Road traffic sign updating method based on public source data

Abstract

The invention discloses a road traffic sign updating method based on crowd source data, and belongs to the field of road traffic sign updating, and the method comprises the steps: processing vehicle-mounted video image data through employing a pre-trained deep learning model, recognizing the type and attribute of the road semantic information of a traffic sign in a video, and obtaining the space coordinate of the traffic sign; performing spatial registration on the identified information of the traffic signs and an open source road data map; calculating a sign similarity value of the traffic sign after space registration, and judging whether the newly added traffic sign replaces the existing traffic sign or the newly added traffic sign; and generating an inference map after the traffic signs are updated, and updating the open source road data map. According to the method, the updating of the road traffic sign information is promoted, the updating efficiency and accuracy of the map data are remarkably improved, richer and more timely semantic support is provided for applications such as automatic driving, and the accuracy of traffic sign positioning and the consistency of semantic information updating are ensured.

Description

Technical Field

[0001] The present invention relates to the utilization of road traffic sign update technology, and specifically relates to a method for updating road traffic signs based on crowdsourced data. Background Art

[0002] With the rapid development of intelligent transportation and autonomous driving technologies, higher requirements are put forward for the real-time performance and accuracy of high-precision electronic maps.

[0003] In recent years, with the popularization of in-vehicle devices, using crowdsourced data for map updating has gradually become a research hotspot. Existing crowdsourced map updating methods mainly focus on the updating of road geometric information, such as identifying new roads or road network changes through GPS trajectory data. Related research includes discovering new roads using trajectory clustering, updating road networks based on trajectory matching, and extracting map elements using crowdsourced data. However, these methods mainly focus on the geometric shape and topological structure of the road network, and relatively few studies have been conducted on the automated collection and updating of road semantic information such as traffic signs, lacking the effective utilization of in-vehicle video data and failing to fully exploit the rich semantic information contained therein. Summary of the Invention

[0004] Aiming at the defects existing in the prior art, the present invention provides a method for updating road traffic signs based on crowdsourced data, which can achieve the accuracy of traffic sign updating.

[0005] The present invention adopts the following technical solutions: A method for updating road traffic signs based on crowdsourced data, comprising the following steps: Step 1: Process in-vehicle video image data using a pre-trained deep learning model to identify the types and attributes of the road semantic information of traffic signs in the video, and obtain the spatial coordinates of the traffic signs.

[0006] Step 2: Perform spatial registration on the information of the traffic signs identified in Step 1 with an open-source road data map.

[0007] Step 3: Calculate the sign similarity value for the traffic signs that have completed spatial registration, and determine whether the newly added traffic signs are to replace the existing traffic signs or are newly added traffic signs.

[0008] Step 4: Generate an inferred map after traffic sign update, and update the open-source road data map.

[0009] Preferably, Step 1 specifically includes the following steps: Step 1.1: Construct a traffic sign recognition model based on a convolutional neural network, and preprocess the in-vehicle video image data.

[0010] Step 1.2: Use the traffic sign recognition model to recognize the preprocessed in-vehicle video image data, and identify the types and attributes of the road semantic information of the traffic signs in the video. The types of road semantic information include traffic restriction, speed limit, height limit, turning, no parking, etc., and the attributes of road semantic information include speed limit value, traffic restriction time, height limit value, turning direction, no parking time, etc.

[0011] Step 1.3: Determine the shooting position according to the GPS timestamp and vehicle position information of the in-vehicle video image data, and calculate the spatial coordinates of the traffic signs through the principle of triangulation.

[0012] Preferably, Step 2 specifically includes the following steps: Step 2.1: Design the spatial registration strategy.

[0013] Step 2.1.1: Define the registration reference system. Step 2.1.2: Establish the spatial coordinate transformation model.

[0015] Step 2.2: Execute the registration process.

[0016] Step 2.2.1: Extract control points: Select k control points with known coordinates from the open-source road data map, and extract the coordinates of the corresponding k traffic signs from the recognition results of Step 1.

[0017] Step 2.2.2: Registration optimization: Use the least squares method to solve the affine transformation parameters and convert the coordinates of the traffic signs to the map coordinate system.

