A method, apparatus and vehicle for predicting pedestrian injury

Abstract

The application discloses a pedestrian injury prediction method and device and a vehicle. The method comprises the following steps: acquiring a current collision speed when a vehicle involved in a traffic accident collides with a pedestrian involved in the traffic accident; identifying a target speed interval in which the current collision speed is located; acquiring a to-be-retrieved item and a target precision rate limit value corresponding to the target speed interval based on a pre-constructed mapping relationship; the to-be-retrieved item is a boundary condition type of accident case retrieval; the target precision rate limit value is an accuracy condition of accident case retrieval; taking the to-be-retrieved item as an accident case retrieval condition, searching for a target accident case satisfying the target precision rate limit value in a pre-constructed historical accident database; and generating an injury prediction result of the pedestrian involved in the traffic accident based on pedestrian injury data in the target accident case. Thus, the pedestrian injury is accurately predicted based on the retrieval item, the precision rate limit value and real pedestrian injury data, support is provided for medical decision-making, and the pedestrian casualty rate is reduced.

Description

Technical Field

[0001] This application relates to the field of traffic safety technology, and in particular to a method, device and vehicle for predicting pedestrian injuries. Background Technology

[0002] In road traffic systems, pedestrians face an extremely high risk of serious injury in collisions with vehicles due to the lack of external protection, making them the most vulnerable group in terms of safety. To reduce pedestrian injury rates, rapid and accurate prediction of pedestrian injuries is crucial for improving medical rescue efficiency and seizing the "golden rescue time" within the first hour after an accident.

[0003] Currently, computer simulation models can be used to systematically adjust and combine key boundary conditions (e.g., vehicle type, collision speed, pedestrian posture, etc.) in a small number of real-world cases, thereby generalizing and generating massive amounts of simulated accident data. This allows for the establishment of a mapping model between pedestrian injury severity (e.g., injury level) and various variables such as vehicle parameters and collision conditions, enabling the prediction of pedestrian injuries through this mapping model.

[0004] However, this prediction method, which relies on simulation data, is difficult to fully cover the diversity of real pedestrian groups in terms of body shape, walking posture and physiological characteristics. As a result, the generalization ability and accuracy of the constructed prediction model are low, making it difficult to meet the actual needs of complex pedestrian injury prediction.

[0005] Therefore, how to accurately predict pedestrian injuries and provide reliable support for emergency medical decisions after an accident, thereby reducing the casualty rate of pedestrian traffic accidents, is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] In view of this, one aspect of this application provides a method for predicting pedestrian injuries, the method comprising: Obtain the current collision speed of the vehicle involved in the current traffic accident at the moment of collision with the pedestrian involved; Identify the target speed range in which the current collision speed is located; Based on a pre-built mapping relationship, the search terms and target precision limits corresponding to the target vehicle speed range are obtained; wherein, the search terms are the boundary condition types for accident case retrieval; and the target precision limit is the accuracy condition for accident case retrieval. Using the term to be searched as the accident case search criteria, target accident cases that meet the target precision limit are retrieved from a pre-constructed historical accident database. Based on pedestrian injury data in the target accident case, injury prediction results for the pedestrians involved are generated.

[0007] Optionally, the historical accident database includes at least vehicle data, pedestrian data, and ambulance data; The vehicle data includes at least the collision speed of the vehicle when it collides with the pedestrian and the collision point data of the vehicle when the collision occurs. The pedestrian data includes at least vital sign data, pedestrian injury data, and pedestrian collision point data at the time of the collision; wherein, the pedestrian injury data includes the injury location, injury level, and injury description text; The rescue data includes at least rescue measures, measure scores, and diagnostic results; the rescue measures include at least one of the following: on-site measures, transfer measures, and hospital admission measures; and the rescue measures correspond one-to-one with the measure scores.

[0008] Optionally, the pedestrian data may also include pedestrian gait data and pedestrian velocity values ​​at the time of collision; the pedestrian gait data includes a first value representing the straddle gait and a second value representing the upright gait. The pedestrian collision point data is the clock data corresponding to the pedestrian's direction of travel in a specified clock direction; The vehicle collision point data is the clock data corresponding to the vehicle's direction of travel in the specified clock direction.

[0009] Optionally, the mapping relationship includes the correspondence between vehicle speed range, search item, and precision limit; The search terms include at least collision speed, pedestrian collision point, vehicle collision point, and pedestrian physical characteristics; The maximum vehicle speed in the speed range is positively correlated with the number of types of the search terms, and the maximum vehicle speed is negatively correlated with the precision limit.

[0010] Optionally, constructing the historical accident database includes the following steps: Collect historical accident data from historical accident cases; Based on the collision speed in the historical accident data and the preset speed range division rules, the historical accident cases are classified. Based on the classification results and the mapping relationship, cases in the historical accident cases that include the search item data corresponding to the vehicle speed range are selected as candidate cases. The candidate cases are listed as cases to be entered into the database and then aggregated to form the historical accident database.

[0011] Optionally, the method for predicting pedestrian injuries further includes: In the candidate cases, when the historical accident data corresponding to any search item comes from multiple acquisition methods, the difference between the data obtained by each acquisition method is compared. If all the differences are less than the corresponding preset threshold, the candidate cases are used as the cases to be entered; and the cases to be entered are collected to construct the historical accident database.

[0012] Optionally, the historical accident database is constructed by collecting the cases to be entered, including: Configure a unique identifier for each of the cases to be entered; Extract pedestrian injury data and first aid data from the historical accident data of the cases to be entered; The pedestrian injury data is assessed using a specified large model to obtain the injury level, and the effectiveness of the rescue measures in the rescue data is scored to obtain the measure score. Based on the unique identifier, the historical accident data of the case to be entered, the damage level, and the measure score are associated and stored to form the historical accident database.

[0013] Optionally, the vehicle data may also include vehicle size data, and the method may further include: Obtain the target vehicle size data of the vehicle involved in the incident; From the historical accident database, cases where the absolute value of the difference between the size data of the target vehicle and the size data of the target vehicle for each size type is less than a preset value are selected as matching cases for the vehicle involved in the accident.

[0014] Optionally, the step of using the term to be searched as the accident case search condition to search for target accident cases in a pre-constructed historical accident database that meets the target precision limit includes: Collect the current accident data corresponding to the item to be searched in the current traffic accident; In the matching cases, the retrieved vehicle speed is within the target vehicle speed range, and the target historical accident data corresponding to the item to be retrieved is selected. Determine parameter values ​​used to characterize the degree of difference between the current accident data and the target historical accident data; Among the matchable cases, the case whose parameter value is not greater than the target precision limit is designated as the target incident case.

[0015] Optionally, the step of designating a specific case among the matchable cases whose parameter value is not greater than the target precision limit as the target incident case includes: If the specified case is unique, the specified case shall be used as the target accident case; If the specified case is not unique, determine whether there is a case in the same area that belongs to the same preset division area as the current traffic accident; If such a case exists, the case in the same area is filtered step by step according to the preset priority filtering conditions. When only one case remains after any level of filtering, the filtering stops to obtain the target accident case. The preset priority filtering conditions include, in order: the case is unique, the parameter value has a unique minimum value, the damage level has a unique maximum level, and the sum of the measure scores has a unique maximum value. If it does not exist, the specified cases are filtered step by step to obtain the target accident cases.

