Construction area determination method, device, equipment, medium and program product

Abstract

This disclosure provides a method, apparatus, equipment, medium, and program product for determining a construction area. The method involves determining first roadblock information and first dynamic object information within a preset range around a vehicle; determining first map element information within the preset range; and, based on the first roadblock information, first dynamic object information, and first map element information, determining the outline information of the construction area within the preset range using a pre-trained area detection model. The area detection model is a deep neural network model. By determining the construction area within the preset range based on the construction area outline information, the accuracy of the construction area detection results can be effectively improved.

Description

Technical Field

[0001] This disclosure relates to intelligent driving technology, and in particular to a method, apparatus, equipment, medium, and program product for determining a construction area. Background Technology

[0002] In intelligent driving scenarios such as autonomous driving and assisted driving, construction zones often involve temporary roadblocks, detours, and lane closures, altering the normal traffic flow and increasing the complexity and uncertainty of the driving environment. Therefore, identifying construction zones is crucial. Related technologies typically detect construction zones based on predefined rules (i.e., rule-based detection methods). However, rule-based detection methods rely on distance information between roadblocks, which makes them less adaptable to complex driving environments, resulting in lower accuracy in construction zone detection. Summary of the Invention

[0003] The embodiments of this disclosure provide a method, apparatus, equipment, medium, and program product for determining a construction area, so as to improve the accuracy of construction area detection results.

[0004] A first aspect of this disclosure provides a method for determining a construction area, comprising: determining first roadblock information and first dynamic object information within a preset range around a vehicle; determining first map element information within the preset range; determining the outline information of a construction area within the preset range based on the first roadblock information, the first dynamic object information, and the first map element information, using a pre-trained area detection model; wherein the area detection model is a deep neural network model; and determining the construction area within the preset range based on the outline information of the construction area.

[0005] A second aspect of this disclosure provides a device for determining a construction area, comprising: a first processing module for determining first roadblock information and first dynamic object information within a preset range around a vehicle; a second processing module for determining first map element information within the preset range; a third processing module for determining the outline information of a construction area within the preset range based on the first roadblock information, the first dynamic object information, and the first map element information, using a pre-trained area detection model; wherein the area detection model is a deep neural network model; and a fourth processing module for determining the construction area within the preset range based on the construction area outline information.

[0006] A third aspect of this disclosure is to provide a computer-readable storage medium storing a computer program for executing the method for determining a construction area as described in any of the above embodiments of this disclosure.

[0007] A fourth aspect of this disclosure provides an electronic device comprising: a processor; a memory for storing executable instructions of the processor; the processor being configured to read the executable instructions from the memory and execute the instructions to implement the method for determining a construction area as described in any of the above embodiments of this disclosure; or, the electronic device comprising: a device for determining a construction area as provided in any of the above embodiments.

[0008] A fifth aspect of this disclosure provides a computer program product that, when instructions in the computer program product are executed by a processor, performs the method for determining a construction area provided in any of the above embodiments of this disclosure.

[0009] Based on the method for determining the construction area provided in the above embodiments of this disclosure, the construction area can be detected by combining the road obstacle information, dynamic object information and map element information around the vehicle and the pre-trained area detection model. Since the area detection model is a deep neural network model, it can learn the complex relationship between the construction area and the road obstacle information, dynamic object information and map element information. Therefore, it can better adapt to complex environments and effectively improve the accuracy of the construction area detection results. Attached Figure Description

[0010] Figure 1 This is an exemplary application scenario of the method for determining the construction area provided in the embodiments of this disclosure;

[0011] Figure 2 This is a flowchart illustrating a method for determining a construction area provided in an exemplary embodiment of this disclosure;

[0012] Figure 3 This is a flowchart illustrating a method for determining a construction area provided in another exemplary embodiment of this disclosure;

[0013] Figure 4 This is a flowchart illustrating a method for determining a construction area provided in yet another exemplary embodiment of this disclosure;

[0014] Figure 5 This is a visual schematic diagram of various information in the vehicle coordinate system provided in an exemplary embodiment of this disclosure;

[0015] Figure 6 This is a flowchart illustrating a method for determining a construction area provided in yet another exemplary embodiment of this disclosure;

[0016] Figure 7 This is a schematic diagram of the outline information of the construction area provided in an exemplary embodiment of this disclosure;

[0017] Figure 8This is a schematic diagram of the training process of a region detection model provided in an exemplary embodiment of this disclosure;

[0018] Figure 9 This is a flowchart illustrating a method for determining a construction area provided in yet another exemplary embodiment of this disclosure;

[0019] Figure 10 This is a schematic diagram of the network structure of a region detection model provided in an exemplary embodiment of this disclosure;

[0020] Figure 11 This is a flowchart illustrating a method for determining a construction area provided in yet another exemplary embodiment of this disclosure;

[0021] Figure 12 This is a schematic diagram of the structure of a construction area determination device provided in an exemplary embodiment of this disclosure;

[0022] Figure 13 This is a schematic diagram of the structure of a construction area determination device provided in another exemplary embodiment of this disclosure;

[0023] Figure 14 This is a structural diagram of an electronic device provided in an embodiment of this disclosure. Detailed Implementation

[0024] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the present disclosure.

[0025] Those skilled in the art will understand that the terms "first," "second," etc., in the embodiments of this disclosure are only used to distinguish different steps, devices, or modules, and do not represent any specific technical meaning, nor do they indicate a necessary logical order between them.

[0026] It should also be understood that in the embodiments disclosed herein, "multiple" can refer to two or more, and "at least one" can refer to one, two or more.

[0027] It should also be understood that any component, data or structure mentioned in the embodiments of this disclosure can generally be understood as one or more unless expressly defined or given to the contrary in the context.

[0028] Furthermore, the term "and / or" in this disclosure is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this disclosure generally indicates that the preceding and following related objects have an "or" relationship.

[0029] It should also be understood that the description of the various embodiments in this disclosure emphasizes the differences between the various embodiments, and the similarities or similarities can be referred to each other. For the sake of brevity, they will not be described in detail.

[0030] At the same time, it should be understood that, for ease of description, the dimensions of the various parts shown in the accompanying drawings are not drawn according to actual scale.

[0031] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit this disclosure or its application or use.

[0032] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0033] It should be noted that similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures.

[0034] Summary of the disclosure

[0035] In developing this disclosure, the inventors discovered that in intelligent driving scenarios such as autonomous driving and assisted driving, construction zones are often accompanied by temporary roadblocks, detours, lane closures, and other changes that alter the normal traffic flow, increasing the complexity and uncertainty of the driving environment. Therefore, identifying construction zones is crucial. Related technologies typically detect construction zones based on predefined rules (i.e., rule-based detection methods). However, rule-based detection methods rely on distance information between roadblocks, which makes them less adaptable to complex driving environments, resulting in low accuracy in construction zone detection.