[0018] Step 2.3: Evaluate the registration quality.

[0019] Step 2.3.1: For each control point, calculate the Euclidean distance residual between the coordinates of the traffic sign after coordinate transformation and the coordinates of the control point: (1).

[0020] Where, represents the x-axis coordinate of the control point, represents the y-axis coordinate of the control point, represents the x-axis coordinate of the traffic sign after coordinate transformation, represents the y-axis coordinate of the traffic sign after coordinate transformation.

[0021] Then the root mean square error RMSE is: (2).

[0022] Step 2.3.2: Set the error threshold: When RMSE is greater than the error threshold, mark it as the registration result to be manually reviewed; when RMSE is less than the threshold, determine it as a valid registration.

[0023] Preferably, step 3 specifically includes the following steps: Step 3.1: Establish a spatial index, construct an R-tree index structure, design index parameters, and adopt a quadratic splitting algorithm for the splitting strategy.

[0024] Step 3.2: Implement a spatial query, define a search radius r, and execute a range query.

[0025] Step 3.3: Flag matching and update.

[0026] Step 3.3.1: Calculate the flag similarity value: .

[0027] is a traffic sign recognized in step 1, is a traffic sign at the corresponding position in the open-source road data map ; is the weight occupied by, is and the position distance similarity of.

[0028] = 1 - (4).

[0029] Among them, is and the Euclidean distance of.

[0030] is the weight occupied by, is and the type similarity of.

[0031] If and have the same type, then = 1, otherwise = 0.

[0032] is the weight occupied by, is and the attribute similarity of.

[0033] (5).

[0034] is and The matching degree of the i-th attribute. If the i-th attributes are the same, then ; if the i-th attributes are different, then .

[0035] is the weight coefficient of , and the sum is 1.

[0036] If there are multiple traffic signs at the corresponding position in the open-source road data map , calculate the sign similarity value between and all traffic signs at this position.

[0037] Step 3.3.2: Determine the matching strategy: Set the similarity threshold . When > , it is determined that and are the same traffic sign; when all ≤ , then is a newly added traffic sign.

[0038] When it is determined that is a newly added traffic sign, if the attributes of the traffic signs at the corresponding position in the open-source road data map are all different from the attributes of , then it is determined that is a newly added traffic sign; if the attributes of the traffic signs at the corresponding position in the open-source road data map are the same as the attributes of , then it is determined that is to replace the existing traffic sign.

[0039] Preferably, Step 4 specifically includes: If is a newly added traffic sign, then add a traffic sign at the corresponding position in the open-source road data map.

[0040] If is to replace the existing traffic sign, replace the traffic sign with the same attributes as at the corresponding position in the open-source road data map.

[0041] Finally, complete the update of the traffic signs in the open-source road data map.

[0042] The beneficial effects of the present invention are: The present invention first processes in-vehicle video image data using a deep learning model to identify the types and attributes of traffic signs, and then obtains the spatial coordinates of the traffic signs; geographical registration of the traffic signs is achieved through spatial indexing and precise coordinate matching, and by calculating the sign similarity value, it is determined whether the traffic signs need to be updated or added, and finally the open-source road data map is updated. The present invention promotes the update of road traffic sign information, significantly improves the update efficiency and accuracy of map data, provides richer and more timely semantic support for applications such as autonomous driving, and ensures the accuracy of traffic sign positioning and the consistency of semantic information update. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is the overall flowchart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0044] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Apparently, the described embodiments are some, but not all, of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts shall fall within the scope of protection of the present application.

[0045] The embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0046] As Figure 1 shown, the present invention provides a method for updating road traffic signs based on crowdsourced data, including the following steps: Step 1: Use a pre-trained deep learning model to process in-vehicle video image data, identify the types and attributes of the road semantic information of traffic signs in the video, and obtain the spatial coordinates of the traffic signs.