[0016] Optionally, after generating the injury prediction result for the pedestrian involved based on the pedestrian injury data in the target accident case, the method further includes: Obtain medical information from hospitals within the preset area where the current traffic accident occurred; the medical information includes at least blood bank information, drug information, and medical equipment information. Determine whether there are any target rescue measures in the target accident cases that match the medical information; If present, the target rescue measures will be transmitted to the real-time rescue terminal; If not, the damage prediction results and accident point data are transmitted to the remote guidance and rescue system; the accident point data includes at least the accident time, latitude and longitude data, and weather data.

[0017] Another aspect of this application provides a pedestrian injury prediction device, the device comprising: The collision speed acquisition module is used to acquire the current collision speed of the vehicle involved in the current traffic accident when it collides with the pedestrian involved. The vehicle speed range identification module is used to identify the target vehicle speed range in which the current collision vehicle speed is located; The retrieval condition determination module is used to obtain the search terms and target precision limit corresponding to the target vehicle speed range based on a pre-built mapping relationship; wherein, the search terms are the boundary condition types for accident case retrieval; and the target precision limit is the accuracy condition for accident case retrieval. The target accident case retrieval module is used to retrieve target accident cases that meet the target precision limit from a pre-built historical accident database, using the search term as the accident case retrieval condition. The injury result generation module is used to generate injury prediction results for the pedestrians involved in the accident based on the pedestrian injury data in the target accident case.

[0018] Another aspect of this application provides a vehicle including a memory and a processor, the memory storing a computer program executable on the processor, the processor executing the computer program to implement the steps of the pedestrian injury prediction method.

[0019] The pedestrian injury prediction method, device, and vehicle provided in this application offer the following advantages: They no longer rely on simulation-generalized data or are limited by the singularity of human models. Instead, they are based on a large amount of real pedestrian accident injury data, enabling accurate prediction of pedestrian injuries in current traffic accidents. This ensures that the prediction results are closer to the actual situation, providing reliable injury prediction support for emergency medical decision-making, allowing rescue personnel to take timely and effective measures to reduce pedestrian injury rates. Furthermore, when searching for current accidents, the method utilizes search terms and precision limits based on mapping relationships to adapt to varying traffic accidents, thereby achieving accurate and efficient data retrieval and improving the accuracy of pedestrian injury prediction. Attached Figure Description

[0020] Figure 1 A flowchart illustrating a method for predicting pedestrian injuries provided in an embodiment of this application; Figure 2 A schematic diagram illustrating the principle of a pedestrian injury prediction method provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of a specified clock provided in an embodiment of this application; Figure 4 A schematic diagram illustrating the principle of constructing a historical accident database, provided as an embodiment of this application; Figure 5 A schematic diagram illustrating the principle of constructing a historical accident database, provided as another embodiment of this application; Figure 6 This is a schematic diagram illustrating the principle of vehicle size data provided in an embodiment of this application; Figure 7 A schematic diagram illustrating the principle of a pedestrian injury prediction method provided in another embodiment of this application; Figure 8 A schematic diagram of the structure of a pedestrian injury prediction device provided in an embodiment of this application; Figure 9 This is a schematic diagram of the structure of a vehicle provided in an embodiment of this application.

[0021] The attached diagram is labeled as follows: 80 is the collision speed acquisition module, 81 is the speed range identification module, 82 is the search condition determination module, 83 is the target accident case search module, 84 is the damage result generation module, 90 is the memory, 91 is the processor, 92 is the display screen, 93 is the input / output interface, 94 is the communication interface, 95 is the power supply, 96 is the communication bus, 901 is the computer program, 902 is the operating system, and 903 is the data. Detailed Implementation

[0022] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0023] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0024] Figure 1 This is a flowchart illustrating a method for predicting pedestrian injuries provided in an embodiment of this application, as shown below. Figure 1 As shown, the method includes: S10: Obtain the current collision speed of the vehicle involved in the current traffic accident at the time of the collision with the pedestrian involved; Understandably, in specific traffic accidents involving collisions between vehicles and pedestrians, the vehicle's speed has a crucial impact on the severity of pedestrian injuries. Specifically, the faster the vehicle travels, the higher the likelihood of more severe pedestrian injuries. Therefore, in the process of pedestrian injury prediction, collecting the current collision speed of the vehicle and pedestrian at the moment of impact is one of the essential data points to ensure the accuracy of injury prediction.

[0025] In a specific embodiment, the current collision speed can be obtained by a vehicle speed sensor. The vehicles involved include, but are not limited to, sedans, sport utility vehicles (SUVs), multi-purpose vehicles (MPVs), off-road vehicles, pickup trucks, or other power-driven non-rail-borne vehicles. This application does not limit the types of vehicles involved.

[0026] In one optional embodiment, in order to ensure the accuracy of the current collision speed and thus the accuracy of the prediction of subsequent pedestrian injuries, the current collision speed can be the average speed of the vehicle over a specified period of time before the collision, for example, the average speed of the vehicle over 5 seconds before the collision is used as the current collision speed. This application does not limit this.

[0027] S11: Identify the target speed range where the current collision speed is located; S12: Based on the pre-built mapping relationship, obtain the search terms and target precision limits corresponding to the target vehicle speed range; where the search terms are the boundary condition types for accident case retrieval; and the target precision limits are the accuracy conditions for accident case retrieval. Figure 2 This is a schematic diagram illustrating the principle of a pedestrian injury prediction method provided in an embodiment of this application. After determining the current collision speed in the current traffic accident, as follows... Figure 2 As shown, the target speed range where the current collision speed is located is identified. In an optional embodiment, the vehicle speed can be pre-divided into multiple ranges according to a preset speed range division rule, and different speed ranges can be associated with different search terms and precision limits to construct a mapping relationship.

[0028] Therefore, after identifying the target speed range where the current collision speed is located, the boundary conditions and accuracy conditions for the current accident case retrieval are determined based on the pre-built mapping relationship. That is, the search terms and target precision limits corresponding to the target speed range are determined.

[0029] It is worth noting that different vehicle speeds cause different degrees of injury to pedestrians. In order to ensure the accuracy of injury prediction, in one optional embodiment, a large number of real historical accident cases can be analyzed in advance to divide the collision speed at the time of collision in different historical data cases, thereby obtaining different speed ranges.

[0030] The focus and accuracy requirements differ depending on the collision speed. In other words, the required search terms and precision limits vary depending on the current collision speed.

[0031] It should be noted that the search terms refer to the boundary types when retrieving accident data. For example, when a vehicle and a pedestrian collide, the focus should be on the point of impact on the pedestrian, the pedestrian's height, weight, and other physical characteristics, as well as the pedestrian's walking speed and posture. Therefore, in a specific embodiment, the search terms may include, but are not limited to, pedestrian speed, pedestrian physical characteristics, and the point of impact.

