Intelligent road crack detection method and system based on AI image recognition

Abstract

The application discloses a kind of road crack intelligent detection method and system based on AI image recognition, which comprises: when target vehicle drives on target lane, the real-time image data of the target lane and the vehicle driving data of the target vehicle are obtained;Based on risk type identification algorithm, at least one driving risk type is identified according to the vehicle driving data;According to image recognition algorithm, and the real-time image data, the crack condition parameters of the target lane are identified;Based on the risk identification model corresponding to the driving risk type, the crack driving risk of the target vehicle is determined according to the crack condition parameters and the vehicle driving data based on the driving risk type corresponding risk identification model.It can be seen that the application can accurately identify the cracks in the road and the driving risks that may be caused based on the vehicle driving risk, improve the safety of intelligent driving, and provide more accurate data basis for auxiliary driving research.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to an intelligent method and system for detecting road cracks based on AI image recognition. Background Technology

[0002] With the development of intelligent driving technology, the application of digital algorithms in modern transportation systems is becoming increasingly widespread. However, the complexity and unpredictability of road environments pose significant challenges to intelligent driving. One crucial technical issue is how to accurately and reasonably identify cracks in roads. Current technologies, however, do not consider combining multiple possible risk types with vehicle driving data for more precise crack risk identification. Therefore, their accuracy and intelligence levels are somewhat lacking. Clearly, existing technologies have shortcomings that urgently need to be addressed. Summary of the Invention

[0003] The technical problem to be solved by this invention is to provide a road crack intelligent detection method and system based on AI image recognition, which can accurately identify cracks in the road and potential driving risks based on vehicle driving risks, improve the safety of intelligent driving, and provide a more accurate data foundation for assisted driving research.

[0004] To address the aforementioned technical problems, the first aspect of this invention discloses a road crack intelligent detection method based on AI image recognition, the method comprising:

[0005] While the target vehicle is traveling in the target lane, real-time image data of the target lane and vehicle driving data of the target vehicle are acquired.

[0006] Based on the risk type identification algorithm, at least one driving risk type is identified according to the vehicle driving data;

[0007] Based on the image recognition algorithm and the real-time image data, the crack condition parameters of the target lane are identified;

[0008] Based on the risk identification model corresponding to the driving risk type, the crack driving risk of the target vehicle is determined according to the crack condition parameters and the vehicle driving data.

[0009] As an optional implementation, in the first aspect of the invention, the driving risk type is rollover risk, skidding risk, bump risk, severe yaw risk, or impact risk; and / or, the crack condition parameters include at least one of crack depth, crack protrusion height, crack length, crack width, and crack location in the road.

[0010] As an optional implementation, in the first aspect of the present invention, the vehicle driving data includes vehicle physical parameters and driving sensing parameters; the vehicle physical parameters include at least one of vehicle volume, vehicle model, vehicle weight, vehicle center of gravity, vehicle length, vehicle width, and vehicle height; the driving sensing parameters include at least one of driving speed, driving acceleration, driving direction, executed driving command, and driving command to be executed.

[0011] As an optional implementation, in the first aspect of the present invention, the step of identifying at least one driving risk type based on the risk type identification algorithm according to the vehicle driving data includes:

[0012] The vehicle driving data is input into a trained multi-classification algorithm model to obtain multiple predicted risk types and corresponding first predicted probabilities corresponding to the vehicle driving data; the multi-classification algorithm model is trained using a training dataset that includes multiple training vehicle driving data and corresponding possible driving risk labels.

[0013] Based on the multiple predicted risk types and their corresponding first predicted probabilities, at least one driving risk type is selected.

[0014] As an optional implementation, in the first aspect of the present invention, the step of selecting at least one driving risk type based on the plurality of predicted risk types and the corresponding first predicted probabilities includes:

[0015] For any two of the predicted risk types, obtain the historical accident records corresponding to the two predicted risk types respectively;

[0016] Calculate the ratio of the number of records in all historical accident records that simultaneously exhibit both of the predicted risk types to the total number of records, and obtain the correlation ratio.

[0017] Calculate the correlation weight between the two predicted risk types; the correlation weight is proportional to the correlation ratio.

[0018] The correlation parameter between the two predicted risk types is obtained by multiplying the average of the first predicted probabilities of the two predicted risk types with the correlation weight.

[0019] Based on the correlation parameters, all predicted risk types are clustered to obtain a type cluster set;

[0020] All predicted risk types in the cluster set are identified as driving risk types.

[0021] As an optional implementation, in the first aspect of the invention, the step of clustering all the predicted risk types based on the correlation parameter to obtain a type cluster set includes:

[0022] The objective function is determined to minimize the total number of all said types of cluster sets and minimize the number of said predicted risk types in each said type of cluster set;

[0023] The constraints include:

[0024] The correlation parameter between any two predicted risk types in each of the aforementioned cluster sets is greater than the first parameter threshold;

[0025] The correlation parameter between any two predicted risk types belonging to different cluster sets is less than the second parameter threshold; the second parameter threshold is less than the first parameter threshold.

[0026] Based on the objective function and the constraints, the dynamic programming algorithm is used to iteratively cluster and group all the predicted risk types to obtain a type cluster set.

[0027] As an optional implementation, in the first aspect of the invention, identifying the crack condition parameters of the target lane based on the image recognition algorithm and the real-time image data includes:

[0028] The real-time image data is input into the trained crack recognition model to obtain the output crack region, crack parameters, and second predicted probability; the crack recognition model is trained using a training dataset that includes multiple training image data and corresponding crack region and crack parameter annotations.