[0036] Exemplary overview

[0037] Figure 1 This is an exemplary application scenario of the method for determining the construction area provided in the embodiments of this disclosure. For example... Figure 1As shown, when the vehicle 11 is traveling on the road, it can perceive the surrounding environment information through the sensors 12 installed on the vehicle 11 and obtain the perception results. The sensors 12 may include one or more sensors. For example, the sensors 12 may include one or more of the following: camera, radar, lidar, millimeter-wave radar, ultrasonic radar, etc. Using the construction area determination method of the embodiments of this disclosure, road obstacle information (referred to as first road obstacle information) and dynamic object information (referred to as first dynamic object information) within a preset range around the vehicle can be determined based on the perception results. The first road obstacle information may include the position, size, orientation, type, and other status information of one or more road obstacles 13 (as shown, road obstacles 13 may include cones, fences, etc.). The first dynamic object information may include the trajectory, type, and other status information of other vehicles (referred to as other vehicles) 14, pedestrians, cyclists (not shown), and other dynamic objects around the vehicle 11. Furthermore, map element information (referred to as first map element information) within a preset range can be determined. The first map element information may include relevant information of elements such as lane lines 15 and curbs (i.e., road edges, not shown in the figure) in the map. Then, based on the first obstacle information, the first dynamic object information, and the first map element information, a pre-trained area detection model can be used to determine the outline information of the construction area within a preset range. Furthermore, based on the construction area outline information, the construction area within the preset range can be determined for downstream planning and control, driver alerts, etc., to ensure the driving safety of vehicle 11. Since the area detection model is a deep neural network model, it can learn the complex relationships between the construction area and obstacle information, dynamic object information, and map element information through pre-training. Therefore, it can better adapt to complex environments and effectively improve the accuracy of the construction area determination.

[0038] Exemplary method

[0039] Figure 2 This is a flowchart illustrating a method for determining a construction area provided in an exemplary embodiment of this disclosure. Embodiments of this disclosure can be applied to electronic devices, specifically, for example, to in-vehicle computing platforms (or in-vehicle terminals). Figure 2 As shown, the method provided in the embodiments of this disclosure may include the following steps:

[0040] Step 210: Determine the information of the first roadblock and the first dynamic object within a preset range around the vehicle.

[0041] The preset range around the vehicle (i.e., the vehicle itself) can be the range that the sensors on the vehicle can perceive. Sensors can include one or more of the following: cameras, radar, lidar, millimeter-wave radar, ultrasonic radar, etc. The first roadblock information is related information (or status information) about roadblocks within the preset range around the vehicle. Roadblocks represent obstacles placed on roads, typically used to block or restrict traffic. Roadblocks can be temporary or permanent, used for traffic control, construction zones, or for safety protection in special events and emergencies. Roadblocks come in various forms (or types), such as cones, fences, water-filled barriers, warning signs, sandbags, and concrete barriers. The status information of a roadblock can include at least one of the following: location, size, orientation, and type. The type information indicates the specific type of roadblock, including cones, fences, water-filled barriers, warning signs, sandbags, and concrete barriers. The first dynamic object information is the status information of dynamic objects within the preset range around the vehicle. The status information of dynamic objects can include the trajectory information and type information of the dynamic objects. The trajectory information includes the position, size, and orientation of the dynamic object at one or more moments (e.g., the current moment and historical moments). The type information of the dynamic object indicates its specific type. The type of dynamic object can include vehicles (i.e., other vehicles around the user), pedestrians, cyclists, etc. The size information indicates the size of the corresponding object (roadblock, dynamic object), and can include length, width, and height.

[0042] In some alternative embodiments, the first obstacle information and the first dynamic object information can be obtained by sensors on the vehicle.

[0043] In some optional embodiments, the first obstacle information and the first dynamic object information can be information in a specified coordinate system. For example, the first obstacle information and the first dynamic object information can be information in the vehicle coordinate system (i.e., the vehicle coordinate system) or other coordinate systems (e.g., the camera coordinate system, the radar coordinate system, etc.). The vehicle coordinate system is a coordinate system with the center of the rear axle of the vehicle as the origin. Coordinate systems can be converted to each other.

[0044] Step 220: Determine the information of the first map element within the preset range.

[0045] The first map element information describes relevant information about map elements within a preset range. This information may include status information for map elements of a specified type (e.g., lane lines, zebra crossings, curbs, medians, etc.) within the preset range. The status information of a map element may include descriptive information used to describe the element. Each map element has corresponding descriptive information. For example, lane line element information may include an ordered set of points or curve coefficients used to describe the lane line. Similarly, zebra crossing element information may include an ordered set of points used to describe the zebra crossing's outline, or the zebra crossing's center point location, orientation, size, etc., or the coordinates of each corner point. The specific method of describing map elements is not limited.

[0046] In some optional embodiments of this disclosure, information about a first map element within a preset range can be determined from map data. The map data can be pre-configured map data or map data constructed in real time; no specific limitation is made. Pre-configured map data can be, for example, high-precision map data, navigation map data, etc.

[0047] In some optional embodiments, the first map element information can be information in the vehicle coordinate system or other coordinate systems. For example, the first map element information can be information in a map coordinate system, which can be subsequently converted to the vehicle coordinate system.

[0048] In some optional embodiments, the first roadblock information, the first dynamic object information, and the first map element information can be information in the same coordinate system (e.g., the vehicle coordinate system).

[0049] It should be noted that the execution of steps 210 and 220 does not have to be in any particular order.

[0050] Step 230: Based on the first roadblock information, the first dynamic object information, and the first map element information, the outline information of the construction area within the preset range is determined by the pre-trained area detection model.

[0051] The region detection model (hereinafter referred to as the model) is a deep neural network model. The contour information of the construction area can be represented as a set of contour points or polygons; the specific representation method is not limited. A construction area refers to a specific area where construction, repair, maintenance, or other engineering work is underway. Potential obstacles or hazards such as heavy equipment, material storage, excavation, and high-altitude operations may exist within the construction area. Therefore, roadblocks are set up around the construction area to alert pedestrians and vehicles to safety and prevent them from entering the construction area.

[0052] In some optional embodiments, the construction area contour information within a preset range may include the contour information corresponding to each construction area within the preset range. For example, if two construction areas are detected within the preset range, the construction area contour information includes the contour information corresponding to each of the two construction areas.

[0053] In some alternative embodiments, the region detection model can employ any implementable deep neural network model. For example, the region detection model can be a Convolutional Neural Network (CNN) model, a Recurrent Neural Network (RNN) model, a Long Short-Term Memory (LSTM) neural network model, an attention-based model (Transformer model), and so on.

[0054] Step 240: Based on the outline information of the construction area, determine the construction area within the preset range.

[0055] Among them, the construction area contour information represents the boundary of the construction area. Therefore, based on the construction area contour information, the construction area within the preset range can be obtained. That is, the area within the contour of each area included in the construction area contour information is the construction area.

[0056] The method for determining the construction area provided in the embodiments of this disclosure can combine road obstacle information, dynamic object information and map element information around the vehicle, and detect the construction area through a pre-trained area detection model. Since the area detection model is a deep neural network model, it can learn the complex relationship between the construction area and road obstacle information, dynamic object information and map element information through pre-training. Therefore, it can better adapt to complex environments and effectively improve the accuracy of the construction area detection results.

[0057] Figure 3 This is a flowchart illustrating a method for determining a construction area provided in another exemplary embodiment of this disclosure.

[0058] In some alternative embodiments, in the above... Figure 2 Based on the illustrated embodiments, as Figure 3 As shown, step 210, determining the information of the first roadblock and the first dynamic object within a preset range around the vehicle, may include:

[0059] Step 2110: Obtain sensor data collected by the sensors on the vehicle.

[0060] The sensor data may include at least one of the following: images, radar point cloud data, ultrasonic radar data, millimeter-wave radar data, and lidar data.