[0032] The precision limit refers to the accuracy condition when retrieving accident data, that is, determining what kind of data meets the expectations. It can be data similarity, keyword matching degree, or error rate between data, etc. This application does not limit this.

[0033] S13: Using the search term as the search criteria for accident cases, retrieve target accident cases that meet the target precision limit from the pre-built historical accident database; Furthermore, such as Figure 2As shown, using the search term as the search condition, a search is performed in the historical accident database. The retrieved target accident cases must meet the target precision limit. In an optional embodiment, after determining the search term, current accident data corresponding to the search term in current traffic accidents is collected. For example, when the search term is pedestrian speed, the current accident data includes pedestrian speed values. It is understood that when performing accident retrieval in a pre-built historical database, the vehicle speeds in historical accidents also need to be within the target speed range and meet the target precision limit. Therefore, it is necessary to collect current accident data related to the search term in current accidents.

[0034] Furthermore, based on the current accident data, target accident cases are retrieved from the pre-built historical accident database whose vehicle speed is within the target speed range and meets the target precision limit. In other words, cases with the same speed range, containing all data corresponding to the items to be retrieved, and whose error between the data corresponding to the items to be retrieved and the current accident data is within the target precision limit are selected as target accident cases.

[0035] It should be noted that, in specific embodiments, the target accident case can be a case whose similarity to the current traffic accident reaches a threshold, or a case whose error between the accident data is within the target precision limit. This application does not limit this.

[0036] S14: Based on pedestrian injury data in the target accident case, generate injury prediction results for the pedestrians involved.

[0037] In fact, the target accident cases obtained after the target precision limit retrieval are highly similar to the current traffic accident. Therefore, injury prediction results for pedestrians involved in the current traffic accident can be generated based on historical accident cases.

[0038] Specifically, the historical accident database includes a large number of historical accident cases, and each accident case stores pedestrian injury data. Therefore, injury prediction results for the current pedestrians involved can be generated based on the historical pedestrian injury data of the target accident case. As an optional implementation, historical pedestrian injury data can be directly used as the injury prediction result for the pedestrians involved.

[0039] Therefore, the pedestrian injury prediction method provided in this application no longer relies on simulation generalization data or is limited to the singularity of human body models. Instead, it is based on a large amount of real pedestrian accident injury data to accurately predict pedestrian injuries in current traffic accidents, ensuring that the prediction results are closer to the actual situation of real people. This provides reliable injury prediction support for emergency medical decision-making, enabling paramedics to take timely and effective rescue measures and reduce pedestrian injury rates. Furthermore, when searching for current accidents, the method adapts to the changing nature of traffic accidents by using search terms and precision limits based on mapping relationships, thereby achieving accurate and efficient data retrieval and improving the accuracy of pedestrian injury prediction.

[0040] In one optional embodiment, the historical accident database includes at least vehicle data, pedestrian data, and ambulance data; Vehicle data includes at least the collision speed of the vehicle when it collides with the pedestrian and the data on the point of collision at the time of the collision. Pedestrian data includes at least vital sign data, pedestrian injury data, and pedestrian collision point data at the time of the collision; among which, pedestrian injury data includes injury location, injury level, and injury description text; Emergency medical data should include at least the emergency measures, measure scores, and diagnostic results; emergency measures should include at least one of the following: on-site measures, transfer measures, and hospital admission measures; and emergency measures and measure scores should correspond one-to-one.

[0041] It is understandable that, in addition to the collision speed, the point of impact when a vehicle collides with a pedestrian can significantly affect the degree of injury suffered by the pedestrian. Therefore, in one optional embodiment, the vehicle data includes at least the collision speed and the point of impact data.

[0042] Regarding pedestrian data, it is understandable that when predicting pedestrian injuries, different pedestrians' physical characteristics (including but not limited to height, weight, and age) and collision points have varying degrees of influence on the severity of the injury. Therefore, pedestrian data in historical accident databases should at least include the above information to facilitate subsequent retrieval of accident boundary conditions, i.e., for determining the items to be retrieved.

[0043] In one optional embodiment, the height of the pedestrian involved can be calculated using a deep learning regression model. This model trains a neural network using a large amount of image data labeled with height, directly establishing a mapping relationship between image features and height. In another optional embodiment, the weight of the pedestrian involved can be calculated using a linear regression model. Specifically, the calculation formula is weight = a × height + b, where a and b are constants.

[0044] Of course, to predict the current pedestrian injury situation based on real historical pedestrian injury data, it is necessary to obtain pedestrian injury data from historical cases. Therefore, in one optional embodiment, the pedestrian data should at least include pedestrian injury data, and the pedestrian injury data should at least include the injury location, injury level, and injury description text.

[0045] It should be noted that the historical accident database provided in this application embodiment can be used for pedestrian injury prediction, therefore, the database necessarily stores historical pedestrian injury data. Furthermore, for the injury description text, in an optional embodiment, the damage description text can be generated by collecting target data from historical accident cases and analyzing the target data using a target large model. The target data to be analyzed may include, but is not limited to, pedestrian injury data, vehicle-mounted videos of collisions, and videos collected from road surveillance.

[0046] In another optional embodiment, to improve the accuracy of pedestrian injury prediction, historical accident cases can be analyzed by medical experts to obtain injury description text. This application does not limit the method of obtaining the injury description text. Similarly, the injury level can be obtained by evaluating the target data to be analyzed through a target large model, or it can be assessed by medical experts; this application also does not limit this method.

[0047] Based on the above embodiments, as an optional embodiment, to further reduce pedestrian injury and death rates, in addition to predicting pedestrian injuries after a traffic accident, corresponding rescue measures can be provided to ensure timely and accurate treatment of pedestrians. Therefore, rescue data is stored in a historical accident database, which, after identifying a target accident case, is used to generate rescue guidance strategies to guide the rescue of pedestrians currently involved in the accident.

[0048] To ensure the accuracy of the provided rescue data, in one optional embodiment, rescue measures can be scored to obtain a measure score, so that data with higher scores can be prioritized as rescue strategies for pedestrian injuries in the current accident. Similarly, the measure score can be evaluated by a target large model on the target data to be analyzed, or it can be scored by medical experts; this application does not limit this.

[0049] In addition, it should be noted that the target large model may include, but is not limited to, GPT (Generative Pre-trained Transformer) series, BERT model, large models based on Transformer architecture, LLaMA (Large Language Model Meta AI) and Tongyi Qianwen. This application does not limit the target large model.

[0050] Based on the above embodiments, as an optional embodiment, the pedestrian data also includes pedestrian gait data and pedestrian velocity values ​​at the time of collision; the pedestrian gait data includes a first value representing the straddle gait and a second value representing the upright gait. The pedestrian collision point data is the clock data corresponding to the pedestrian's direction of travel in the specified clock direction; The vehicle collision point data is the clock data corresponding to the vehicle's direction of travel in a specified clock direction.