[0029] Determine whether the second predicted probability is within a preset suspicious probability range to obtain a first determination result;

[0030] When the first judgment result is yes, the real-time control image acquisition device acquires images of adjacent lanes corresponding to the target lane based on different angles;

[0031] After stitching the crack region with the adjacent lane image, the stitched image is input into the crack recognition model to obtain the third prediction probability;

[0032] Determine whether the third predicted probability falls within the suspicious probability interval to obtain a second determination result;

[0033] If the second judgment result is yes, the crack condition parameter of the target lane is determined to be a suspicious parameter;

[0034] If the first judgment result or the second judgment result is negative, determine whether the second predicted probability is greater than a preset probability threshold to obtain a fourth judgment result; the probability threshold is greater than the upper limit of the suspicious probability interval.

[0035] When the fourth determination result is yes, the crack parameter is determined as the crack condition parameter of the target lane.

[0036] As an optional implementation, in the first aspect of the present invention, determining the crack driving risk of the target vehicle based on the risk identification model corresponding to the driving risk type, according to the crack condition parameters and the vehicle driving data, includes:

[0037] For each type of driving risk, a risk identification model corresponding to that type of driving risk is determined in a preset model library; the risk identification model is trained using a training dataset that includes multiple training crack condition parameters and corresponding vehicle driving data annotations and risk value annotations.

[0038] The crack condition parameters and the vehicle driving data are input into the risk identification model to obtain the type risk value corresponding to the driving risk type of the target vehicle.

[0039] For any two driving risk types, based on preset risk association rules, it is determined whether the two driving risk types belong to enhanced associated risks. If so, the type risk value corresponding to the two driving risk types is determined as the product of the original type risk value and a preset weight; the preset weight is greater than 1; the set of risk types that belong to enhanced associated risks between each other as defined by the risk association rules includes: {rollover risk, severe yaw risk, turbulence risk}, {skid risk, severe yaw risk} and {rollover risk, collision risk, turbulence risk};

[0040] All driving risk types whose risk values ​​are greater than a preset risk value threshold are identified as crack driving risks for the target vehicle.

[0041] A second aspect of this invention discloses an intelligent road crack detection system based on AI image recognition, the system comprising:

[0042] The acquisition module is used to acquire real-time image data of the target lane and vehicle driving data of the target vehicle when the target vehicle is driving in the target lane.

[0043] The first identification module is used to identify at least one type of driving risk based on the vehicle driving data using a risk type identification algorithm.

[0044] The second identification module is used to identify the crack condition parameters of the target lane based on the image recognition algorithm and the real-time image data.

[0045] The determination module is used to determine the crack driving risk of the target vehicle based on the risk identification model corresponding to the driving risk type, according to the crack condition parameters and the vehicle driving data.

[0046] As an optional implementation, in a second aspect of the invention, the driving risk type is rollover risk, skidding risk, turbulence risk, severe yaw risk, or impact risk; and / or, the crack condition parameters include at least one of crack depth, crack protrusion height, crack length, crack width, and crack location in the road.

[0047] As an optional implementation, in a second aspect of the present invention, the vehicle driving data includes vehicle physical parameters and driving sensing parameters; the vehicle physical parameters include at least one of vehicle volume, vehicle model, vehicle weight, vehicle center of gravity, vehicle length, vehicle width, and vehicle height; the driving sensing parameters include at least one of driving speed, driving acceleration, driving direction, executed driving command, and driving command to be executed.

[0048] As an optional implementation, in a second aspect of the invention, the first identification module identifies at least one type of driving risk based on a risk type identification algorithm and the vehicle driving data, including:

[0049] The vehicle driving data is input into a trained multi-classification algorithm model to obtain multiple predicted risk types and corresponding first predicted probabilities corresponding to the vehicle driving data; the multi-classification algorithm model is trained using a training dataset that includes multiple training vehicle driving data and corresponding possible driving risk labels.

[0050] Based on the multiple predicted risk types and their corresponding first predicted probabilities, at least one driving risk type is selected.

[0051] As an optional implementation, in a second aspect of the invention, the specific method by which the first identification module filters out at least one driving risk type based on the plurality of predicted risk types and their corresponding first predicted probabilities includes:

[0052] For any two of the predicted risk types, obtain the historical accident records corresponding to the two predicted risk types respectively;

[0053] Calculate the ratio of the number of records in all historical accident records that simultaneously exhibit both of the predicted risk types to the total number of records, and obtain the correlation ratio.

[0054] Calculate the correlation weight between the two predicted risk types; the correlation weight is proportional to the correlation ratio.

[0055] The correlation parameter between the two predicted risk types is obtained by multiplying the average of the first predicted probabilities of the two predicted risk types with the correlation weight.

[0056] Based on the correlation parameters, all predicted risk types are clustered to obtain a type cluster set;

[0057] All predicted risk types in the cluster set are identified as driving risk types.

[0058] As an optional implementation, in a second aspect of the invention, the specific method by which the first identification module clusters all the predicted risk types based on the correlation parameter to obtain a type cluster set includes:

[0059] The objective function is determined to minimize the total number of all said types of cluster sets and minimize the number of said predicted risk types in each said type of cluster set;

[0060] The constraints include:

[0061] The correlation parameter between any two predicted risk types in each of the aforementioned cluster sets is greater than the first parameter threshold;

[0062] The correlation parameter between any two predicted risk types belonging to different cluster sets is less than the second parameter threshold; the second parameter threshold is less than the first parameter threshold.