[0061] Step 2120: Based on sensor data, determine the information of the first roadblock and the information of the first dynamic object.

[0062] Specifically, based on sensor data, pre-configured perception algorithms or models can be used to determine the information of the first roadblock and the first dynamic object. Perception algorithms or models may include, for example, target detection algorithms or models, semantic segmentation algorithms or models, target tracking algorithms, etc.

[0063] In some optional embodiments, the vehicle can acquire sensor data collected by the sensors in real time during the driving process, in order to determine the first roadblock information and the first dynamic object information corresponding to the current moment.

[0064] In embodiments of this disclosure, information about a first road obstacle and a first dynamic object around the vehicle is obtained through sensor data on the vehicle, providing effective feature references for the detection of the construction area and improving the accuracy of the construction area detection.

[0065] Figure 4 This is a flowchart illustrating a method for determining a construction area provided in yet another exemplary embodiment of this disclosure.

[0066] In some optional embodiments, step 220, determining the first map element information within a preset range, may include:

[0067] Step 2210: Obtain the vehicle's pose information in the map coordinate system.

[0068] The vehicle's pose information in the map coordinate system can be obtained through the vehicle's positioning function. For example, the vehicle's pose information in the map coordinate system can be obtained through a high-precision positioning system on the vehicle. Alternatively, the vehicle's pose information in the map coordinate system can be determined through a pre-configured positioning algorithm on the vehicle. The positioning algorithm can be any feasible algorithm; for example, it can be a Simultaneous Localization and Mapping (SLAM) algorithm or other positioning algorithms.

[0069] Step 2220: Based on the pose information, determine the first map element information within a preset range from the pre-configured map data.

[0070] The pre-configured map data can be high-precision map data, navigation map data, etc., and there are no specific restrictions.

[0071] In some optional embodiments, based on the vehicle's pose information and a preset range, description information of a specified type of map element within a preset range can be extracted from pre-configured map data and used as the first map element information.

[0072] The embodiments of this disclosure determine map element information within a preset range based on the current pose of the vehicle, providing effective map element reference information for construction area detection. This can be combined with road obstacle information and dynamic object information to improve the accuracy of construction area detection.

[0073] In some optional embodiments, the first roadblock information includes at least one of the following: position information, size information, orientation information, and type information of a roadblock of a first preset type within a preset range in the vehicle coordinate system.

[0074] The first preset type can include roadblock types such as cones, fences, and water-filled barriers. The vehicle coordinate system is the vehicle's own coordinate system with its rear axle center as the origin. Position information includes the three-axis coordinates of the roadblock in the vehicle coordinate system, which can be represented as (x, y, z). Dimension information includes the length, width, and height of the roadblock, which can be represented as (length, width, height). Orientation information indicates the roadblock's orientation, which can be represented as "heading". Type information characterizes the roadblock type, which can be represented as "type".

[0075] For example, the roadblock information corresponding to each roadblock can be represented as (x, y, z, width, length, height, heading, type).

[0076] In some optional embodiments, the first dynamic object information includes trajectory information of a second preset type of dynamic object within a preset range in the vehicle coordinate system.

[0077] The second preset type can include dynamic object types such as vehicles, pedestrians, and cyclists. The trajectory information of the dynamic object in the vehicle coordinate system can include the state information of the dynamic object in one or more frames (i.e., multiple moments) in the vehicle coordinate system. The state information at each moment includes the position information, size information, and orientation information of the dynamic object. The first dynamic object information also includes the type information of the dynamic object.

[0078] For example, the dynamic object information for each frame can be represented as (x, y, z, width, length, height, heading, type).

[0079] In some optional embodiments, the first map element information includes descriptive information of each map element within a preset range in the vehicle coordinate system.

[0080] The description information of each map element in the vehicle coordinate system can adopt any feasible description method, such as ordered point sets, shape parameters, etc. For example, the description information of a lane line element can be an ordered point set collected on the lane line element or curve parameters (such as the slope and intercept of a straight line, the curve coefficient of a quadratic curve, etc.). The description information of a zebra crossing element can be an ordered point set of the zebra crossing outline or zebra crossing area outline parameters. The ordered point set of the zebra crossing outline is a set of coordinate points collected on the zebra crossing outline, and the outline parameters can include the coordinates of the outline corner points or the coordinates of the zebra crossing center point, orientation, size, etc. The specific description method is not limited.

[0081] In some optional embodiments, for linear elements such as lane lines and curbs, an element length threshold can be set. Based on the element length threshold, the linear element can be divided into one or more elements whose length does not exceed the element length threshold. For example, if a lane line element is 200 meters long and the element length threshold is 100 meters, then the 200-meter lane line element can be divided into two 100-meter lane line elements.

[0082] In some alternative embodiments, Figure 5 This is a visual illustration of various information in the vehicle coordinate system provided in an exemplary embodiment of this disclosure. For example... Figure 5 As shown, the first type of roadblock information includes the location, size (visualized as a rectangular area in the figure), orientation (visualized as a white arrow on the rectangular area), and type (not shown in the figure) of various roadblocks in the vehicle coordinate system. In practical applications, different types of roadblocks can be visualized as different shapes or colors, or their type information (such as a preset type number) can be displayed on the rectangular area. The second preset type of dynamic object includes other vehicles, pedestrians, and cyclists. Similar to roadblocks, the first dynamic object information includes the location, size (visualized as a rectangular area in the figure), orientation (visualized as a white arrow on the rectangular area), and type information of each dynamic object. The information of the dynamic object at multiple moments constitutes the trajectory information of the dynamic object. Different types of dynamic objects can use different visualization methods; dynamic objects and roadblocks can use different visualization methods, which are not specifically limited. Map elements include zebra crossings, lane lines, and medians.

[0083] In the embodiments of this disclosure, specific road obstacle information, dynamic object information, and map element information in the vehicle coordinate system are used as input features of the area detection model to detect construction areas. Since construction areas have complex relationships with road obstacles, map elements, and dynamic objects, for example, areas with dynamic object trajectories are usually not construction areas, while areas with road obstacles may be construction areas, etc. Therefore, by combining road obstacle information, dynamic object information, and map element information, and learning the complex relationships through a deep neural network model, the accuracy of the detection results of construction areas can be effectively improved.

[0084] Figure 6 This is a flowchart illustrating a method for determining a construction area provided in yet another exemplary embodiment of this disclosure.

[0085] In some alternative embodiments, based on any of the above embodiments, such as Figure 6 As shown, step 230, which determines the outline information of the construction area within a preset range based on the first roadblock information, the first dynamic object information, and the first map element information, using a pre-trained area detection model, may include:

[0086] Step 2310: Determine the input feature data based on the first roadblock information, the first dynamic object information, and the first map element information.

[0087] Specifically, the information on the first roadblock, the first dynamic object, and the first map can be fused according to a preset fusion method, and the fusion result can be used as the input feature data of the model. The preset fusion method may include concatenation, addition, convolution, etc.

[0088] In some optional embodiments, the roadblock information corresponding to each roadblock, the object information corresponding to each dynamic object, and the map element information corresponding to each map element can be concatenated into a vector, and this vector can be used as input feature data.

[0089] Step 2320: Based on the region detection model, the input feature data is processed to obtain the contour point set corresponding to the construction area.

[0090] The input feature data can be used as input to the region detection model, and the contour point set corresponding to the construction area can be obtained through the inference of the region detection model.