[0051] Understandably, in traffic accidents, pedestrian gait, pedestrian speed, pedestrian collision point, and vehicle collision point are crucial parameters for pedestrian injury analysis. Therefore, pedestrian collision point, pedestrian gait, and vehicle collision point can be used as retrieval terms in the mapping relationship.

[0052] In one optional embodiment, for pedestrian gait recognition, key points can be determined in the acquired images including the pedestrian involved using the OpenPose model, and the key points are clustered to obtain a pedestrian skeleton, thereby completing gait recognition based on the pedestrian skeleton. Pedestrian gait includes straddling gait and upright gait.

[0053] Furthermore, it is necessary to calculate and determine the precision limit based on the accident data corresponding to the search terms. Therefore, it is necessary to convert pedestrian gait into data that can be calculated. In one optional embodiment, a first value is assigned to the straddling gait and a second value is assigned to the upright gait. For example, the first value can be 1 and the second value can be 0. The specific values ​​of the first and second values ​​are not limited in this application.

[0054] Figure 3 This is a schematic diagram of a specified clock structure provided in an embodiment of this application. Similarly, the vehicle collision point and pedestrian collision point also need to be converted into calculable values. In an optional embodiment, the collision point is described by the clock direction. Specifically, as shown... Figure 3 As shown, a specified clock is preset with the direction of 00 as the positive direction. The effects of vehicle collision points and pedestrian collision points in the mirrored condition are eliminated on this clock.

[0055] Based on this, pedestrian collision point data is the clock data corresponding to the pedestrian's direction of travel in that specified clock direction, and vehicle collision point data is the clock data corresponding to the vehicle's direction of travel in that specified clock direction. It should be noted that... Figure 3 As an example of a clock, this application does not limit the positive direction, structure, or numerical setting of the specified clock direction.

[0056] In one alternative embodiment, the perception of vehicle collision points and pedestrian collision points can be achieved by arranging trigger strips inside the front bulkhead and wheel arch trim panels of the vehicle. These trigger strips determine the location of the collision point by sensing vibrations and trim panel deformations exceeding a threshold. The perception accuracy of the first collision point is calibrated before the vehicle rolls off the production line.

[0057] Based on the above embodiments, as an optional embodiment, the mapping relationship includes the correspondence between vehicle speed range, search item, and precision limit; The search terms should include at least the collision speed, pedestrian collision point, vehicle collision point, and pedestrian physical characteristics. The maximum vehicle speed within the speed range is positively correlated with the number of search terms, while the maximum vehicle speed is negatively correlated with the precision limit.

[0058] Table 1 is a priority table of search terms provided in an embodiment of this application. In a specific embodiment, different search terms have different degrees of impact on pedestrian injury. Therefore, referring to Table 1, the priority of search terms is as follows: collision speed, pedestrian collision point, vehicle collision point, pedestrian physical characteristics, pedestrian gait, and pedestrian speed.

[0059] Table 1. Priority diagram of a search term It should be noted that physical characteristics data may include height and weight, and the priority of these data may be equal, or height may have a higher priority than weight; this application does not impose any limitations on this. Furthermore, the table above is an example of a search term, and this application does not limit the specific content and priority level of the search terms.

[0060] In specific embodiments, it is understood that the faster the vehicle speed, the more severe the pedestrian injury may be. In this case, to ensure the accuracy of injury prediction, more stringent boundary conditions are required for the retrieval. However, it is also understood that the more stringent the boundary conditions, the smaller the amount of data retrieved. Therefore, the accuracy conditions can be appropriately relaxed. That is, the larger the maximum vehicle speed within the speed range, the more types of search terms are available, and the smaller the corresponding precision limit.

[0061] Table 2 is a schematic table of a mapping relationship provided by an embodiment of this application. Based on the above analysis, referring to Table 2, in an optional embodiment, the vehicle speed range can be divided into three ranges, each range corresponding to a precision limit. The maximum value of the first vehicle speed range is not greater than the minimum value of the second vehicle speed range, and the maximum value of the second vehicle speed range is not greater than the minimum value of the third vehicle speed range.

[0062] Table 2 is a schematic diagram of a mapping relationship. In a specific embodiment, the speeds within the first speed range are relatively low, resulting in less severe pedestrian injuries. Therefore, the number of search terms can be reduced during the retrieval process, but the precision threshold needs to be increased. For the third speed range, for example, speeds exceeding 40 km / h, pedestrian injuries may be more severe, requiring more detailed and accurate injury prediction and first aid guidance. Therefore, a greater number of search term types are set. Under numerous boundary conditions, this inevitably leads to fewer historical accident cases meeting the requirements. In this case, the precision threshold can be lowered to obtain target accident cases that meet the conditions, thereby providing accurate injury prediction and first aid guidance.

[0063] Figure 4 This is a schematic diagram illustrating the principle of constructing a historical accident database, provided as an embodiment of this application. In an optional embodiment, constructing the historical accident database includes the following steps: Collect historical accident data from historical accident cases; Based on the collision speed in historical accident data and the preset speed range division rules, historical accident cases are classified. Based on the classification results and mapping relationships, cases in historical accident cases that include search terms corresponding to vehicle speed ranges are selected as candidate cases. Candidate cases are listed as cases to be entered into the database and then aggregated to form a historical accident database.

[0064] In a specific embodiment, the historical accident database is updated once every preset period (e.g., every week). When updating or building the database, a large amount of historical accident data from real historical accident cases is collected. In order to ensure the accuracy of the data in the database, and thus ensure the accuracy of subsequent pedestrian injury prediction, it is necessary to verify and screen the integrity, reliability, and validity of the data.

[0065] Based on the above embodiments, it is understood that the pre-built mapping relationship is crucial when predicting pedestrian injuries; that is, the collision speed and the search terms are critical to the prediction accuracy. Therefore, during database maintenance, in order to ensure the reliability of subsequent data retrieval, it is also necessary to screen historical accident cases based on the pre-built mapping relationship.

[0066] Specifically, in one alternative embodiment, such as Figure 4As shown, historical accident cases are categorized according to the collision speed and preset speed range division rules. Specifically, the collision speed at the time of the collision between the vehicle and the pedestrian in each historical accident case is obtained, and the historical accident cases are split based on the predefined speed ranges. For example, historical accident cases with speeds between 0-20 km / h are grouped into one category, historical accident cases with speeds between 20-40 km / h are grouped into another category, and historical accident cases with speeds greater than 40 km / h are grouped into a third category.

[0067] Furthermore, based on the aforementioned mapping relationship, candidate cases that satisfy the correspondence between speed ranges and search terms are selected from historical accident data. In other words, after classifying historical accident cases based on speed, different speed ranges correspond to different search terms, and each type of case must meet the data requirements of the corresponding search term. For example, in the example above, the search terms corresponding to speeds between 20-40 km / h include pedestrian collision point, pedestrian height, and pedestrian weight. Therefore, among the historical accident cases within this speed range, only cases that include data corresponding to pedestrian collision point, pedestrian height, and pedestrian weight can be considered as candidate cases.