[0063] Based on the objective function and the constraints, the dynamic programming algorithm is used to iteratively cluster and group all the predicted risk types to obtain a type cluster set.

[0064] As an optional implementation, in a second aspect of the invention, the specific method by which the second identification module identifies the crack condition parameters of the target lane based on an image recognition algorithm and the real-time image data includes:

[0065] The real-time image data is input into the trained crack recognition model to obtain the output crack region, crack parameters, and second predicted probability; the crack recognition model is trained using a training dataset that includes multiple training image data and corresponding crack region and crack parameter annotations.

[0066] Determine whether the second predicted probability is within a preset suspicious probability range to obtain a first judgment result;

[0067] When the first judgment result is yes, the real-time control image acquisition device acquires images of adjacent lanes corresponding to the target lane based on different angles;

[0068] After stitching the crack region with the adjacent lane image, the stitched image is input into the crack recognition model to obtain the third prediction probability;

[0069] Determine whether the third predicted probability falls within the suspicious probability interval to obtain a second determination result;

[0070] If the second judgment result is yes, the crack condition parameter of the target lane is determined to be a suspicious parameter;

[0071] If the first judgment result or the second judgment result is negative, determine whether the second predicted probability is greater than a preset probability threshold to obtain a fourth judgment result; the probability threshold is greater than the upper limit of the suspicious probability interval.

[0072] When the fourth determination result is yes, the crack parameter is determined as the crack condition parameter of the target lane.

[0073] As an optional implementation, in a second aspect of the invention, the determining module determines the specific method by which it determines the crack driving risk of the target vehicle based on the risk identification model corresponding to the driving risk type, according to the crack condition parameters and the vehicle driving data, including:

[0074] For each type of driving risk, a risk identification model corresponding to that type of driving risk is determined in a preset model library; the risk identification model is trained using a training dataset that includes multiple training crack condition parameters and corresponding vehicle driving data annotations and risk value annotations.

[0075] The crack condition parameters and the vehicle driving data are input into the risk identification model to obtain the type risk value corresponding to the driving risk type of the target vehicle.

[0076] For any two driving risk types, based on preset risk association rules, it is determined whether the two driving risk types belong to enhanced associated risks. If so, the type risk value corresponding to the two driving risk types is determined as the product of the original type risk value and a preset weight; the preset weight is greater than 1; the set of risk types that belong to enhanced associated risks between each other as defined by the risk association rules includes: {rollover risk, severe yaw risk, turbulence risk}, {skid risk, severe yaw risk} and {rollover risk, collision risk, turbulence risk};

[0077] All driving risk types whose risk values ​​are greater than a preset risk value threshold are identified as crack driving risks for the target vehicle.

[0078] A third aspect of this invention discloses another intelligent road crack detection system based on AI image recognition, the system comprising:

[0079] Memory containing executable program code;

[0080] A processor coupled to the memory;

[0081] The processor calls the executable program code stored in the memory to execute some or all of the steps in the intelligent road crack detection method based on AI image recognition disclosed in the first aspect of the present invention.

[0082] The fourth aspect of the present invention discloses a computer storage medium storing computer instructions, which, when invoked, are used to execute some or all of the steps in the intelligent road crack detection method based on AI image recognition disclosed in the first aspect of the present invention.

[0083] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0084] This invention can identify at least one possible driving risk type based on a risk type identification algorithm and vehicle driving data. Then, based on an image crack identification algorithm, it identifies the crack condition parameters of the target lane from real-time image data. Finally, it accurately determines the crack driving risk of the target vehicle based on the crack condition parameters and vehicle driving data through a risk identification model corresponding to the driving risk type. This enables the accurate identification of cracks in the road and the possible driving risks based on vehicle driving risks, improving the safety of intelligent driving and providing a more accurate data foundation for assisted driving research. Attached Figure Description

[0085] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0086] Figure 1 This is a flowchart illustrating an intelligent road crack detection method based on AI image recognition disclosed in an embodiment of the present invention.

[0087] Figure 2 This is a schematic diagram of the structure of an intelligent road crack detection system based on AI image recognition disclosed in an embodiment of the present invention.

[0088] Figure 3 This is a schematic diagram of another intelligent road crack detection system based on AI image recognition disclosed in an embodiment of the present invention. Detailed Implementation

[0089] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0090] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0091] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0092] This invention discloses an intelligent road crack detection method and system based on AI image recognition. It can identify at least one possible driving risk type based on a risk type identification algorithm and vehicle driving data. Then, based on an image crack identification algorithm, it identifies crack condition parameters of the target lane from real-time image data. Finally, through a risk identification model corresponding to the driving risk type, it accurately determines the crack driving risk of the target vehicle based on the crack condition parameters and vehicle driving data. This enables accurate identification of road cracks and potential driving risks based on vehicle driving risk, improving the safety of intelligent driving and providing a more accurate data foundation for assisted driving research. Detailed explanations follow.

[0093] Example 1

[0094] Please see Figure 1 , Figure 1This is a flowchart illustrating an intelligent road crack detection method based on AI image recognition disclosed in an embodiment of the present invention. Figure 1 The described AI-based image recognition-based intelligent road crack detection method can be applied to data processing systems / data processing equipment / data processing servers (including local processing servers or cloud processing servers). For example... Figure 1 As shown, the intelligent road crack detection method based on AI image recognition can include the following operations:

[0095] 101. When the target vehicle is driving in the target lane, acquire real-time image data of the target lane and vehicle driving data of the target vehicle.