[0047] Specifically, it includes the following steps: Step 1.1: Construct a traffic sign recognition model based on a convolutional neural network to preprocess the in-vehicle video image data. Specifically for the preprocessing, image frames are extracted from the video at a frequency of 5 frames per second, image denoising is performed using a 3×3 Gaussian filter kernel (standard deviation σ = 0.8), the brightness and contrast are adjusted through histogram equalization, and finally the images are uniformly scaled to 224×224 pixels. The constructed traffic sign recognition model includes 5 convolutional blocks (each block contains two 3×3 convolutional layers and one 2×2 max pooling layer) and two fully connected layers (1024 and 512 neurons), and a softmax output layer is used for classification.

[0048] Step 1.2: Use the traffic sign recognition model to recognize the preprocessed in-vehicle video image data, and identify the types and attributes of the road semantic information of the traffic signs in the video. The types of road semantic information include prohibition types such as no overtaking, no honking, no passage for trucks, no passage for tricycles, no passage for animal-drawn / human-drawn vehicles, no passage for pedestrians, no passage for non-motor vehicles, no passage for dangerous goods vehicles, etc.; warning types such as watch for pedestrians, watch for children, watch for livestock, watch for wild animals, watch for rear-end collision risk, watch for icy road surface, watch for rain and snow weather, watch for crosswind, etc.; indication types such as stop and yield, slow down and yield, right of way, permitted U-turn, traffic restriction, speed limit, height limit, turning, etc.; guiding types such as intersection preview, direction of ferry / airport / port / railway station, service area / rest area / viewing platform, etc.; construction and emergency types such as construction area, emergency stopping strip, hospital / fire brigade / police assistance point, etc. The road semantic information of traffic signs covers all traffic signs.

[0049] The attributes of the road semantic information include speed limit value, traffic restriction time, height limit value, turning direction, no-parking time, etc.

[0050] Step 1.3: Determine the shooting location according to the GPS timestamp and vehicle position information of the in-vehicle video image data, and calculate the spatial coordinates of the traffic signs through the principle of triangulation. The method of calculating the spatial coordinates of traffic signs is a prior art and will not be elaborated here.

[0051] Step 2: Perform spatial registration on the information of the traffic signs recognized in Step 1 with the open-source road data map.

[0052] Specifically, it includes the following steps: Step 2.1: Design a spatial registration strategy.

[0053] Step 2.1.1: Define a registration reference system, and the WGS84 coordinate system is adopted in the spatial registration process.

[0054] Step 2.1.2: Establish a spatial coordinate transformation model.

[0055] The spatial coordinate transformation model is: (6).

[0056] , (7).

[0057] where (x,y) is the source coordinate of the traffic sign, (x',y') is the target coordinate, and (a,b,c,d,e,f) are the affine transformation parameters.

[0058] Step 2.2: Execute the registration process.

[0059] Step 2.2.1: Extract control points: Select k control points with known coordinates (such as intersections, turning points, etc.) from the open-source road data map, and extract the coordinates of the corresponding k traffic signs from the recognition results in Step 1.

[0060] Step 2.2.2: Registration optimization: Solve the affine transformation parameters using the least squares method to convert the coordinates of the traffic signs into the map coordinate system.

[0061] Step 2.3: Registration quality assessment.

[0062] Step 2.3.1: For each control point, calculate the Euclidean distance residual between the coordinates of the traffic sign after coordinate transformation and the control point coordinates: (1).

[0063] Where, represents the x-axis coordinate of the control point, represents the y-axis coordinate of the control point, represents the x-axis coordinate of the traffic sign after coordinate transformation, represents the y-axis coordinate of the traffic sign after coordinate transformation.

[0064] Then the root mean square error RMSE is: (2).

[0065] Step 2.3.2: Set the error threshold: When RMSE is greater than the error threshold, mark the registration result as pending manual review, and use the manually reviewed and corrected registration result as the valid registration for this point. When RMSE is less than the threshold, it is determined as a valid registration.

[0066] Step 3: Calculate the sign similarity value for the traffic signs that have completed spatial registration, and determine whether the newly added traffic sign is a replacement for an existing traffic sign or a newly added traffic sign.