[0068] To illustrate further, if the collision speed of a historical accident case falls within the first speed range, the corresponding search term is pedestrian collision point. In this case, the historical accident case must include pedestrian collision point data. If it does not, it cannot be used for subsequent accident data retrieval, meaning the data for that case is invalid and can be removed.

[0069] In other words, based on a pre-built mapping relationship, historical accident cases are screened, and only those historical accident data included in a case that satisfy the mapping relationship can be considered as candidate cases. Furthermore, the screened candidate cases are listed as cases to be entered into the database and aggregated to form a historical accident database.

[0070] Based on the above embodiments, as an optional embodiment, the pedestrian injury prediction method provided in this application further includes: When the historical accident data corresponding to any search term in the candidate cases comes from multiple acquisition channels, the difference between the data obtained from each acquisition channel is compared. If the difference is less than the corresponding preset threshold, the candidate case will be used as a case to be entered; and the cases to be entered will be combined to build a historical accident database.

[0071] In a specific embodiment, to ensure data reliability, in one optional embodiment, historical accident data corresponding to any search term can be collected through multiple methods. For example, vehicle speed can be obtained through vehicle speed sensors, traffic police accident investigation reports, etc., while vehicle collision points can be obtained through vehicle body traces, on-site monitoring, and dashcam footage, etc. Therefore, historical accident data corresponding to different search terms can include data collected through multiple paths.

[0072] Therefore, in candidate cases, if the historical accident data corresponding to any search term comes from multiple acquisition methods, in order to ensure data accuracy and reliability, in one optional embodiment, the difference between the data obtained from each acquisition method is calculated. If all the differences corresponding to the same candidate case are less than the corresponding preset threshold, indicating that the data reliability of the candidate case meets expectations, then the candidate case can be used as a case to be entered.

[0073] In other words, in the above embodiments, historical accident cases are first screened based on mapping relationships, and then a second round of screening is conducted based on the differences between data from multiple sources within the cases. It should be noted that when entering cases into the database, for data from multiple acquisition sources, the mean or mode can be used as the stored data in the database; this application does not impose any limitations on this.

[0074] Based on the above embodiments, as an optional embodiment, a historical accident database is constructed by collecting cases to be entered, including: Configure a unique identifier for each case to be entered; Extract pedestrian injury data and first aid data from historical accident data of cases to be entered; The injury level is assessed by using a specified large model to evaluate the injury level of pedestrian injury data, and the effectiveness of rescue measures in the rescue data is scored to obtain the measure score. Based on a unique identifier, the historical accident data, damage level, and response score of the case to be entered are linked and stored to form a historical accident database.

[0075] In a specific embodiment, in order to accurately retrieve historical accident cases and extract the corresponding historical accident data, a unique identifier is configured for each case to be entered. Table 3 is a schematic table of historical accident data provided in an embodiment of this application. As shown in Table 3, different case data are distinguished by a unique identifier.

[0076] In one optional embodiment, the unique identifier may consist of an accident number m, participants n (including pedestrians and vehicles), and participant number d. Participant n=1 indicates a pedestrian, and participant n=2 indicates a vehicle. This application does not limit the method of generating the unique identifier.

[0077] Table 3 is a schematic table of historical accident data. As shown in Table 3 above, the historical accident database contains d data entries. In order to provide effective rescue strategies after pedestrian injury prediction, it is necessary to analyze the pedestrian injury situation of historical accident cases when entering them into the database.

[0078] Figure 5 This is a schematic diagram illustrating the principle of constructing a historical accident database, as provided in another embodiment of this application. Specifically, as shown... Figure 5 As shown, pedestrian injury data and first aid data are extracted from historical accident data of cases to be entered. Pedestrian injury data includes the location of the injury and a description of the injury, while first aid data includes at least the first aid measures and diagnostic results.

[0079] In one optional embodiment, the pedestrian injury data also includes an injury level to characterize the severity of pedestrian injury. To ensure accurate and precise first aid for different body parts, as shown in Table 3, different injury parts correspond to a specific injury level. That is, the injury location, injury level, and loss description text are in a one-to-one correspondence.

[0080] Furthermore, at least one of the following is input into a designated large model for processing: the location of the injury, the description of the injury, the rescue measures, and the diagnosis results. This allows the designated large model to assess the degree of injury at different locations and generate injury levels for different locations.

[0081] In another alternative embodiment, in addition to assessing the severity of the injury to provide more appropriate guidance, it is understood that the effectiveness, appropriateness, and timeliness of previous first aid measures taken in past cases are also important for subsequent pedestrian first aid guidance. Therefore, the effectiveness of first aid measures can also be scored.

[0082] Specifically, at least one of the following will be input into a designated large model for processing: the injury site, injury description text, first aid measures, and diagnosis results. The designated large model will score the first aid measures for different injury sites and generate a score for the measures for different injury sites.

[0083] As shown in Table 3, the first aid measures include on-site measures, transport measures, and hospital admission measures, and correspondingly, the measures are scored in three ways. Specifically, on-site measures correspond to the first score, transport measures to the second score, and hospital admission measures to the third score. For different injury sites, there is a one-to-one correspondence between a first aid measure and a diagnostic result. That is, the injury site, the first aid measure, and the diagnostic result are in a one-to-one relationship.

[0084] It should be noted that, in one optional embodiment, in addition to determining the injury level and intervention score through a specified large model, the level can also be determined and scored by medical experts. Furthermore, weights are assigned to the assessment results from different approaches, and finally, a weighted sum is performed to obtain the final injury level and intervention score. In one optional embodiment, to ensure that the injury level is an integer, the level can be rounded up to the nearest integer, where the injury level may be a decimal.

[0085] Furthermore, it should also be noted that, in specific embodiments, the designated large model may include, but is not limited to, the GPT (Generative Pre-trained Transformer) series, the BERT model, large models based on the Transformer architecture, LLaMA (Large Language Model Meta AI), and Tongyi Qianwen. This application does not limit the designation of the large model.

[0086] Furthermore, based on a unique identifier, the historical accident data of each case to be entered is sequentially entered into the database. Specifically, each case to be entered includes vehicle data, pedestrian data, and ambulance data. The pedestrian data includes the injury level generated in this application embodiment, and the ambulance data includes the measure score generated in this application embodiment. In addition, each case to be entered may also include injury data of the driver, passengers, and other persons in the vehicle, and may also include historical accident location data. This application does not limit the specific data stored for each case. Among these, the historical accident location data includes at least one of the following: accident time, latitude and longitude data, and weather data.

[0087] In one optional embodiment, the historical accident database can be stored in tabular form, and the storage method is not limited. It is worth noting that, to ensure that subsequent data can be viewed intuitively, when storing data based on unique identifiers, the data is categorized and stored as vehicle data, pedestrian data, and ambulance data.

[0088] In an optional embodiment, the vehicle data further includes vehicle size data, and the method for predicting pedestrian injuries further includes: Obtain the target vehicle size data of the vehicle involved in the incident; From the historical accident database, cases where the absolute value of the difference between the size data of the target vehicle and the actual size data of the target vehicle is less than a preset value are selected as matching cases for the vehicle involved in the accident.