[0096] 102. Based on the risk type identification algorithm, at least one driving risk type is identified according to vehicle driving data.

[0097] 103. Based on the image recognition algorithm and real-time image data, identify the crack condition parameters of the target lane.

[0098] 104. Based on the risk identification model corresponding to the driving risk type, determine the crack driving risk of the target vehicle according to the crack condition parameters and vehicle driving data.

[0099] As can be seen, the above-described embodiments of the invention can identify at least one possible driving risk type based on the risk type identification algorithm and vehicle driving data. Then, based on the image crack identification algorithm, the crack condition parameters of the target lane are identified from the real-time image data. Finally, the crack driving risk of the target vehicle is accurately determined by the risk identification model corresponding to the driving risk type, based on the crack condition parameters and vehicle driving data. This enables the accurate identification of cracks in the road and the possible driving risks based on vehicle driving risks, improving the safety of intelligent driving and providing a more accurate data foundation for assisted driving research.

[0100] As an optional embodiment, in the above steps, the driving risk type is rollover risk, skidding risk, bump risk, severe yaw risk, or collision risk; and / or, the crack condition parameters include at least one of crack depth, crack protrusion height, crack length, crack width, and crack location in the road.

[0101] As can be seen, the above optional embodiments define the types of driving risks and the content of crack condition parameters to accurately characterize the risk situation, facilitate subsequent risk identification, and thus help to accurately identify cracks in the road and potential driving risks based on vehicle driving risks, improve the safety of intelligent driving, and provide a more accurate data foundation for assisted driving research.

[0102] As an optional embodiment, the vehicle driving data in the above steps includes vehicle physical parameters and driving sensing parameters; the vehicle physical parameters include at least one of vehicle volume, vehicle model, vehicle weight, vehicle center of gravity, vehicle length, vehicle width, and vehicle height; the driving sensing parameters include at least one of driving speed, driving acceleration, driving direction, executed driving command, and driving command to be executed.

[0103] As can be seen, the above optional embodiments define the data content of vehicle driving data to accurately characterize the vehicle's driving situation, which facilitates subsequent risk identification. This helps to accurately identify cracks in the road and potential driving risks based on vehicle driving risks, improves the safety of intelligent driving, and provides a more accurate data foundation for assisted driving research.

[0104] As an optional embodiment, in the above steps, based on the risk type identification algorithm, at least one driving risk type is identified according to vehicle driving data, including:

[0105] Vehicle driving data is input into a trained multi-classification algorithm model to obtain multiple predicted risk types and corresponding first predicted probabilities for the vehicle driving data; optionally, the multi-classification algorithm model is trained using a training dataset that includes multiple training vehicle driving data and corresponding possible driving risk labels.

[0106] Based on multiple predicted risk types and their corresponding first predicted probabilities, at least one driving risk type is selected.

[0107] As can be seen, through the above optional embodiments, the trained multi-classification algorithm model can accurately pre-identify vehicle driving data to obtain multiple possible risk types and predicted probabilities, thereby filtering out driving risk types and facilitating subsequent risk identification. This helps to accurately identify cracks in the road and potential driving risks based on vehicle driving risks, improves the safety of intelligent driving, and provides a more accurate data foundation for assisted driving research.

[0108] As an optional embodiment, the step described above, selecting at least one driving risk type based on multiple predicted risk types and their corresponding first predicted probabilities, includes:

[0109] For any two predicted risk types, obtain the historical accident records corresponding to the two predicted risk types respectively;

[0110] Calculate the ratio of the number of records in all historical accident records that simultaneously exhibit both of the predicted risk types to the total number of records, and obtain the correlation ratio.

[0111] Calculate the correlation weight between the two predicted risk types; optionally, the correlation weight is proportional to the correlation ratio.

[0112] The correlation parameter between the two predicted risk types is obtained by multiplying the average of the first predicted probabilities of the two predicted risk types with the correlation weight.

[0113] Based on the correlation parameters, clustering is performed on all predicted risk types to obtain a type cluster set;

[0114] All predicted risk types in the type cluster set are identified as driving risk types.

[0115] As can be seen, through the above optional embodiments, the correlation weight between predicted risk types can be determined by the proportion of simultaneous occurrence of historical accident records between risk types, so as to accurately calculate the correlation parameters between predicted risk types, and based on this, clustering can be used to determine more related and high-probability driving risk types, which facilitates subsequent risk identification. This helps to accurately identify cracks in the road and potential driving risks based on vehicle driving risks, improves the safety of intelligent driving, and provides a more accurate data foundation for assisted driving research.

[0116] As an optional embodiment, the step above, clustering all predicted risk types based on correlation parameters to obtain a type cluster set, includes:

[0117] The objective function is to minimize the total number of cluster sets of all types and the number of predicted risk types in each cluster set;

[0118] The constraints include:

[0119] The correlation parameter between any two predicted risk types in each cluster set is greater than the first parameter threshold;

[0120] The correlation parameter between any two predicted risk types belonging to different cluster sets is less than the second parameter threshold; optionally, the second parameter threshold is less than the first parameter threshold.

[0121] Based on the objective function and constraints, the dynamic programming algorithm is used to iteratively cluster and group all predicted risk types to obtain a set of type clusters.

[0122] As can be seen, through the above optional embodiments, based on the preset objective function and constraints, the predicted risk types can be clustered according to the dynamic programming algorithm to determine a more related and high-probability set of driving risk types, which facilitates subsequent risk identification. This helps to accurately identify cracks in the road and potential driving risks based on vehicle driving risks, improves the safety of intelligent driving, and provides a more accurate data foundation for assisted driving research.