[0067] Specifically, it includes the following steps: Step 3.1: Establish a spatial index, construct an R-tree index structure, design index parameters, and adopt the quadratic splitting algorithm for the splitting strategy.

[0068] Step 3.2: Implement spatial query, define the search radius r, and execute a range query.

[0069] Step 3.3: Sign matching and update.

[0070] Step 3.3.1: Calculate the sign similarity value: .

[0071] is a traffic sign recognized in Step 1, A traffic sign at the corresponding position in the open-source road data map ; is the weight occupied, is and the position distance similarity.

[0072] = 1 - (4).

[0073] Among them, is and the Euclidean distance.

[0074] is the weight occupied, is and the type similarity.

[0075] If and have the same type, then = 1, otherwise = 0.

[0076] is the weight occupied, is and the attribute similarity.

[0077] (5).

[0078] is and the matching degree of the i-th attribute. If the i-th attributes are the same, then , if the i-th attributes are different, then .

[0079] is 's weight coefficient, and the sum is 1.

[0080] If there are multiple traffic signs at the corresponding position in the open-source road data map , then calculate the sign similarity value with all traffic signs at this position.

[0081] Step 3.3.2: Determine the matching strategy: Set the similarity threshold , when there is > When determining and are the same traffic sign; when all of the following here ≤ then is a newly added traffic sign.

[0082] When it is determined that is a newly added traffic sign, if the attributes of the traffic sign at the corresponding position in the open-source road data map are all different from those of , then it is determined that is a newly added traffic sign; if the attributes of the traffic sign at the corresponding position in the open-source road data map are the same as those of , then it is determined that is to replace the existing traffic sign.

[0083] In one embodiment, there are two traffic signs at the corresponding position in the open-source road data map : Sign 1: speed limit value 40, effective time all day; Sign 2: speed limit value 60, effective time 7:00 - 20:00.

[0084] The traffic sign here extracted in step 1 is speed limit value 60, effective time 7:00 - 20:00.

[0085] with Sign 1's is 1 (coordinates are the same), is 1 (types are the same, both are speed limit types).

[0086] Because the speed limit values of 60 and 40 are different, then the first attribute speed limit value , effective time 7:00 - 20:00 ≠ all day, the second attribute speed limit time , then is 0.

[0087] Calculate the sign similarity value as: , the set similarity threshold is 0.8, then is a newly added traffic sign compared with Sign 1.

[0088] However, with Sign 2's is 1 (coordinates are the same), is 1 (types are the same, both are speed limit types).

[0089] Because the speed limit values of 60 and 60 are the same, then the first attribute speed limit value , effective time is the same, the second attribute speed limit time , then is 1.

[0090] Calculate the logo similarity value as: , the set similarity threshold is 0.8, then is the same traffic sign as compared with logo 2.

[0091] Therefore, it is proved that in the open-source road data map at the corresponding position there is the same traffic sign as, and the sign at this place on the map does not need to be updated subsequently.

[0092] On the contrary, when all the calculated logo similarity values here need to be less than or equal to the similarity threshold, then it can be determined that is a newly added traffic sign.

[0093] Step 4: Generate an inferred map after updating the traffic signs and update the open-source road data map.

[0094] Specifically include: If is a newly added traffic sign, then add a traffic sign at the corresponding position in the open-source road data map.

[0095] If is to replace an existing traffic sign, replace with a traffic sign having the same attribute at the corresponding position in the open-source road data map.

[0096] Finally, complete the update of the traffic signs in the open-source road data map.

[0097] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by those skilled in the art within the essence of the present invention should also fall within the protection scope of the present invention.

Claims

1. A method for updating road traffic signs based on crowdsourced data, characterized in that It includes the following steps: Step 1: Process the in-vehicle video image data using a pre-trained deep learning model to identify the types and attributes of the road semantic information of traffic signs in the video, and obtain the spatial coordinates of the traffic signs; Step 2: Perform spatial registration on the information of the traffic signs identified in Step 1 with the open-source road data map; Step 3: Calculate the sign similarity value for the traffic signs that have completed spatial registration, and determine whether the newly added traffic sign is a replacement for an existing traffic sign or a newly added traffic sign; Step 4: Generate an inferred map after the traffic signs are updated, and update the open-source road data map.