[0089] Understandably, vehicle size data is used to measure different types of vehicles, i.e., different vehicle models. In specific embodiments, different vehicle models cause significant differences in the amount of injury to pedestrians. Therefore, to further improve the accuracy of pedestrian injury prediction, vehicle size data is included as part of the vehicle data for matching target accident cases.

[0090] In other words, in specific embodiments, different vehicle models cause significant differences in the degree of injury to pedestrians. Therefore, when searching for target accident cases in the historical accident database, cases with significant differences in size from the currently involved vehicle are first frozen in order to quickly obtain cases with similar vehicle sizes.

[0091] Figure 6 This is a schematic diagram of the principle of vehicle size data provided in an embodiment of this application. Table 4 is a schematic table of vehicle size data provided in an embodiment of this application. For ease of understanding, the following will be combined with... Figure 6 Table 4 explains the vehicle size data.

[0092] In one alternative embodiment, the analysis is based on a pedestrian impact in the vehicle's direction of travel. Therefore, as shown in Table 4, the vehicle size data includes first outer contour data and second outer contour data, such as... Figure 6 As shown, the first outer contour data is the height data between the front of the vehicle and the ground, and the outer contour data is the width data of the front of the vehicle.

[0093] Table 4 is a schematic table of vehicle size data. like Figure 6 As shown in Table 4, in one optional embodiment, the first outer contour dimension includes at least the lower edge height H1 of the front bumper, the upper edge height H2 of the front bumper, the ground clearance height H3 of the front end of the hood, and the ground clearance height H4 of the rear end of the hood. The second outer contour dimension includes at least the front bumper width La, the distance from the rear end of the front bumper to the front end of the hood Lb, the hood width Lc, and the windshield width Ld.

[0094] It should be noted that Table 4 and Figure 6 This is merely one example of vehicle size data. In fact, the more vehicle size data available, the higher the accuracy of vehicle model determination and the better the accuracy of pedestrian prediction. Therefore, this application does not limit the specific content and type of vehicle size data, and it can be set according to needs. Furthermore, it should be noted that the vehicle size data can be modified by the user so that the data can be updated promptly when the user modifies the vehicle.

[0095] In a specific embodiment, when a vehicle retrieves data from the historical accident case database from the cloud and stores it locally, it can immediately perform vehicle size data matching. Specifically, from the historical case database, cases where the absolute value of the difference between the target vehicle size data and the data of the current vehicle under each size type condition in Table 4 above is less than a preset value are selected as matching cases.

[0096] Referring to Table 4, if the absolute value of the first difference between the height H1 of the lower edge of the front bumper in any case in the database and the height H1 of the lower edge of the front bumper in the target vehicle size data is less than a preset value, the other first outer contour dimensions and second outer contour dimensions are calculated in the same way. If the absolute value of the difference corresponding to all types of dimensions is less than the preset value, it indicates that the vehicle in the historical case is similar in size to the current vehicle involved, which meets the expectation and can be used as a matching case.

[0097] For example, after matching vehicle size data, the matching cases for vehicle A include Case 1 to Case 100. The vehicle models of these 100 cases are not significantly different from those of vehicle A and can be considered to be of the same type. Therefore, if vehicle A collides with a pedestrian during its subsequent driving process, it can immediately be matched with these 100 cases according to the aforementioned search terms and target precision limit, so as to quickly identify target accident cases that can be used for pedestrian injury prediction from among the 100 cases.

[0098] Therefore, the pedestrian injury prediction method provided in this application uses vehicle size data as a preliminary case screening condition to perform independent case set matching for each vehicle, so that the target accident case can be quickly and accurately identified when predicting pedestrian losses.

[0099] Figure 7 This is a schematic diagram illustrating the principle of a pedestrian injury prediction method provided in another embodiment of this application. In an optional embodiment, the method uses the term to be searched as the accident case retrieval condition to search for target accident cases that meet the target precision limit in a pre-built historical accident database, including: Collect current accident data corresponding to the item to be searched in the current traffic accident; In the matchable cases, retrieve the target historical accident data where the vehicle speed is within the target speed range and the item to be retrieved corresponds to the target vehicle speed range; Determine the parameter values ​​used to characterize the degree of difference between current accident data and target historical accident data; Among the matchable cases, the specified cases whose parameter values ​​are not greater than the target precision limit are designated as target incident cases.

[0100] Based on the above embodiments, such as Figure 7As shown, after filtering matching cases based on vehicle size data, the matching cases are further searched for the search term and the target precision limit. Specifically, the current accident data corresponding to each search term is collected from the current number of traffic accidents. Furthermore, from the matching cases, the target historical accident data corresponding to the search term that are within the target speed range are extracted. For example, referring to Table 2, under the condition of the same speed range, when the search term is pedestrian collision point, the corresponding target historical accident data is pedestrian collision point data.

[0101] Furthermore, for the same search term, a parameter value is calculated to characterize the degree of difference between the current accident data and the target historical accident data. This parameter value can be the absolute value of the difference between the two over time, or it can represent the percentage difference in data; this application does not limit this. When the parameter value is not greater than the target precision limit, it indicates that the corresponding matching case meets the search requirements, that is, the matching case can be used as the target accident case. For ease of understanding, the following example uses a parameter value as a percentage.

[0102] For example, if the search term is pedestrian height, the corresponding historical accident data is 180 cm, the current accident data is 175 cm, and the target precision limit is 5%, then the parameter value can be calculated as follows: (180-175) / [(180+175) / 2]*100%=2.82%, or (180-175) / 180*100%=2.78%, or (180-175) / 175*100%=2.86%. Clearly, in any of these three possible calculation methods, the obtained parameter value is less than 5%, indicating that the matching case can be used as the target accident case.

[0103] like Figure 7 As shown, when no matching cases can be found that meet the target speed range, the search term, and the target precision limit, that is, there are no cases similar to the current traffic accident in the current database, in order to ensure that pedestrians can receive timely assistance, the accident point data of the current traffic accident can be collected and transmitted to the remote guidance and rescue system.

[0104] In one optional embodiment, the accident location data may include, but is not limited to, accident time, latitude and longitude data, weather data, altitude data, and air pressure data. This application does not limit the storage method or data storage accuracy of the accident location data. After acquiring the accident location data, the remote rescue guidance system can provide remote rescue guidance suggestions based on the local geographical environment and weather conditions to prevent secondary injuries to pedestrians.

[0105] In one alternative embodiment, such as Figure 7As shown, among the matchable cases, those with parameter values ​​not exceeding the target precision limit are designated as target incident cases, including: If the specified case is unique, the specified case will be used as the target accident case; If the specified case is not unique, determine whether there are any cases in the same preset division area that belong to the current traffic accident; If it exists, the cases in the same area are filtered step by step according to the preset priority filtering conditions. When only one case remains after any level of filtering, the filtering stops and the target accident case is obtained. The preset priority filtering conditions include, in order: unique case, unique minimum value of parameter value, unique highest level of damage level, and unique maximum value of the sum of measure scores. If it does not exist, the specified cases are filtered step by step to obtain the target accident cases.

