning information can be obtained in the area where satellite signals can be effectively utilized outdoors.

[0100] Step 2-2: Based on LiDAR (Light Detection and Ranging) and IMU, construct a LiDAR / IMU integrated positioning model in an indoor scenario to achieve high-precision positioning of an autonomous vehicle in an area without satellite signals (transition scenario).

[0101] Specifically:

[0102] (1) LiDAR:

[0103] First, obtain the point cloud data scanned by the lidar, and process the point cloud data to obtain the currently observed feature M. According to the pose of the vehicle and the pose of feature M in the vehicle coordinate system, calculate the pose of feature M in the world coordinate system. At this time, add feature M to the map (update the map); when the vehicle pose changes and feature M is observed again, the vehicle pose x can be solved according to the pose of feature M in the world coordinate system and the pose of feature M in the vehicle coordinate system L 。

[0104] Concatenate the positioning information vector x L solved based on LiDAR I and the positioning information vector x

[0105]

[0106] Combined with the non-linear characteristics of LiDAR / IMU integrated positioning, the robust filtering algorithm is also used as the filtering algorithm for integrated positioning. The specific steps refer to Step 2-2

[0107] Step 3: Smoothly switch different outdoor and indoor positioning methods in the transition scene area, specifically as follows

[0108] Adopt the interactive multiple model algorithm (IMM) to construct a seamless indoor and outdoor positioning solution, achieve automatic smooth switching between indoor and outdoor positioning modes, and continuously output high-precision positioning results, as Figure 2 shown, specifically including: input information interaction, model filtering, model probability update, and estimation fusion

[0109] Step 3-1: Input information interaction. Interact the outdoor positioning obtained from the GNSS / LEO / IMU integrated positioning model and the indoor positioning result x i (k - 1) obtained from the LiDAR / IMU integrated positioning model with the transition probability matrix Π, and respectively obtain the state and variance q i (k - 1) of the hybrid system, which is specifically described as

[0110]

[0111] Among them, and q j (k - 1) are the state and variance of any model k - 1 in the hybrid system, μ j|i (k - 1) is the hybrid probability of switching from any other model j to model i in the hybrid model at k - 1 moment, Π = [π ij M×M ​, where M is the number of sub-models, and π ij is the probability that model i jumps to model j, and μ j (k - 1) is the probability of model j at time k - 1, where i, j = 1, 2.

[0112] Step 3 - 2, Model filtering process. The respective combined positioning systems outdoors and indoors use the robust filtering algorithm to complete the estimation and update of the system state variables and state covariance, and the results obtained by each filter are incorporated into the interaction link of the final output.

[0113] The specific process is as follows:

[0114] For outdoor positioning, take the obtained through interaction as the input of the robust filtering algorithm of the GNSS / LEO / IMU combined positioning model, and perform the robust filtering calculation again as shown in Equation 8 - 18, and output

[0115] For indoor positioning, take the obtained through interaction as the input of the robust filtering algorithm of the LiDAR / IMU combined positioning model, and perform the robust filtering prediction calculation again as shown in Equation 8 - 18, and output

[0116] Step 3 - 3, Model probability update. In the transition scenario, the model probability update process is as follows according to the likelihood function of the indoor - outdoor combined positioning model:

[0117]

[0118]

[0119] where μ i (k) is the probability of any model i in the hybrid model at time k, Ξ i (k) is the likelihood function of model i at time k, v i is the residual of model i at time k, and S i (k) is the residual covariance matrix of model i at time k.

[0120] Step 3 - 4, Estimation fusion. According to the updated model probability μ i (k), weight and fuse the robust filtering results of each sub - model in the hybrid model set to output the optimal estimated state and determine the positioning information in the transition scenario.

[0121]

[0122] Thus, the transformation of the positioning method from the outdoor to the indoor scene is realized. The switching of the positioning method from the indoor to the outdoor scene also includes: the recognition of the transition scene, the construction of the combined positioning model for the transition scene, and the smooth switching of the positioning method for the transition scene. The specific process refers to the process of realizing the transformation of the positioning method from the outdoor to the indoor scene. Thus, the whole process of a positioning method for the transition scene of a high-level autonomous driving vehicle with integrated multi-domain integration of heaven and earth is completed.

[0123] In summary, a positioning method for the transition scene of a high-level autonomous driving vehicle with integrated multi-domain integration of heaven and earth designed by the present invention solves the problems of decreased positioning accuracy and interrupted positioning signals in the transition scene, and achieves the purpose of high-precision positioning and smooth switching of the autonomous driving vehicle in the transition scene.

[0124] The above embodiments are only used to illustrate the design idea and characteristics of the present invention, and the purpose is to enable those skilled in the art to understand the content of the present invention and implement it accordingly. The protection scope of the present invention is not limited to the above embodiments. Therefore, all equivalent changes or modifications made according to the principles and design ideas disclosed by the present invention are within the protection scope of the present invention.