2. The method for updating road traffic signs based on crowdsourced data according to claim 1, wherein Step 1 specifically includes the following steps: Step 1.1: Construct a traffic sign recognition model based on a convolutional neural network to preprocess the in-vehicle video image data; Step 1.2: Use the traffic sign recognition model to identify the preprocessed in-vehicle video image data, and identify the types and attributes of the road semantic information of traffic signs in the video. The types of road semantic information include traffic restriction, speed limit, height limit, turning, and no parking. The attributes of road semantic information include speed limit value, traffic restriction time, height limit value, turning direction, and no parking time; Step 1.3: Determine the shooting position according to the GPS timestamp and vehicle position information of the in-vehicle video image data, and calculate the spatial coordinates of the traffic signs through the principle of triangulation.

3. A method for updating road traffic signs based on crowdsourced data according to claim 1, characterized in that Step 2 specifically includes the following steps: Step 2.1: Design a spatial registration strategy: Step 2.1.1: Define a registration reference system; Step 2.1.2: Establish a spatial coordinate transformation model; Step 2.2: Execute the registration process: Step 2.2.1: Extract control points: Select k control points with known coordinates from the open-source road data map, and extract the coordinates of the corresponding k traffic signs from the recognition results of Step 1; Step 2.2.2: Registration optimization: Use the least squares method to solve the affine transformation parameters and transform the coordinates of the traffic signs into the map coordinate system; Step 2.3: Registration quality assessment: Step 2.3.1: For each control point, calculate the Euclidean distance residual between the coordinates of the traffic sign after coordinate transformation and the control point coordinates: （1）； Among them, represents the x-axis coordinate of the control point, represents the y-axis coordinate of the control point, represents the x-axis coordinate of the traffic sign after coordinate transformation, represents the y-axis coordinate of the traffic sign after coordinate transformation; Then the root mean square error RMSE is: （2）； Step 2.3.2: Set an error threshold: When RMSE is greater than the error threshold, mark it as a registration result to be manually reviewed; when RMSE is less than the threshold, determine it as an effective registration.

4. A method for updating road traffic signs based on crowdsourced data according to claim 1, characterized in that, Step 3 specifically includes the following steps: Step 3.1: Establish a spatial index, construct an R-tree index structure, design index parameters, and use the quadratic splitting algorithm for the splitting strategy; Step 3.2: Implement spatial query, define a search radius r, and execute a range query; Step 3.3: Sign matching and update: Step 3.3.1: Calculate the sign similarity value: ； A traffic sign identified in step 1, is a traffic sign at the corresponding location in the open-source road data map; For the weight occupied is and the positional distance similarity =1- (4)； Among them, is and 's Euclidean distance; The weight occupied by is the type similarity between and ​ If and are of the same type, then = 1, otherwise = 0; For the weight occupied, is and the attribute similarity of; （5）； For and the matching degree of the i-th attribute, if the i-th attributes are the same, then , if the i-th attributes are different, then ; is the weight coefficient, and the sum is 1; If there are multiple traffic signs at the corresponding position in the open-source road data map then calculate the sign similarity values of all traffic signs at this position; Step 3.3.2: Determine the matching strategy: Set the similarity threshold When > it is determined that and are the same traffic sign; When all here ≤ then is a new traffic sign; After determining that is a newly added traffic sign, if the attributes of the traffic sign at the corresponding position in the open-source road data map are all different from those of , then determine that is a newly added traffic sign; if the attributes of the traffic sign at the corresponding position in the open-source road data map are the same as those of , then determine that is to replace the existing traffic sign.

5. A method for updating road traffic signs based on crowdsourced data according to claim 4, characterized in that, Step 4 specifically includes: If it is a newly added traffic sign, add the traffic sign at the corresponding position in the open-source road data map; If To replace the existing traffic signs, at the corresponding position in the open-source road data map Replace the traffic signs with the same attributes as it; Finally, complete the update of the traffic signs in the open-source road data map.

Images