[0091] In some optional embodiments, a region detection model can detect one or more contour point sets corresponding to construction regions. Each contour point set corresponding to a construction region may include a preset number of coordinate points on the contour of the construction region. The preset number can be set to any number according to actual needs; for example, the preset number can be 20, 30, 40, etc.

[0092] In some optional embodiments, the set of contour points corresponding to the construction area can be a set of coordinate points in the vehicle coordinate system. For example, taking 30 points as an example, the set of contour points can be represented as ((x1,y1),(x2,y2),…,(x...). 30 ,y 30 )). (x i ,y i ) represents the longitudinal and lateral coordinates of the i-th contour point in the vehicle coordinate system.

[0093] Step 2330: Determine the contour information of the construction area based on the first roadblock information and the contour point set corresponding to the construction area.

[0094] Specifically, the outline information of the construction area can be determined based on the positional relationship between each roadblock in the first roadblock information and the corresponding outline point set of the construction area.

[0095] In some optional embodiments, the set of contour points corresponding to the construction area constitutes the initial contour of the construction area, and the accurate contour information of the construction area is determined by combining the first roadblock information and the initial contour.

[0096] In the embodiments of this disclosure, the contour point set of the construction area is detected by the area detection model, and then combined with the first roadblock information to determine the contour information of the construction area. Since the roadblocks located within the initial contour formed by the contour point set in the first roadblock information represent the boundary of the construction area, combining the roadblock information can improve the accuracy of the contour information of the construction area, thereby improving the accuracy of the construction area.

[0097] In some optional embodiments, step 2330, which determines the contour information of the construction area based on the first roadblock information and the contour point set corresponding to the construction area, may include:

[0098] For each construction area, based on the contour point set corresponding to the construction area and the first roadblock information, the roadblock information corresponding to the target roadblock in the construction area is determined; based on the roadblock information corresponding to the target roadblock in the construction area, the contour point set corresponding to the construction area is adjusted to obtain the contour information of the construction area.

[0099] Specifically, based on the positional relationship between the contour point set corresponding to the construction area and each obstacle in the first obstacle information, the obstacle information corresponding to the target obstacle in the construction area can be determined from the first obstacle information. The contour point set corresponding to the construction area is then adjusted according to a pre-configured adjustment method to obtain the contour information of the construction area.

[0100] In some optional embodiments, the minimum hull polygon of each target roadblock within the construction area can be determined as the outline information of the construction area. The hull polygon can be either a convex hull polygon or a concave hull polygon, and there is no specific limitation.

[0101] In some alternative embodiments, Figure 7 This is a schematic diagram illustrating the outline information of a construction area provided in an exemplary embodiment of this disclosure. For example... Figure 7 As shown, the set of contour points obtained by the model detection constitutes the initial contour of the construction area. The target roadblocks in the construction area are determined by combining the initial contour and the location information of the roadblocks. The initial contour is adjusted based on the target roadblocks in the construction area to obtain the adjusted contour (i.e., the contour corresponding to the contour information of the construction area).

[0102] In the embodiments of this disclosure, after obtaining the contour point set of the construction area through model detection, the contour point set is further adjusted by combining the road obstacles within the contour formed by the contour point set. The adjusted contour information is then determined as the contour information of the construction area, making the contour of the construction area more consistent with the road obstacle situation, thereby improving the accuracy and reliability of the construction area.

[0103] In some optional embodiments, the input feature data is processed based on a region detection model to obtain a set of contour points corresponding to the construction area. This may include: processing the input feature data based on a region detection model to obtain a set of candidate contour points corresponding to the candidate construction area and a confidence level for each candidate contour point set; and determining the set of contour points corresponding to the construction area from the set of candidate contour points corresponding to the candidate construction area based on the confidence level for each candidate contour point set.

[0104] The confidence level of the candidate contour point set corresponding to each candidate construction area represents the degree to which the candidate construction area is believed to belong to the construction area. Construction areas can be determined from the candidate construction areas based on the confidence level and a pre-configured confidence threshold. Each candidate construction area corresponds to a set of candidate contour points, which constitutes the candidate contour point set for that candidate construction area. Candidate construction areas with excessively low confidence levels can be filtered out based on the confidence threshold, while candidate construction areas with confidence levels greater than the threshold are determined as construction areas, thus obtaining the contour point set corresponding to each construction area.

[0105] In the embodiments of this disclosure, candidate construction areas are screened based on their confidence level, and candidate construction areas with low confidence levels are filtered out, which can improve the effectiveness and reliability of construction areas.

[0106] Figure 8 This is a schematic diagram of the training process of a region detection model provided in an exemplary embodiment of this disclosure.

[0107] In some alternative embodiments, based on any of the above embodiments, such as Figure 8 As shown, the region detection model is obtained in the following way:

[0108] Step 310: Obtain training sample data; the training sample data includes at least one training sample and the construction area outline label corresponding to each training sample; each training sample includes the second roadblock information, the second dynamic object information and the second map element information corresponding to the vehicle at a certain time point.

[0109] The construction area contour label can include a preset number of contour point ground truth values ​​for the construction area. The second roadblock information, second dynamic object information, and second map element information are similar to the first roadblock information, first dynamic object information, and first map element information described above, and will not be repeated here. Training samples can be pre-collected. For example, during a vehicle's journey through a road section with an actual construction area, sensor data is collected at each moment, and the vehicle's pose information at each moment (i.e., time point) is recorded. Training samples are obtained based on the sensor data at each moment, and these training samples are labeled to obtain the construction area contour label. The labeling method can include manual labeling, semi-automatic labeling, and automatic labeling.

[0110] Step 320: Based on the training sample data, train the initial region detection model to obtain the region detection model.

[0111] The initial region detection model can be an initialized model with initialized network parameters. Alternatively, the initial region detection model can be a model that has been updated through a certain number of iterations, with updated network parameters.

[0112] In some optional embodiments, taking one iteration as an example, each training sample can be input into the initial region detection model to obtain the construction area contour prediction point set corresponding to each training sample. Based on the construction area contour prediction point set and construction area contour label corresponding to each training sample, the model loss is determined, and the network parameters of the model are updated based on the model loss to obtain the updated region detection model. If the updated region detection model meets the training termination condition, training ends, and the updated region detection model is used as the trained region detection model. If the updated region detection model does not meet the training termination condition, the updated region detection model is used as the initial region detection model, and the next iteration is performed according to the above process until the updated region detection model meets the training termination condition, and the region detection model is obtained. The training termination condition may include model convergence, the number of iterations reaching a threshold, etc., and is not specifically limited. For updating the network parameters, a pre-configured optimizer can be used based on the model loss. The optimizer may include stochastic gradient descent, adaptive learning rate optimization algorithm, batch gradient descent, etc. The specific optimization algorithm is not limited.

[0113] In the embodiments of this disclosure, training samples consisting of second roadblock information, second dynamic object information, and second map element information are used to train the region detection model. This enables the region detection model to learn the complex relationship between the construction area and the roadblock information, dynamic object information, and map element information, thereby improving the model's generalization ability and robustness, and ultimately improving the accuracy and effectiveness of the detection results for the construction area.

[0114] Figure 9 This is a flowchart illustrating a method for determining a construction area provided in yet another exemplary embodiment of this disclosure.