[0123] As an optional embodiment, the step described above, identifying the crack condition parameters of the target lane based on the image recognition algorithm and real-time image data, includes:

[0124] Real-time image data is input into a trained crack recognition model to obtain the output crack region, crack parameters, and second predicted probability; optionally, the crack recognition model is trained using a training dataset that includes multiple training image data and corresponding crack region and crack parameter annotations.

[0125] Determine whether the second predicted probability falls within the preset suspicious probability range to obtain the first judgment result;

[0126] When the first judgment result is yes, the real-time control image acquisition device acquires images of adjacent lanes corresponding to the target lane based on different angles;

[0127] After stitching the crack area with the adjacent lane image, the stitched image is input into the crack recognition model to obtain the third prediction probability.

[0128] Determine whether the third predicted probability falls within the suspicious probability interval to obtain the second judgment result;

[0129] If the second judgment result is yes, the crack condition parameter of the target lane is determined to be a suspicious parameter;

[0130] If the first or second judgment result is negative, determine whether the second predicted probability is greater than the preset probability threshold to obtain the fourth judgment result; optionally, the probability threshold is greater than the upper limit of the suspicious probability interval.

[0131] If the fourth judgment result is yes, the crack parameter is determined as the crack condition parameter of the target lane.

[0132] In some specific implementations, the adjacent lane can be a lane that is obstructed by an obstacle or is separated from the target lane by a distance greater than a preset distance threshold. For example, the far end of the target lane extending into the distance at a curve, or a distant oncoming lane separated from the target lane by a fence. The purpose of setting this feature is to enable the image acquisition device to acquire images of similar lanes that are as far away as possible for the judgment of suspicious identification.

[0133] As can be seen, through the above optional embodiments, real-time image data can be identified based on the crack identification model, and when the probability is questionable, images of adjacent lanes can be introduced for further identification and determination. Here, images of adjacent lanes from different angles are used to avoid cracks being misjudged as normal due to factors such as shadows, road color, or specific road materials. Adjacent lanes, such as the opposite lane, have similar road conditions to the target lane, and generally do not have similar cracks at different angles. Therefore, using them as the basis for stitching images and re-identifying effectively confirms whether the identification results are feasible, thereby determining more accurate crack condition parameters, which facilitates subsequent risk identification. This helps to accurately identify cracks in the road and potential driving risks based on vehicle driving risks, improves the safety of intelligent driving, and provides a more accurate data foundation for assisted driving research.

[0134] As an optional embodiment, the step above, determining the crack driving risk of the target vehicle based on the risk identification model corresponding to the driving risk type, according to the crack condition parameters and vehicle driving data, includes:

[0135] For each type of driving risk, a risk identification model corresponding to that type of driving risk is determined from a pre-defined model library; optionally, the risk identification model is trained using a training dataset that includes multiple training crack condition parameters and corresponding vehicle driving data annotations and risk value annotations.

[0136] The crack condition parameters and vehicle driving data are input into the risk identification model to obtain the type risk value corresponding to the driving risk type of the target vehicle.

[0137] For any two driving risk types, it is determined whether the two driving risk types belong to enhanced associated risks based on the preset risk association rules. If so, the type risk value corresponding to the two driving risk types is determined as the product of the original type risk value and the preset weight. Optionally, the preset weight is greater than 1. The set of risk types that belong to enhanced associated risks between each other as defined by the risk association rules includes: {rollover risk, severe yaw risk, turbulence risk}, {skid risk, severe yaw risk} and {rollover risk, collision risk, turbulence risk}.

[0138] All driving risk types with a risk value greater than a preset risk value threshold are identified as crack driving risks for the target vehicle.

[0139] As can be seen, through the above optional embodiments, driving risks can be more accurately predicted and correlated based on crack conditions and driving data using a risk identification model. This enables the accurate identification of cracks in the road and potential driving risks based on vehicle driving risks, improving the safety of intelligent driving and providing a more accurate data foundation for assisted driving research.

[0140] Example 2

[0141] Please see Figure 2 , Figure 2 This is a schematic diagram of the structure of an intelligent road crack detection system based on AI image recognition, as disclosed in an embodiment of the present invention. Figure 2 The described AI-based image recognition-based intelligent road crack detection system can be applied to data processing systems / data processing equipment / data processing servers (wherein, the server includes local processing servers or cloud processing servers). For example... Figure 2 As shown, the AI-based image recognition-based intelligent road crack detection system may include:

[0142] The acquisition module 201 is used to acquire real-time image data of the target lane and vehicle driving data of the target vehicle when the target vehicle is driving in the target lane.

[0143] The first identification module 202 is used to identify at least one type of driving risk based on the vehicle driving data using a risk type identification algorithm.

[0144] The second identification module 203 is used to identify the crack condition parameters of the target lane based on the image recognition algorithm and real-time image data.

[0145] The determination module 204 is used to determine the crack driving risk of the target vehicle based on the risk identification model corresponding to the driving risk type, according to the crack condition parameters and vehicle driving data.

[0146] As can be seen, the above-described embodiments of the invention can identify at least one possible driving risk type based on the risk type identification algorithm and vehicle driving data. Then, based on the image crack identification algorithm, the crack condition parameters of the target lane are identified from the real-time image data. Finally, the crack driving risk of the target vehicle is accurately determined by the risk identification model corresponding to the driving risk type, based on the crack condition parameters and vehicle driving data. This enables the accurate identification of cracks in the road and the possible driving risks based on vehicle driving risks, improving the safety of intelligent driving and providing a more accurate data foundation for assisted driving research.