[0106] In a specific implementation, if there is only one designated case that meets the target precision limit, that designated case can be directly used as the target accident case. If there are multiple designated cases, a more accurate case that is closer to the current traffic accident can be selected as the target accident case.

[0107] Specifically, due to significant differences in traffic accidents occurring in different regions—for example, even with similar information such as vehicle speed and size—the degree of pedestrian injury can vary greatly between mountainous and plain areas. Therefore, in one optional embodiment, when the specified case is not unique, priority is given to determining whether local data exists within the specified case, i.e., whether there are cases within the same preset division area as the current traffic accident.

[0108] If local cases from the same region exist, the target incident case is selected from the local historical data according to preset priority filtering criteria. If no local historical data exists, the target incident case is then selected from the specified non-local cases according to preset priority filtering criteria.

[0109] Whether filtering cases within the same region or specific cases using preset priorities, the following applies: Figure 7 The preset priority filtering conditions shown are applied sequentially. During the process, if only one case is obtained after any level of filtering, the subsequent filtering is stopped immediately, and the remaining case is taken as the target accident case.

[0110] Specifically, during the step-by-step screening, the process first determines whether a case is unique. If unique, the case is directly designated as the target accident case. If not unique, the process checks for a unique minimum parameter value; if found, it is designated as the target accident case. Otherwise, the process checks for a unique maximum damage level; if found, it is designated as the target accident case. Finally, the process checks for a unique maximum sum of the measure scores; if found, it is designated as the target accident case.

[0111] In fact, the above-mentioned preset priority filtering conditions can basically filter out a unique target accident case. If there are still multiple cases, and multiple cases are characterized by being highly similar to the current traffic accident in terms of various indicators, then any one of these cases can be used as the target accident case.

[0112] In an optional embodiment, after generating a predicted injury result for the pedestrian involved based on pedestrian injury data in the target accident case, the method further includes: Obtain medical information from hospitals within the predefined area where the current traffic accident occurred; the medical information includes at least blood bank information, drug information, and medical equipment information; In the case of the target accident, is there a target first aid measure that matches the medical information? If present, the target rescue measures will be transmitted to the real-time rescue terminal; If not, the damage prediction results and accident location data will be transmitted to the remote guidance and rescue system; the accident location data shall include at least the accident time, latitude and longitude data and weather data.

[0113] In a specific embodiment, after predicting the pedestrian's injuries, professional first aid guidance can be provided based on a historical accident database to further reduce the pedestrian injury and death rate. Specifically, medical information from various hospitals within a preset area where the current traffic accident occurred can be obtained, i.e., medical information from the city where the traffic accident occurred can be obtained.

[0114] Through the above embodiments, target accident cases highly similar to the current traffic accident have been identified. At this point, it is necessary to determine whether the emergency medical data in the target accident cases can be used to provide injury relief to the pedestrians involved in the current accident. Specifically, it is necessary to determine whether there are target emergency medical measures in the target accident cases that match the medical information; that is, it is necessary to determine whether the emergency medical measures used in the target accident cases are compatible with the current level of medical care. For example, it is necessary to determine whether local hospitals include the medical equipment used in the target accident cases, and whether the blood type, blood volume, and medications are all suitable for the needs of the target accident cases.

[0115] If the target accident case has target rescue measures that meet the current medical standards, these measures will be transmitted to the real-time rescue terminal, which must include at least a hospital. If no target rescue measures are available, the injury prediction results and accident site data will be transmitted to the remote guidance and rescue system for remote rescue guidance of pedestrians. In addition, collected pedestrian data (including but not limited to height, weight, and age) and video footage of the traffic accident scene can also be transmitted to the remote guidance and rescue system; this application does not limit the scope of such transmission.

[0116] It is worth noting that, as shown in Table 3, the corresponding first aid measures for different injury sites include on-site measures, transfer measures, and hospital admission measures. When determining the target first aid measures, for different injury sites, only measures that meet the medical information requirements are considered as target first aid measures.

[0117] In the above embodiments, the method for predicting pedestrian injuries has been described in detail. This application also provides an embodiment of a device for predicting pedestrian injuries.

[0118] Figure 8 This is a schematic diagram of the structure of a pedestrian injury prediction device provided in an embodiment of this application, as shown below. Figure 8 As shown, the device includes: The collision speed acquisition module 80 is used to acquire the current collision speed of the vehicle involved in the current traffic accident when it collides with the pedestrian involved. The vehicle speed range recognition module 81 is used to identify the target vehicle speed range in which the current collision vehicle speed is located. The retrieval condition determination module 82 is used to obtain the search terms and target precision limits corresponding to the target vehicle speed range based on the pre-built mapping relationship; wherein, the search terms are the boundary condition types of accident case retrieval; and the target precision limits are the accuracy conditions of accident case retrieval. The target accident case retrieval module 83 is used to retrieve target accident cases that meet the target precision limit from a pre-built historical accident database, using the search term as the accident case retrieval condition. The damage result generation module 84 is used to generate damage prediction results for the pedestrians involved in the accident based on the pedestrian damage data in the target accident case.

[0119] Figure 9 This is a schematic diagram of the structure of a vehicle provided in an embodiment of this application, such as... Figure 9 As shown, the vehicle includes: a memory 90 for storing computer programs; The processor 91 is configured to execute a computer program to implement the steps of the pedestrian injury prediction method as described in the above embodiments.

[0120] The processor 91 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 91 may be implemented using at least one of the following hardware forms: Digital Signal Processor (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 91 may also include a main processor and a coprocessor. The main processor, also known as the Central Processing Unit (CPU), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 91 may integrate a Graphics Processing Unit (GPU), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 91 may also include an Artificial Intelligence (AI) processor, which is used to handle computational operations related to machine learning.

[0121] The memory 90 may include one or more computer-readable storage media, which may be non-transitory. The memory 90 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In this embodiment, the memory 90 is used to store at least the following computer program 901, which, after being loaded and executed by the processor 91, is capable of implementing the relevant steps of the pedestrian injury prediction method disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 90 may also include an operating system 902 and data 903, and the storage method may be temporary or permanent storage. The operating system 902 may include Windows, Unix, Linux, etc. The data 903 may include, but is not limited to, relevant data involved in the pedestrian injury prediction method.

[0122] In some embodiments, the vehicle may also include a display screen 92, an input / output interface 93, a communication interface 94, a power supply 95, and a communication bus 96.

[0123] Those skilled in the art will understand that Figure 9 The structure shown does not constitute a limitation on the vehicle and may include more or fewer components than illustrated.

[0124] The vehicle provided in this application includes a memory and a processor. When the processor executes the program stored in the memory, it can implement the pedestrian injury prediction method described in the above embodiments.

[0125] It should be noted that although the operations are depicted in a specific order in the accompanying drawings, this should not be construed as requiring these operations to be performed in the specific order shown or sequentially, or requiring all illustrated operations to be performed to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. Furthermore, the separation of various system modules and components in the above embodiments should not be construed as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.