[0115] In some alternative embodiments, based on any of the above embodiments, such as Figure 9 As shown, step 230, which determines the outline information of the construction area within a preset range based on the first roadblock information, the first dynamic object information, and the first map element information, using a pre-trained area detection model, may include:

[0116] Step 23a0: Extract features from the first roadblock information to obtain the first roadblock features.

[0117] Specifically, the first roadblock information can be feature extracted based on the first feature extraction network in the regional detection model to obtain the first roadblock features.

[0118] In some alternative embodiments, the first feature extraction network can be any implementable feature extraction network. For example, it can be a convolutional neural network, a multi-layer perceptron, a Transformer, etc., without any specific limitation.

[0119] Step 23b0: Extract features from the information of the first dynamic object to obtain the features of the first dynamic object.

[0120] In this process, the first dynamic object information can be feature extracted based on the second feature extraction network in the region detection model to obtain the features of the first dynamic object.

[0121] In some optional embodiments, the second feature extraction network can be any implementable feature extraction network. For example, it can be a convolutional neural network, a multi-layer perceptron, a Transformer, etc., without any specific limitation.

[0122] Step 23c0: Extract features from the first map element information to obtain the features of the first map element.

[0123] In this process, the first map element information can be extracted based on the third feature extraction network in the region detection model to obtain the first map element features.

[0124] In some alternative embodiments, the third feature extraction network can be any implementable feature extraction network. For example, it can be a convolutional neural network, a multi-layer perceptron, a Transformer, etc., without any specific limitation.

[0125] In some optional embodiments, the first feature extraction network, the second feature extraction network, and the third feature extraction network may adopt the same network structure, and the network parameters may be obtained according to the actual training, that is, the network parameters of the first feature extraction network, the second feature extraction network, and the third feature extraction network may be different.

[0126] It should be noted that the execution of steps 23a0, 23b0 and 23c0 does not have to be in any particular order.

[0127] Step 23d0: Determine the fusion features based on the first roadblock features, the first dynamic object features, and the first map element features.

[0128] Among them, the first roadblock feature, the first dynamic object feature, and the first map element feature can be fused based on the feature fusion network in the regional detection model to obtain fused features.

[0129] In some alternative embodiments, the feature fusion network can be implemented based on any feasible network. For example, the feature fusion network can be implemented using a multilayer perceptron, a Transformer encoder, etc. The Transformer encoder, based on an attention mechanism, fuses the features of a first obstacle, a first dynamic object, and a first map element to obtain global information.

[0130] Step 23e0: Decode the fused features to obtain the decoding result.

[0131] In this method, the fused features can be decoded based on the decoder in the region detection model to obtain the decoding result.

[0132] In some alternative embodiments, the decoder can adopt a Transformer decoder family architecture, such as a Deformable Transformer Decoder, a Masked Transformer Decoder, a Detection Transformer (DETR), a Deformable Detection Transformer (Deformable DETR), and so on. The specific decoder architecture is not limited.

[0133] In some optional embodiments, the fused features can be decoded based on a reference contour point set to obtain a decoding result. That is, the fused features and the reference contour point set are input to a decoder, and the decoder decodes to obtain the decoding result. The reference contour point set can be an initialized contour point set. The reference contour point set can include initial contour point sets corresponding to M initial candidate construction areas. During the decoding process, the initial contour point set is continuously updated so that the updated contour point set continuously approaches the contour of the construction area, resulting in an effective decoding result.

[0134] Step 23f0: Based on the decoding results, determine the outline information of the construction area.

[0135] Specifically, the contour information of the construction area can be determined based on the head network layer in the region detection model. For example, the head network can include a regression head network layer, which is used to regress the decoding results to obtain a set of candidate contour points for M candidate construction areas, and then determine the contour information of the construction area based on the set of candidate contour points.

[0136] In some optional embodiments, the head network layer may further include a classification head network layer for obtaining the confidence levels corresponding to the M candidate construction areas based on the decoding results.

[0137] In some alternative embodiments, the head network layer can adopt any implementable network structure. For example, it can adopt a convolutional neural network, a multilayer perceptron, a fully connected layer, etc., without any specific limitation.

[0138] In the embodiments of this disclosure, by extracting roadblock features, dynamic object features, and map element features respectively, local features of various types of information can be obtained. Then, through feature fusion, global information can be obtained. By decoding the global information, decoded features associated with the construction area are obtained, and the contour information of the construction area can be effectively determined based on these decoded features. Due to the effective fusion of various features, and based on the complex correlation between the construction area and various features learned by the model, accurate and effective contour information of the construction area can be obtained, thereby improving the accuracy of the detected construction area.

[0139] In some optional embodiments, step 23d0, which determines the fusion features based on the first roadblock features, the first dynamic object features, and the first map element features, may include: determining the fusion features based on the first roadblock features, the first dynamic object features, and the first map element features using a first number of encoders; each encoder includes at least one encoding attention layer.

[0140] The first quantity can be set according to actual needs; for example, it can be set to 1, 2, 3, 4, etc. The encoding attention layer (or encoder attention layer) is a network layer based on an attention mechanism, which captures contextual information of various features. The attention mechanism can be a single-head attention mechanism or a multi-head attention mechanism; there is no specific limitation. The number of encoding attention layers included in each encoder is not limited.

[0141] In some alternative embodiments, each encoder may be a Transformer encoder. Accordingly, the encoding attention layer is a Transformer Encoder Attention Layer (or Transformer Encoding Attention Layer).

[0142] In some optional embodiments, the encoder may include other related network layers in addition to the encoding attention layer, such as feedforward neural network layers, residual connections, etc. The residual connections are used to fuse features from different levels, such as fusing shallow features with deep features to improve the comprehensiveness of feature information. The specific encoder structure can be set according to actual needs. Alternatively, each encoding attention layer may include an attention layer, a feedforward neural network layer, residual connections, etc.

[0143] The embodiments of this disclosure, through multiple coded attention layers, can effectively capture the contextual information between fault features, dynamic object features, and map element features, ensuring the accuracy of construction area detection results.

[0144] In some optional embodiments, step 23e0 of decoding the fused features to obtain a decoding result includes: decoding the fused features through a second number of decoders to obtain a decoding result; each decoder includes at least one decoding attention layer.

[0145] The second quantity can be set to any number according to actual needs; for example, the second quantity can be 1, 2, 3, 4, etc. The decoding attention layer (or decoder attention layer) is a network layer based on the attention mechanism.

[0146] In some alternative embodiments, the decoder may be a Transformer decoder. The decoding attention layer is a Transformer decoding attention layer. For example, a multi-head attention layer, a masked attention layer, etc.

[0147] In some optional embodiments, the decoder may further include other related network layers, such as feedforward neural network layers and residual connections, in addition to the decoding attention layer. The specific decoder structure can be set according to actual needs. Alternatively, each decoding attention layer may include an attention layer, a feedforward neural network layer, residual connections, etc.

[0148] The embodiments of this disclosure can effectively capture the dependency between the output sequence (contour point set) and the fused features through a multi-layer decoding attention layer, thereby improving the effectiveness of the detection results.

[0149] In some optional embodiments, step 23f0, which determines the outline information of the construction area based on the decoding result, may include:

[0150] Based on the decoding results, a set of candidate contour points corresponding to the candidate construction area is determined through a regression head network layer; based on the decoding results, a confidence level corresponding to each candidate contour point set is determined through a classification head network layer; based on the candidate contour point sets and the confidence level corresponding to each candidate contour point set, the contour information of the construction area is determined.