[0147] As an optional embodiment, the driving risk type is rollover risk, skidding risk, bump risk, severe yaw risk, or impact risk; and / or, the crack condition parameters include at least one of crack depth, crack protrusion height, crack length, crack width, and crack location in the road.

[0148] As can be seen, the above optional embodiments define the types of driving risks and the content of crack condition parameters to accurately characterize the risk situation, facilitate subsequent risk identification, and thus help to accurately identify cracks in the road and potential driving risks based on vehicle driving risks, improve the safety of intelligent driving, and provide a more accurate data foundation for assisted driving research.

[0149] As an optional embodiment, the vehicle driving data includes vehicle physical parameters and driving sensing parameters; the vehicle physical parameters include at least one of vehicle volume, vehicle model, vehicle weight, vehicle center of gravity, vehicle length, vehicle width, and vehicle height; the driving sensing parameters include at least one of driving speed, driving acceleration, driving direction, executed driving commands, and driving commands to be executed.

[0150] As can be seen, the above optional embodiments define the data content of vehicle driving data to accurately characterize the vehicle's driving situation, which facilitates subsequent risk identification. This helps to accurately identify cracks in the road and potential driving risks based on vehicle driving risks, improves the safety of intelligent driving, and provides a more accurate data foundation for assisted driving research.

[0151] As an optional embodiment, the first identification module, based on a risk type identification algorithm, identifies at least one specific type of driving risk according to vehicle driving data, including:

[0152] Vehicle driving data is input into a trained multi-classification algorithm model to obtain multiple predicted risk types and corresponding first predicted probabilities for the vehicle driving data; optionally, the multi-classification algorithm model is trained using a training dataset that includes multiple training vehicle driving data and corresponding possible driving risk labels.

[0153] Based on multiple predicted risk types and their corresponding first predicted probabilities, at least one driving risk type is selected.

[0154] As can be seen, through the above optional embodiments, the trained multi-classification algorithm model can accurately pre-identify vehicle driving data to obtain multiple possible risk types and predicted probabilities, thereby filtering out driving risk types and facilitating subsequent risk identification. This helps to accurately identify cracks in the road and potential driving risks based on vehicle driving risks, improves the safety of intelligent driving, and provides a more accurate data foundation for assisted driving research.

[0155] As an optional embodiment, the specific method by which the first identification module filters out at least one driving risk type based on multiple predicted risk types and their corresponding first predicted probabilities includes:

[0156] For any two predicted risk types, obtain the historical accident records corresponding to the two predicted risk types respectively;

[0157] Calculate the ratio of the number of records in all historical accident records that simultaneously exhibit both of the predicted risk types to the total number of records, and obtain the correlation ratio.

[0158] Calculate the correlation weight between the two predicted risk types; optionally, the correlation weight is proportional to the correlation ratio.

[0159] The correlation parameter between the two predicted risk types is obtained by multiplying the average of the first predicted probabilities of the two predicted risk types with the correlation weight.

[0160] Based on the correlation parameters, clustering is performed on all predicted risk types to obtain a type cluster set;

[0161] All predicted risk types in the type cluster set are identified as driving risk types.

[0162] As can be seen, through the above optional embodiments, the correlation weight between predicted risk types can be determined by the proportion of simultaneous occurrence of historical accident records between risk types, so as to accurately calculate the correlation parameters between predicted risk types, and based on this, clustering can be used to determine more related and high-probability driving risk types, which facilitates subsequent risk identification. This helps to accurately identify cracks in the road and potential driving risks based on vehicle driving risks, improves the safety of intelligent driving, and provides a more accurate data foundation for assisted driving research.

[0163] As an optional embodiment, the specific method by which the first identification module clusters all predicted risk types based on correlation parameters to obtain a type cluster set includes:

[0164] The objective function is to minimize the total number of cluster sets of all types and the number of predicted risk types in each cluster set;

[0165] The constraints include:

[0166] The correlation parameter between any two predicted risk types in each cluster set is greater than the first parameter threshold;

[0167] The correlation parameter between any two predicted risk types belonging to different cluster sets is less than the second parameter threshold; optionally, the second parameter threshold is less than the first parameter threshold.

[0168] Based on the objective function and constraints, the dynamic programming algorithm is used to iteratively cluster and group all predicted risk types to obtain a set of type clusters.

[0169] As can be seen, through the above optional embodiments, based on the preset objective function and constraints, the predicted risk types can be clustered according to the dynamic programming algorithm to determine a more related and high-probability set of driving risk types, which facilitates subsequent risk identification. This helps to accurately identify cracks in the road and potential driving risks based on vehicle driving risks, improves the safety of intelligent driving, and provides a more accurate data foundation for assisted driving research.

[0170] As an optional embodiment, the second recognition module identifies the specific method by which it determines the crack condition parameters of the target lane based on an image recognition algorithm and real-time image data, including:

[0171] Real-time image data is input into a trained crack recognition model to obtain the output crack region, crack parameters, and second predicted probability; optionally, the crack recognition model is trained using a training dataset that includes multiple training image data and corresponding crack region and crack parameter annotations.

[0172] Determine whether the second predicted probability falls within the preset suspicious probability range to obtain the first judgment result;

[0173] When the first judgment result is yes, the real-time control image acquisition device acquires images of adjacent lanes corresponding to the target lane based on different angles;

[0174] After stitching the crack area with the adjacent lane image, the stitched image is input into the crack recognition model to obtain the third prediction probability.