Claims

1. A positioning method for the transition scenario of an integrated space-earth multi-domain fusion autonomous vehicle, characterized in that, It includes the following steps: Step 1: Obtain the images during driving, and determine whether the vehicle reaches the transition scene area through visual image feature matching; Step 2: Construct a combined positioning model based on robust adaptive, where the combined positioning model includes a GNSS / LEO / IMU combined positioning model for vehicle positioning in outdoor areas, and a LiDAR / IMU combined positioning model for vehicle positioning in indoor areas; Step 3: Adopt an interactive multiple model algorithm to smoothly switch different outdoor and indoor positioning methods in the transition scene area.

2. The method for positioning a transition scenario of an integrated space-earth multi-domain autonomous driving vehicle according to claim 1, wherein The method for determining the transition scene is as follows: Step 1-1: Obtain the images during the driving of the autonomous vehicle; Step 1-2: Segment the image into multiple sub-image blocks, calculate the number of main color pixels in each sub-image, extract the main color and its pixel number in the sub-image block to construct the main color feature vector of the whole image; Step 1-3: Take pictures of the road scene by setting data sampling points for the transition scene, and form a transition scene database with the sampling point images according to the main color feature vector descriptor; Step 1-4: During the driving of the autonomous vehicle, match the collected image information with the transition scene image database to determine whether the vehicle reaches the transition scene area.

3. A method for positioning a transition scenario of an integrated space-earth multi-domain autonomous driving vehicle according to claim 2, characterized in that, In Step 1-2, the K-means clustering algorithm is used to calculate the number of main color pixels in each sub-image block. Specifically, all pixel points in the sub-image are regarded as data points, and the RGB value of each pixel point is used as its feature vector.

4. A method for positioning an autonomous vehicle in a transition scenario with integrated multi-domain integration of heaven and earth, according to claim 2, characterized in that, In Steps 1-4, the one-dimensional dynamic programming matching results of two main color feature vectors in each group are calculated in a loop. Let Γ au and Λ bv be the main color feature vectors of two images. The matching distance between the image to be matched and the images in the transition scene database is determined by means of the dynamic programming matching technique. When the matching distance is greater than the set threshold ζ, it is determined that the vehicle has reached the transition scene area.

5. A method for positioning a transition scenario of an integrated space-earth multi-domain fusion autonomous vehicle according to claim 1, characterized in that, Build the GNSS / LEO / IMU integrated positioning model based on GNSS, LEO, and IMU. The positioning information vector x output by the GNSS positioning module or the LEO positioning module G , and the positioning information vector x output by the IMU positioning module I , perform positioning information fusion, expressed as: And process the fused positioning information using a robust filtering algorithm.

6. A method for positioning a transition scenario of an integrated space-earth multi-domain autonomous driving vehicle according to claim 5, characterized in that, The basis for switching between the GNSS positioning module and the LEO positioning module is as follows: when the number of stably visible satellites is greater than or equal to 4, the positioning information vector x output by the GNSS positioning module is used G ; when the number of stably visible satellites is less than 4, the positioning information vector x output by the LEO is used G .

7. A method for positioning a transition scenario of an integrated space-earth multi-domain fusion autonomous vehicle according to claim 1, characterized in that, Construct the LiDAR / IMU integrated positioning model based on lidar and IMU, and the positioning information vector x solved based on LiDAR L and the positioning information vector x solved based on IMU I , and perform positioning information fusion, which is expressed as: And process the fused positioning information using a robust filtering algorithm.

8. A method for positioning a transition scenario of an integrated space-ground multi-domain autonomous driving vehicle according to claim 1, characterized in that, The scheme for constructing seamless indoor and outdoor positioning by adopting the interactive multiple model algorithm is as follows: Step 3-1: Interact the outdoor positioning results x obtained from the GNSS / LEO / IMU integrated positioning model and the indoor positioning results i (k-1) obtained from the LiDAR / IMU integrated positioning model with the transition probability matrix Π to respectively obtain the state (k-1) and variance q i of the hybrid system; i (k-1) and the transition probability matrix Π to respectively obtain the state of the hybrid system and variance q i (k-1); Step 3-2: Use the robust filtering algorithm to estimate and update the system state variables and state covariance, and incorporate the obtained results into the interactive link of the final output; Step 3-3: In the transition scene, update the model probability according to the likelihood function of the indoor and outdoor combined positioning model; Step 3-4: According to the updated model probability, weight and fuse the robust filtering results of each sub-model in the hybrid model set, output the optimal estimated state, and determine the transition scene positioning information.

9. A method for positioning an autonomous vehicle in a transition scenario with integrated multi-domain integration of heaven and earth, according to claim 4, characterized in that The dynamic programming matching technology determines the matching distance between the two as: Among them, Γ au represents the main color feature vector of the u-th currently acquired image, Λ bv represents the main color feature vector of the image in the v-th transition scene image database, d(Γ au , Λ bv ) represents the element distance between Γ au and Λ bv , Γ[a, u] represents the value of Γ au , Λ[b, v] represents the value of Λ bv , u = 1, 2, 3..., z, v = 1, 2, 3..., z, and z is the number of images.

Images