[0151] The specific structures of the regression head network layer and the classification head network layer can be any implementable structure, and this disclosure does not limit them. For example, convolutional networks, multilayer perceptrons, fully connected layers, etc. The specific operations for determining the construction area contour information based on the confidence level can be found in the foregoing embodiments.

[0152] In some alternative embodiments, Figure 10 This is a schematic diagram of the network structure of a region detection model provided in an exemplary embodiment of this disclosure. For example... Figure 10As shown, the network structure of the region detection model includes a feature extraction network, a feature fusion network, a decoding network, and a head network. The feature extraction network is used for feature extraction, specifically comprising a first feature extraction network, a second feature extraction network, and a third feature extraction network. The first feature extraction network extracts roadblock features (which are the first roadblock features during the inference phase). The second feature extraction network extracts dynamic object features. The third feature extraction network extracts map element features. The first, second, and third feature extraction networks in the figure use the same network structure as an example. That is, each includes m (m is a positive integer) sets of MLPs and Add&Norm (addition and normalization layers), as well as Max pooling&Norm (max pooling and normalization layers). ×m represents the stacking of m sets of MLPs and Add&Norm (i.e., the gray blocks in the feature extraction network in the figure). The feature fusion network includes n (n is a positive integer) stacked Transformer Encoder Attention Layers. The first roadblock feature, the first dynamic object feature, and the first map element feature are encoded and fused through the feature fusion network to obtain the fused feature. The decoding network consists of s (where s is a positive integer) stacked Transformer Decoder Attention Layers. The decoding network decodes the fused features to obtain the decoding result. The decoding network also receives a reference contour point set (points query), which can be found in the previously described embodiment. The head network consists of a Regression Head network layer and a Classification Head network layer. The Regression Head network layer regresses the decoding result to obtain candidate contour point sets. The Classification Head network layer processes the decoding result to obtain the confidence score for each candidate contour point set. Based on the confidence score and a confidence threshold, target contour point sets with confidence scores greater than the threshold are selected from the candidate contour point sets as the contour point sets corresponding to the construction area. The target roadblocks enclosed by each contour point set are determined from the first roadblock information. The contour point sets are adjusted based on the roadblock information corresponding to the target roadblocks to obtain accurate and effective construction area contour information. The values ​​m, n, and s can be set according to actual needs. For example, m = 3, n = 6, s = 6, or m, n, and s can be other values.

[0153] It should be noted that, Figure 10 This is only an exemplary network structure diagram. In practical applications, the network components are not limited to the structure shown in the diagram.

[0154] It should be noted that the model inference process during the training phase of the region detection model is consistent with that during the prediction and application phase. That is, the above-mentioned feature extraction, feature fusion, decoding, regression and classification are performed on the second roadblock information, the second dynamic object information and the second map element information of the training samples. These will not be elaborated on here.

[0155] Figure 11 This is a flowchart illustrating a method for determining a construction area provided in yet another exemplary embodiment of this disclosure.

[0156] In some alternative embodiments, based on any of the above embodiments, such as Figure 11 As shown, the method in this embodiment of the disclosure may further include:

[0157] Step 250: In response to the existence of at least one construction area within the preset range, output construction area prompt information.

[0158] The construction area prompt information can include information such as the direction and distance of the construction area relative to the vehicle. The output method can be any method, such as voice output, screen display, or a combination of voice and screen display. For example, the voice output could say, "Construction is underway 200 meters ahead in the current lane. Please drive carefully." The specific prompt content can be set according to actual needs, and this embodiment does not limit it. Another example is displaying the vehicle's surrounding environment information on the screen, highlighting the construction area.

[0159] In embodiments of this disclosure, when a construction area is detected, a construction area warning message can be output to remind the driver to pay attention to the road conditions in a timely manner, so that the driver can take timely countermeasures, such as changing lanes or changing routes, thereby improving driving safety.

[0160] In some alternative embodiments, based on any of the above embodiments, such as Figure 11 As shown, the method in this embodiment of the disclosure may further include:

[0161] Step 410: Plan the driving route based on the construction area within the preset range.

[0162] In this way, the driving route can be comprehensively planned by taking into account the occupation of the driving road by the construction area, as well as other obstacles and traffic participants on the road, so as to effectively avoid the construction area and ensure the safety of vehicle driving.

[0163] Step 420: Control the vehicle's operating status based on the driving route.

[0164] After planning the driving route, the vehicle's operation can be controlled according to the planned route, so that the vehicle can travel along the planned route and effectively avoid the construction area.

[0165] The embodiments of this disclosure use the detected construction area for vehicle planning and control, which enables the vehicle to automatically avoid the construction area and ensure vehicle driving safety.

[0166] It should be noted that when steps 410, 420 and 250 are implemented in combination, the planning and control process of steps 410 to 420 and the prompting process of step 250 are not in any particular order.

[0167] In related technologies, rule-based detection methods rely on distance information between obstacles, which has many limitations in complex driving environments and poor adaptability to scene complexity. Construction areas are highly diverse and uncertain, making it difficult for rules to cover all scenarios. Furthermore, to expand the recognition of construction areas in new scenarios, rule-based detection methods typically require manual addition and debugging of new rules. This manual rule updating method is time-consuming, inefficient, and difficult to adapt to rapid environmental changes. The continuous accumulation of rules also makes the intelligent driving system complex and difficult to maintain, and different rules may interfere with each other, leading to instability.

[0168] To address the aforementioned issues, the method for determining construction areas provided in this disclosure utilizes environmental obstacles, the trajectories of traffic participants (dynamic objects), and map information. Through a deep learning model, it accurately identifies construction areas in real time, providing information such as the location, shape, and size of the construction area to effectively distinguish between drivable areas and construction areas to avoid. Since construction areas may contain potential obstacles such as workers, temporary equipment, and excavation pits, accurate identification of construction areas allows vehicles to slow down, maintain a greater distance, or stop in time, ensuring the safety of pedestrians and workers. Furthermore, accurate identification of construction areas contributes to a smooth driving experience, reducing sudden stops and sharp turns, and improving passenger comfort and confidence. In addition, the layout of construction areas may change daily. Real-time accurate identification of construction areas allows vehicles to flexibly adjust their driving paths and speeds, avoiding traffic accidents caused by environmental changes and effectively helping intelligent driving systems make safer, more compliant, and efficient decisions. Moreover, the deep learning model can effectively learn the complex relationships between construction areas, obstacles, dynamic object trajectories, and map information, giving the model high generalization ability and robustness, enabling it to adapt to more complex environments. Furthermore, when new scenarios arise, the model can be trained or fine-tuned based on training samples from the new scenarios, enabling the model to quickly adapt to the identification of construction areas in new scenarios, effectively improving efficiency, reducing time costs, and enhancing model performance. Therefore, the method for determining construction areas provided by the embodiments of this disclosure can solve the aforementioned problems of rule-based detection methods in related technologies, avoid manual rule maintenance, effectively reduce system complexity and maintenance difficulty, and effectively improve the accuracy of construction area identification.

[0169] The embodiments described above can be implemented individually or in any combination without conflict. The specific implementation can be set according to actual needs, and this disclosure does not limit them.