[0175] Determine whether the third predicted probability falls within the suspicious probability interval to obtain the second judgment result;

[0176] If the second judgment result is yes, the crack condition parameter of the target lane is determined to be a suspicious parameter;

[0177] If the first or second judgment result is negative, determine whether the second predicted probability is greater than the preset probability threshold to obtain the fourth judgment result; optionally, the probability threshold is greater than the upper limit of the suspicious probability interval.

[0178] If the fourth judgment result is yes, the crack parameter is determined as the crack condition parameter of the target lane.

[0179] As can be seen, through the above optional embodiments, real-time image data can be identified based on the crack identification model, and when the probability is questionable, images of adjacent lanes can be introduced for further identification and determination. Here, images of adjacent lanes from different angles are used to avoid cracks being misjudged as normal due to factors such as shadows, road color, or specific road materials. Adjacent lanes, such as the opposite lane, have similar road conditions to the target lane, and generally do not have similar cracks at different angles. Therefore, using them as the basis for stitching images and re-identifying effectively confirms whether the identification results are feasible, thereby determining more accurate crack condition parameters, which facilitates subsequent risk identification. This helps to accurately identify cracks in the road and potential driving risks based on vehicle driving risks, improves the safety of intelligent driving, and provides a more accurate data foundation for assisted driving research.

[0180] As an optional embodiment, the determining module, based on a risk identification model corresponding to the driving risk type, determines the specific method of the target vehicle's crack driving risk according to crack condition parameters and vehicle driving data, including:

[0181] For each type of driving risk, a risk identification model corresponding to that type of driving risk is determined from a pre-defined model library; optionally, the risk identification model is trained using a training dataset that includes multiple training crack condition parameters and corresponding vehicle driving data annotations and risk value annotations.

[0182] The crack condition parameters and vehicle driving data are input into the risk identification model to obtain the type risk value corresponding to the driving risk type of the target vehicle.

[0183] For any two driving risk types, it is determined whether the two driving risk types belong to enhanced associated risks based on the preset risk association rules. If so, the type risk value corresponding to the two driving risk types is determined as the product of the original type risk value and the preset weight. Optionally, the preset weight is greater than 1. The set of risk types that belong to enhanced associated risks between each other as defined by the risk association rules includes: {rollover risk, severe yaw risk, turbulence risk}, {skid risk, severe yaw risk} and {rollover risk, collision risk, turbulence risk}.

[0184] All driving risk types with a risk value greater than a preset risk value threshold are identified as crack driving risks for the target vehicle.

[0185] As can be seen, through the above optional embodiments, driving risks can be more accurately predicted and correlated based on crack conditions and driving data using a risk identification model. This enables the accurate identification of cracks in the road and potential driving risks based on vehicle driving risks, improving the safety of intelligent driving and providing a more accurate data foundation for assisted driving research.

[0186] Example 3

[0187] Please see Figure 3 , Figure 3 This is another intelligent road crack detection system based on AI image recognition disclosed in the embodiments of the present invention. Figure 3 The described AI-based image recognition-based intelligent road crack detection system is applied in a data processing system / data processing equipment / data processing server (wherein, the server includes a local processing server or a cloud processing server). For example... Figure 3 As shown, the AI-based image recognition-based intelligent road crack detection system may include:

[0188] Memory 301 storing executable program code;

[0189] Processor 302 coupled to memory 301;

[0190] The processor 302 calls the executable program code stored in the memory 301 to execute the steps of the road crack intelligent detection method based on AI image recognition described in Embodiment 1.

[0191] Example 4

[0192] This invention discloses a computer read storage medium that stores a computer program for electronic data interchange, wherein the computer program causes a computer to execute the steps of the road crack intelligent detection method based on AI image recognition described in Embodiment 1.

[0193] Example 5

[0194] This invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to perform the steps of the road crack intelligent detection method based on AI image recognition described in Embodiment 1.

[0195] The foregoing has described specific embodiments of this specification; other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims may be performed in a different order than those shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily have to follow the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0196] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0197] For ease of description, the above devices are described in terms of function, divided into various units. Of course, in implementing this specification, the functions of each unit can be implemented in one or more software and / or hardware components.

[0198] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, the embodiments of this specification can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the embodiments of this specification can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0199] This specification is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0200] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0201] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0202] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0203] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0204] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0205] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0206] This specification can be described in the general context of computer-executable instructions that are executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This specification can also be practiced in distributed computing environments, where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0207] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0208] Finally, it should be noted that the intelligent road crack detection method and system based on AI image recognition disclosed in the embodiments of the present invention are merely preferred embodiments of the present invention, and are only used to illustrate the technical solutions of the present invention, and not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A road crack intelligent detection method based on AI image recognition, characterized in that, The method includes: While the target vehicle is traveling in the target lane, real-time image data of the target lane and vehicle driving data of the target vehicle are acquired. Based on the risk type identification algorithm, at least one driving risk type is identified according to the vehicle driving data; Based on the image recognition algorithm and the real-time image data, the crack condition parameters of the target lane are identified, including: The real-time image data is input into the trained crack recognition model to obtain the output crack region, crack parameters, and second predicted probability; the crack recognition model is trained using a training dataset that includes multiple training image data and corresponding crack region and crack parameter annotations. Determine whether the second predicted probability is within a preset suspicious probability range to obtain a first determination result; When the first judgment result is yes, the real-time control image acquisition device acquires images of adjacent lanes corresponding to the target lane based on different angles; After stitching the crack region with the adjacent lane image, the stitched image is input into the crack recognition model to obtain the third prediction probability; Determine whether the third predicted probability falls within the suspicious probability interval to obtain a second determination result; If the second judgment result is yes, the crack condition parameter of the target lane is determined to be a suspicious parameter; If the first judgment result or the second judgment result is negative, determine whether the second predicted probability is greater than a preset probability threshold to obtain a fourth judgment result; the probability threshold is greater than the upper limit of the suspicious probability interval. When the fourth judgment result is yes, the crack parameter is determined as the crack condition parameter of the target lane; Based on the risk identification model corresponding to the driving risk type, the crack driving risk of the target vehicle is determined according to the crack condition parameters and the vehicle driving data.