Claims

1. A method for predicting pedestrian injuries, characterized in that, The method includes: Obtain the current collision speed of the vehicle involved in the current traffic accident at the moment of collision with the pedestrian involved; Identify the target speed range in which the current collision speed is located; Based on a pre-built mapping relationship, the search terms and target precision limits corresponding to the target vehicle speed range are obtained; wherein, the search terms are the boundary condition types for accident case retrieval; and the target precision limit is the accuracy condition for accident case retrieval. Using the term to be searched as the accident case search criteria, target accident cases that meet the target precision limit are retrieved from a pre-constructed historical accident database. Based on pedestrian injury data in the target accident case, injury prediction results for the pedestrians involved are generated.

2. The method for predicting pedestrian injuries as described in claim 1, characterized in that, The historical accident database includes at least vehicle data, pedestrian data, and ambulance data; The vehicle data includes at least the collision speed of the vehicle when it collides with the pedestrian and the collision point data of the vehicle when the collision occurs. The pedestrian data includes at least vital sign data, pedestrian injury data, and pedestrian collision point data at the time of the collision; wherein, the pedestrian injury data includes the injury location, injury level, and injury description text; The rescue data includes at least rescue measures, measure scores, and diagnostic results; the rescue measures include at least one of the following: on-site measures, transfer measures, and hospital admission measures; and the rescue measures correspond one-to-one with the measure scores.

3. The method for predicting pedestrian injuries as described in claim 2, characterized in that, The pedestrian data also includes pedestrian gait data and pedestrian velocity values ​​at the time of collision; the pedestrian gait data includes a first value representing the straddle gait and a second value representing the upright gait. The pedestrian collision point data is the clock data corresponding to the pedestrian's direction of travel in a specified clock direction; The vehicle collision point data is the clock data corresponding to the vehicle's direction of travel in the specified clock direction.

4. The method for predicting pedestrian injuries as described in claim 1, characterized in that, The mapping relationship includes the correspondence between vehicle speed range, search terms, and precision limits; The search terms include at least collision speed, pedestrian collision point, vehicle collision point, and pedestrian physical characteristics; The maximum vehicle speed in the speed range is positively correlated with the number of types of the search terms, and the maximum vehicle speed is negatively correlated with the precision limit.

5. The method for predicting pedestrian injuries as described in claim 1, characterized in that, Building the historical accident database includes the following steps: Collect historical accident data from historical accident cases; Based on the collision speed in the historical accident data and the preset speed range division rules, the historical accident cases are classified. Based on the classification results and the mapping relationship, cases in the historical accident cases that include the search item data corresponding to the vehicle speed range are selected as candidate cases. The candidate cases are listed as cases to be entered into the database and then aggregated to form the historical accident database.

6. The method for predicting pedestrian injuries as described in claim 5, characterized in that, The method further includes: In the candidate cases, when the historical accident data corresponding to any search item comes from multiple acquisition methods, the difference between the data obtained by each acquisition method is compared. If all the differences are less than the corresponding preset threshold, the candidate cases are used as the cases to be entered; and the cases to be entered are collected to construct the historical accident database.

7. The method for predicting pedestrian injuries as described in claim 6, characterized in that, The collected cases to be entered are used to construct the historical accident database, including: Configure a unique identifier for each of the cases to be entered; Extract pedestrian injury data and first aid data from the historical accident data of the cases to be entered; The pedestrian injury data is assessed for injury level by specifying a large model to obtain the injury level, and the effectiveness of the rescue measures in the rescue data is scored to obtain the measure score. Based on the unique identifier, the historical accident data of the case to be entered, the damage level, and the measure score are associated and stored to form the historical accident database.

8. The method for predicting pedestrian injuries as described in claim 2, characterized in that, The vehicle data also includes vehicle size data, and the method further includes: Obtain the target vehicle size data of the vehicle involved in the incident; From the historical accident database, cases where the absolute value of the difference between the size data of the target vehicle and the size data of the target vehicle for each size type is less than a preset value are selected as matching cases for the vehicle involved in the accident.

9. The method for predicting pedestrian injuries as described in claim 8, characterized in that, The step of using the search term as the accident case retrieval condition to retrieve target accident cases that meet the target precision limit in a pre-constructed historical accident database includes: Collect the current accident data corresponding to the item to be searched in the current traffic accident; In the matching cases, the retrieved vehicle speed is within the target vehicle speed range, and the target historical accident data corresponding to the item to be retrieved is selected. Determine parameter values ​​used to characterize the degree of difference between the current accident data and the target historical accident data; Among the matchable cases, the case whose parameter value is not greater than the target precision limit is designated as the target incident case.

10. The method for predicting pedestrian injuries as described in claim 9, characterized in that, The step of designating a case among the matchable cases whose parameter value is not greater than the target precision limit as the target incident case includes: If the specified case is unique, the specified case shall be used as the target accident case; If the specified case is not unique, determine whether there is a case in the same area that belongs to the same preset division area as the current traffic accident; If such a case exists, the case in the same area is filtered step by step according to the preset priority filtering conditions. When only one case remains after any level of filtering, the filtering stops to obtain the target accident case. The preset priority filtering conditions include, in order: the case is unique, the parameter value has a unique minimum value, the damage level has a unique maximum level, and the sum of the measure scores has a unique maximum value. If it does not exist, the specified cases are filtered step by step to obtain the target accident cases.

11. The method for predicting pedestrian injuries as described in claim 2, characterized in that, After generating the injury prediction result for the pedestrian involved based on the pedestrian injury data in the target accident case, the method further includes: Obtain medical information from hospitals within the preset area where the current traffic accident occurred; the medical information includes at least blood bank information, drug information, and medical equipment information. Determine whether there are any target rescue measures in the target accident cases that match the medical information; If present, the target rescue measures will be transmitted to the real-time rescue terminal; If not, the damage prediction results and accident point data are transmitted to the remote guidance and rescue system; the accident point data includes at least the accident time, latitude and longitude data, and weather data.

12. A pedestrian injury prediction device, characterized in that, The device includes: The collision speed acquisition module is used to acquire the current collision speed of the vehicle involved in the current traffic accident when it collides with the pedestrian involved. The vehicle speed range identification module is used to identify the target vehicle speed range in which the current collision vehicle speed is located; The retrieval condition determination module is used to obtain the search terms and target precision limit corresponding to the target vehicle speed range based on a pre-built mapping relationship; wherein, the search terms are the boundary condition types for accident case retrieval; and the target precision limit is the accuracy condition for accident case retrieval. The target accident case retrieval module is used to retrieve target accident cases that meet the target precision limit from a pre-built historical accident database, using the item to be retrieved as the accident case retrieval condition. The injury result generation module is used to generate injury prediction results for the pedestrians involved in the accident based on the pedestrian injury data in the target accident case.

13. A vehicle comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the pedestrian injury prediction method according to any one of claims 1 to 11.

Images