[0170] Any of the methods for determining a construction area provided in this disclosure can be executed by any suitable electronic device with data processing capabilities, including but not limited to: terminal devices and servers. Alternatively, any of the methods for determining a construction area provided in this disclosure can be executed by a processor, such as by a processor executing any of the methods for determining a construction area mentioned in this disclosure by calling corresponding instructions stored in memory. Further details will not be elaborated below.

[0171] Exemplary device

[0172] Figure 12 This is a schematic diagram of a construction area determination device provided in an exemplary embodiment of this disclosure. The device in this embodiment can be used to implement corresponding method embodiments of this disclosure, such as… Figure 12 The apparatus shown may include: a first processing module 51, a second processing module 52, a third processing module 53, and a fourth processing module 54.

[0173] The first processing module 51 is used to determine the information of the first roadblock and the first dynamic object within a preset range around the vehicle.

[0174] The second processing module 52 is used to determine the information of the first map element within a preset range.

[0175] The third processing module 53 is used to determine the outline information of the construction area within a preset range based on the first roadblock information, the first dynamic object information and the first map element information, through a pre-trained area detection model; the area detection model is a deep neural network model.

[0176] The fourth processing module 54 is used to determine the construction area within a preset range based on the construction area outline information.

[0177] In some alternative embodiments, in the above... Figure 12 Based on the illustrated embodiment, the first processing module 51 can specifically be used to: acquire sensor data collected by sensors on the vehicle; and determine first roadblock information and first dynamic object information based on the sensor data.

[0178] In some optional embodiments, the second processing module 52 may specifically be used to: acquire the vehicle's pose information in the map coordinate system; and, based on the pose information, determine first map element information within a preset range from pre-configured map data.

[0179] In some optional embodiments, the first roadblock information includes at least one of the following: position information, size information, orientation information, and type information of a roadblock of a first preset type within a preset range in the vehicle coordinate system.

[0180] In some optional embodiments, the first dynamic object information includes trajectory information of a second preset type of dynamic object within a preset range in the vehicle coordinate system.

[0181] In some optional embodiments, the first map element information includes descriptive information of each map element within a preset range in the vehicle coordinate system.

[0182] In some optional embodiments, based on any of the above embodiments, the third processing module 53 may specifically be used to: determine input feature data based on the first roadblock information, the first dynamic object information, and the first map element information; perform detection processing on the input feature data based on a region detection model to obtain the contour point set corresponding to the construction area; and determine the contour information of the construction area based on the first roadblock information and the contour point set corresponding to the construction area.

[0183] In some optional embodiments, the third processing module 53 may specifically be used for:

[0184] For each construction area, based on the contour point set corresponding to the construction area and the first roadblock information, the roadblock information corresponding to the target roadblock in the construction area is determined; based on the roadblock information corresponding to the target roadblock in the construction area, the contour point set corresponding to the construction area is adjusted to obtain the contour information of the construction area.

[0185] In some optional embodiments, the third processing module 53 may be specifically used to: perform detection processing on the input feature data based on the region detection model to obtain the candidate contour point set corresponding to the candidate construction area and the confidence level corresponding to each candidate contour point set; and determine the contour point set corresponding to the construction area from the candidate contour point set corresponding to the candidate construction area based on the confidence level corresponding to each candidate contour point set.

[0186] In some optional embodiments, based on any of the above embodiments, the region detection model is obtained in the following manner: acquiring training sample data; the training sample data includes at least one training sample and a construction area outline label corresponding to each training sample; each training sample includes second roadblock information, second dynamic object information and second map element information corresponding to a vehicle at a certain time point; and training the initial region detection model based on the training sample data to obtain the region detection model.

[0187] In some optional embodiments, the third processing module 53 may be specifically used to: extract features from the first roadblock information to obtain first roadblock features; extract features from the first dynamic object information to obtain first dynamic object features; extract features from the first map element information to obtain first map element features; determine fusion features based on the first roadblock features, the first dynamic object features, and the first map element features; decode the fusion features to obtain decoding results; and determine the outline information of the construction area based on the decoding results.

[0188] In some optional embodiments, the third processing module 53 may be specifically used to: determine fusion features based on the first roadblock features, the first dynamic object features, and the first map element features, through a first number of encoders; each encoder includes at least one encoding attention layer.

[0189] In some optional embodiments, the third processing module 53 may specifically be used to: decode the fused features through a second number of decoders to obtain a decoding result; each decoder includes at least one decoding attention layer.

[0190] In some optional embodiments, the third processing module 53 may be specifically used to: determine the candidate contour point set corresponding to the candidate construction area through the regression head network layer based on the decoding result; determine the confidence level corresponding to each candidate contour point set through the classification head network layer based on the decoding result; and determine the contour information of the construction area based on the candidate contour point set and the confidence level corresponding to each candidate contour point set.

[0191] Figure 13 This is a schematic diagram of the structure of a construction area determination device provided in another exemplary embodiment of this disclosure.

[0192] In some alternative embodiments, based on any of the above embodiments, such as Figure 13 As shown, the apparatus of this embodiment may further include an output module 55.

[0193] The output module 55 is used to output a construction area prompt message in response to the existence of at least one construction area within a preset range.

[0194] In some alternative embodiments, based on any of the above embodiments, such as Figure 13 As shown, the apparatus of this embodiment may further include a fifth processing module 61 and a control module 62.

[0195] The fifth processing module 61 is used to plan driving routes based on the construction area within a preset range.

[0196] Control module 62 is used to control the vehicle's operating status based on the driving route.

[0197] The embodiments described above can be implemented individually or in any combination without conflict. The specific implementation can be set according to actual needs, and this disclosure does not limit them.

[0198] The beneficial technical effects corresponding to the exemplary embodiments of this device can be found in the corresponding beneficial technical effects of the exemplary method section above, and will not be repeated here.

[0199] Exemplary electronic devices

[0200] Figure 14 This is a structural diagram of an electronic device provided in an embodiment of the present disclosure, including at least one processor 91 and a memory 92.

[0201] The processor 91 may be a central processing unit (CPU) or other form of processing unit with data processing and / or instruction execution capabilities, and may control other components in the electronic device 90 to perform desired functions.

[0202] The memory 92 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 91 may execute one or more computer program instructions to implement the methods and / or other desired functions of the various embodiments of this disclosure described above.

[0203] In one example, the electronic device 90 may also include an input device 93 and an output device 94, which are interconnected via a bus system and / or other forms of connection mechanism (not shown).

[0204] The input device 93 may also include, for example, a keyboard, mouse, touchscreen, microphone, various sensors, etc. Sensors may include, for example, image sensors (e.g., cameras, webcams), LiDAR, millimeter-wave radar, ultrasonic radar, positioning sensors, pressure sensors, air quality sensors, temperature sensors, etc. Image sensors, LiDAR, millimeter-wave radar, ultrasonic radar, etc., can be used for environmental perception, i.e., detecting moving and static objects in the surrounding environment. Moving and static objects may include, for example, static objects such as lane lines, curbs, arrows, signs, trees, and buildings, as well as dynamic objects such as surrounding vehicles, pedestrians, and cyclists. Positioning sensors are used to locate the mobile device (e.g., a bicycle, a robot, etc.) where the electronic device is located. Positioning sensors may include, for example, an Inertial Measurement Unit (IMU) and a Global Positioning System (GPS). Pressure sensors can be used to detect seat pressure. Temperature sensors can be used to detect the temperature inside the vehicle cabin. Air quality sensors can be used to detect the air quality inside the vehicle cabin.