2. The intelligent road crack detection method based on AI image recognition according to claim 1, characterized in that, The driving risk type is rollover risk, skidding risk, bump risk, severe yaw risk, or collision risk; and / or, the crack condition parameters include at least one of crack depth, crack protrusion height, crack length, crack width, and crack location in the road.

3. The intelligent road crack detection method based on AI image recognition according to claim 1, characterized in that, The vehicle driving data includes vehicle physical parameters and driving sensor parameters; the vehicle physical parameters include at least one of vehicle volume, vehicle model, vehicle weight, vehicle center of gravity, vehicle length, vehicle width, and vehicle height; the driving sensor parameters include at least one of driving speed, driving acceleration, driving direction, executed driving commands, and driving commands to be executed.

4. The intelligent road crack detection method based on AI image recognition according to claim 1, characterized in that, The risk type identification algorithm identifies at least one driving risk type based on the vehicle driving data, including: The vehicle driving data is input into a trained multi-classification algorithm model to obtain multiple predicted risk types and corresponding first predicted probabilities corresponding to the vehicle driving data; the multi-classification algorithm model is trained using a training dataset that includes multiple training vehicle driving data and corresponding possible driving risk labels. Based on the multiple predicted risk types and their corresponding first predicted probabilities, at least one driving risk type is selected.

5. The intelligent road crack detection method based on AI image recognition according to claim 4, characterized in that, The step of selecting at least one driving risk type based on the multiple predicted risk types and their corresponding first predicted probabilities includes: For any two of the predicted risk types, obtain the historical accident records corresponding to the two predicted risk types respectively; Calculate the ratio of the number of records in all historical accident records that simultaneously exhibit both of the predicted risk types to the total number of records, and obtain the correlation ratio. Calculate the correlation weight between the two predicted risk types; the correlation weight is proportional to the correlation ratio. The correlation parameter between the two predicted risk types is obtained by multiplying the average of the first predicted probabilities of the two predicted risk types with the correlation weight. Based on the correlation parameters, all predicted risk types are clustered to obtain a type cluster set; All predicted risk types in the cluster set are identified as driving risk types.

6. The intelligent road crack detection method based on AI image recognition according to claim 5, characterized in that, The clustering of all predicted risk types based on the correlation parameters to obtain a type cluster set includes: The objective function is determined to minimize the total number of all said types of cluster sets and minimize the number of said predicted risk types in each said type of cluster set; The constraints include: The correlation parameter between any two predicted risk types in each of the aforementioned cluster sets is greater than the first parameter threshold; The correlation parameter between any two predicted risk types belonging to different cluster sets is less than the second parameter threshold; the second parameter threshold is less than the first parameter threshold. Based on the objective function and the constraints, the dynamic programming algorithm is used to iteratively cluster and group all the predicted risk types to obtain a type cluster set.

7. The intelligent road crack detection method based on AI image recognition according to claim 2, characterized in that, The risk identification model based on the driving risk type determines the crack driving risk of the target vehicle according to the crack condition parameters and the vehicle driving data, including: For each type of driving risk, a risk identification model corresponding to that type of driving risk is determined in a preset model library; the risk identification model is trained using a training dataset that includes multiple training crack condition parameters and corresponding vehicle driving data annotations and risk value annotations. The crack condition parameters and the vehicle driving data are input into the risk identification model to obtain the type risk value corresponding to the driving risk type of the target vehicle. For any two driving risk types, based on preset risk association rules, it is determined whether the two driving risk types belong to enhanced associated risks. If so, the type risk value corresponding to the two driving risk types is determined as the product of the original type risk value and a preset weight; the preset weight is greater than 1; the set of risk types that belong to enhanced associated risks between each other as defined by the risk association rules includes: {rollover risk, severe yaw risk, turbulence risk}, {skid risk, severe yaw risk} and {rollover risk, collision risk, turbulence risk}; All driving risk types whose risk values ​​are greater than a preset risk value threshold are identified as crack driving risks for the target vehicle.

8. A road crack intelligent detection system based on AI image recognition, characterized in that, The system executes the intelligent road crack detection method based on AI image recognition as described in any one of claims 1-7, and the system includes: The acquisition module is used to acquire real-time image data of the target lane and vehicle driving data of the target vehicle when the target vehicle is driving in the target lane. The first identification module is used to identify at least one type of driving risk based on the vehicle driving data using a risk type identification algorithm. The second identification module is used to identify the crack condition parameters of the target lane based on the image recognition algorithm and the real-time image data. The determination module is used to determine the crack driving risk of the target vehicle based on the risk identification model corresponding to the driving risk type, according to the crack condition parameters and the vehicle driving data.

9. A road crack intelligent detection system based on AI image recognition, characterized in that, The system includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the intelligent road crack detection method based on AI image recognition as described in any one of claims 1-7.

Images