[0205] The output device 94 can output various information to the outside, including, for example, a display, a speaker, a communication network and its connected remote output devices, etc.

[0206] Of course, for the sake of simplicity, Figure 14 Only some of the components of the electronic device 90 relevant to this disclosure are shown, omitting components such as buses, input / output interfaces, etc. In addition, the electronic device 90 may include any other suitable components depending on the specific application.

[0207] Exemplary computer program products and computer-readable storage media

[0208] In addition to the methods and apparatus described above, embodiments of this disclosure may also provide a computer program product, including computer program instructions that, when executed by a processor, cause the processor to perform the steps of the methods in the various embodiments of this disclosure described in the "Exemplary Methods" section above.

[0209] Computer program products can be written in any combination of one or more programming languages ​​to perform the operations of embodiments of this disclosure. These programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on a user's computing device, partially on a user's computing device, as a standalone software package, partially on a user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0210] Furthermore, embodiments of this disclosure may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform the steps of the methods in the various embodiments of this disclosure described in the "Exemplary Methods" section above.

[0211] Computer-readable storage media may take the form of any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may include, but is not limited to, systems, apparatuses, or devices that are electrical, magnetic, optical, electromagnetic, infrared, or semiconductor, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0212] The basic principles of this disclosure have been described above with reference to specific embodiments. However, the advantages, benefits, and effects mentioned in this disclosure are merely examples and not limitations, and should not be considered as essential features of each embodiment of this disclosure. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the scope of this disclosure to the necessity of employing the aforementioned specific details for implementation.

[0213] Various modifications and variations can be made to this disclosure without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this disclosure and their equivalents, this disclosure is also intended to include such modifications and variations.

Claims

1. A method for determining a construction area, comprising: Determine the information of the first roadblock and the first dynamic object within a preset range around the vehicle; Determine the information of the first map element within the preset range; Based on the first roadblock information, the first dynamic object information, and the first map element information, the outline information of the construction area within the preset range is determined by a pre-trained area detection model; the area detection model is a deep neural network model. Based on the outline information of the construction area, the construction area within the preset range is determined.

2. The method according to claim 1, wherein, The step of determining the outline information of the construction area within the preset range based on the first roadblock information, the first dynamic object information, and the first map element information, using a pre-trained area detection model, includes: Based on the first roadblock information, the first dynamic object information, and the first map element information, the input feature data is determined; Based on the region detection model, the input feature data is processed to obtain the contour point set corresponding to the construction area; Based on the first roadblock information and the set of contour points corresponding to the construction area, the contour information of the construction area is determined.

3. The method according to claim 2, wherein, The step of determining the contour information of the construction area based on the first roadblock information and the contour point set corresponding to the construction area includes: For each construction area, based on the contour point set corresponding to the construction area and the first roadblock information, the roadblock information corresponding to the target roadblock in the construction area is determined. Based on the roadblock information corresponding to the target roadblock within the construction area, the contour point set corresponding to the construction area is adjusted to obtain the contour information of the construction area.

4. The method according to claim 2, wherein, The process of detecting and processing the input feature data based on the region detection model to obtain the contour point set corresponding to the construction area includes: Based on the region detection model, the input feature data is processed to obtain the candidate contour point set corresponding to the candidate construction area and the confidence level of each candidate contour point set. Based on the confidence level corresponding to each candidate contour point set, the contour point set corresponding to the construction area is determined from the candidate contour point set corresponding to the candidate construction area.

5. The method according to claim 1, wherein, The determination of the first roadblock information and the first dynamic object information within a preset range around the vehicle includes: Acquire sensor data collected by sensors on the vehicle; Based on the sensor data, the information of the first roadblock and the information of the first dynamic object are determined.

6. The method according to claim 1, wherein, Determining the first map element information within the preset range includes: Obtain the vehicle's pose information in the map coordinate system; Based on the pose information, the first map element information within the preset range is determined from the pre-configured map data.

7. The method according to claim 1, wherein, The first roadblock information includes at least one of the following: position information, size information, orientation information, and type information of a roadblock of the first preset type within the preset range in the vehicle coordinate system; and / or, The first dynamic object information includes the trajectory information of a second preset type of dynamic object within the preset range in the vehicle coordinate system; And / or, The first map element information includes the description information of each map element within the preset range in the vehicle coordinate system.

8. The method according to any one of claims 1-7, wherein, The region detection model is obtained in the following way: Acquire training sample data; the training sample data includes at least one training sample and a construction area outline label corresponding to each training sample; each training sample includes second roadblock information, second dynamic object information and second map element information corresponding to a vehicle at a certain time point. Based on the training sample data, the initial region detection model is trained to obtain the region detection model.

9. The method according to any one of claims 1-7, further comprising: In response to the existence of at least one construction area within the preset range, a construction area prompt message is output. And / or, Based on the construction area within the preset range, a driving route is planned; The vehicle's operating status is controlled based on the driving route.

10. The method according to any one of claims 1-7, wherein, The step of determining the outline information of the construction area within the preset range based on the first roadblock information, the first dynamic object information, and the first map element information, using a pre-trained area detection model, includes: Feature extraction is performed on the first roadblock information to obtain the first roadblock features; Feature extraction is performed on the information of the first dynamic object to obtain the features of the first dynamic object; Feature extraction is performed on the first map element information to obtain the first map element features; Based on the first roadblock feature, the first dynamic object feature, and the first map element feature, the fusion feature is determined; The fused features are decoded to obtain the decoding result; Based on the decoding results, the outline information of the construction area is determined.

11. The method according to claim 10, wherein, The step of determining the fusion features based on the first roadblock features, the first dynamic object features, and the first map element features includes: Based on the first roadblock feature, the first dynamic object feature, and the first map element feature, the fusion feature is determined by a first number of encoders; each encoder includes at least one encoding attention layer.

12. The method according to claim 10, wherein, Decoding the fused features to obtain the decoding result includes: The fused features are decoded by a second number of decoders to obtain the decoding result; each decoder includes at least one decoding attention layer.

13. The method according to claim 10, wherein, The step of determining the outline information of the construction area based on the decoding result includes: Based on the decoding results, the candidate contour point set corresponding to the candidate construction area is determined through the regression head network layer; Based on the decoding results, the confidence level corresponding to each candidate contour point set is determined through the classification head network layer; Based on the candidate contour point set and the confidence level corresponding to each candidate contour point set, the contour information of the construction area is determined.

14. A device for determining a construction area, comprising: The first processing module is used to determine the information of the first roadblock and the first dynamic object within a preset range around the vehicle. The second processing module is used to determine the first map element information within the preset range; The third processing module is used to determine the outline information of the construction area within the preset range based on the first roadblock information, the first dynamic object information, and the first map element information, using a pre-trained area detection model; the area detection model is a deep neural network model. The fourth processing module is used to determine the construction area within the preset range based on the construction area outline information.

15. An electronic device, the electronic device comprising: processor; Memory used to store the processor's executable instructions; The processor is configured to read the executable instructions from the memory and execute the instructions to implement the method described in any one of claims 1-13. or, The electronic device includes the apparatus described in claim 14.

16. A computer-readable storage medium storing a computer program for performing the method according to any one of claims 1-13.

17. A computer program product, wherein when instructions in the computer program product are executed by a processor, the method described in any one of claims 1-13 of this disclosure is performed.

Images