Computer-implemented method for generating a report regarding an accident event in road traffic, and system and computer program product

Abstract

The invention relates to a computer-implemented method for generating a report regarding an accident event in road traffic in which at least one vehicle is involved, and to a system and a computer program product, wherein the method enables reliable and simple generation of an accident report which contains all relevant information and thus reduces the need for expensive and complex accident reconstruction methods.

Description

Computer-implemented method for generating a report on a Road traffic accident, as well as system and computer program product

[0001] The present invention relates to a computer-implemented method for generating a report on a road traffic accident involving at least one vehicle.

[0002] Furthermore, the invention relates to a system and a computer program product which execute the computer-implemented method for generating a report. State of the art

[0003] Methods for automatically completing accident reports are known from the prior art (EP 2 950 251 A1). These methods can, for example, capture data relating to vehicles and drivers and reproduce them in the accident report. Similarly, such methods can record images or videos of the vehicles and include them in the accident report. However, these methods are generally limited to reproducing the captured data in the accident report.

[0004] Established procedures, however, typically suffer from the fact that they are carried out by laypersons, or the corresponding equipment is operated by them. Such laypersons, and sometimes even theoretically trained individuals (e.g., police officers), often lack the knowledge of what data is necessary for a comprehensive accident report and how this data must be collected and processed. Consequently, due to the poor or incomplete data, the accident reports cannot be used, or can only be partially used, by the responsible authorities (police, insurance companies, accident analysis experts, etc.). This often necessitates complex accident reconstruction procedures to clarify the circumstances of the accident.

[0005] Taking photographs, in particular, presents several challenges for those involved in accidents (e.g., drivers or passengers of a vehicle). On the one hand, people involved in an accident may be emotionally distressed or in shock, impairing their judgment and attention. This can hinder clear thinking and the correct and complete execution of all necessary steps related to documenting an accident, including taking photographs. Taking pictures of the vehicle and its surroundings can be difficult. On the other hand, people involved in accidents are usually laypersons and unfamiliar with the specific documentation requirements associated with accident reports. For example, they often lack the knowledge of which specific angles and areas need to be photographed to provide a clear and comprehensive overview of the damage, and which important details, such as close-ups of the damage, the overall scene, vehicle positions, road conditions, traffic signs, and any debris or skid marks, should be captured. Furthermore, poor lighting conditions, adverse weather conditions, or active traffic can make taking pictures even more difficult. Disclosure of the invention

[0006] The present invention therefore aims to provide a method for creating a report on a road traffic accident, which enables a reliable and simple creation of an accident report containing all relevant information and thus reduces the need for expensive and time-consuming accident reconstruction procedures.

[0007] The invention solves the stated problem by means of a computer-implemented method according to independent claim 1.

[0008] The computer-implemented method according to the invention for creating a report on a road traffic accident involving at least one vehicle comprises at least the steps described below (the steps can be carried out in any order unless otherwise specified).

[0009] According to one aspect of the invention, in a process step a series of images of the vehicle and its surroundings is captured by a user using an end device, wherein instructions for recording the images are displayed to the user on the end device during the capture of the series of images.

[0010] By providing instructions to the user on the device during the image capture process, every user, regardless of their emotional state or level of knowledge, is always able to capture the necessary images for documenting the accident in a way that meets the respective documentation requirements (e.g., of insurance companies, police, researchers, etc.). The device can preferably be equipped with appropriate [features / devices / etc.]. The programming ensures that the user doesn't forget to take important pictures and that they take or repeat the shots as often as necessary until all required images are captured in usable quality. Furthermore, the user is relieved of the decision-making process regarding which details or angles should be captured. This provides a particularly reliable and, at the same time, easy-to-use method for creating an accident report.

[0011] According to one aspect of the invention, dimensions of the vehicle and its surroundings are recorded in a process step.

[0012] Capturing the dimensions of the vehicle and its surroundings can improve the interpretation of the captured images, particularly by enabling enhanced contextual accuracy / precision, damage assessment, and accident reconstruction. Dimensions help, among other things, to contextualize the captured images, validate witness statements, support insurance claims with concrete evidence, and facilitate compliance with legal regulations and expert opinions. Overall, capturing dimensions adds an extra layer of precision, improving the understanding and resolution of accident-related issues.

[0013] According to one aspect of the invention, a 3d model of the vehicle is created from the series of captured images in a process step.

[0014] Creating a 3D model of a damaged vehicle offers the advantage of providing a comprehensive, accurate, and easily understandable representation of the damage. In particular, this facilitates all aspects of accident analysis and processing, enabling detailed damage analysis and precise accident reconstruction, which can serve as objective evidence in insurance and legal proceedings and improve communication between the parties involved. Furthermore, it supports accurate damage assessment and repair planning, promotes the training and education of specialists, and serves as a permanent, easily shareable documentation tool.

[0015] According to one aspect of the invention, in a process step a rectified top view of the vehicle and its surroundings is created from the series of captured images and the captured dimensions.

[0016] A rectified top view of the vehicle and its surroundings offers numerous advantages for accident reports: It allows for a true-to-scale and distortion-free representation of the accident scene, which helps to determine the exact position and distances. The accident scene precisely records the vehicles and objects involved. This facilitates accident reconstruction and analysis of the accident's cause. Furthermore, it supports detailed damage assessment and simplifies the comparison and documentation of damages. The clear visual representation promotes a shared understanding and communication between all parties involved, such as insurance companies, police, and experts. Finally, it serves as a durable and unambiguous piece of evidence that can be of great importance in legal disputes and insurance claims.

[0017] In particular, the controlled capture of images under instruction and the recording of dimensions of the vehicle and the surroundings allows the rectified top view to be created automatically with minimal error, thus providing a particularly reliable and easy-to-execute procedure for creating an accident report.

[0018] According to one aspect of the invention, in one process step a report on the event (accident report) is generated using the 3D model of the vehicle and the rectified top view of the vehicle and its surroundings.

[0019] If the accident report is created using the 3D model of the vehicle and the rectified top view of the vehicle and its surroundings, the method according to the invention can reliably produce an accident report that, on the one hand, enables a detailed and comprehensive damage analysis from all perspectives and, on the other hand, a distortion-free and true-to-scale representation of the accident scene and the vehicle, capturing precise positions and distances. Such an accident report can significantly facilitate the analysis of the cause of the accident and the damage assessment, and ensure improved communication and a uniform understanding among all parties involved, including insurance companies, police, lawyers, and experts, or serve as comprehensive evidence.

[0020] According to the invention, a simple and reliable method for creating an accident report has been created, which can support persons involved in accident events in creating meaningful and good accident reports and thus enables comprehensive documentation of the accident event even without special knowledge and in stressful situations.

[0021] Generally speaking, a road traffic accident is defined as any event involving at least one vehicle and potentially triggering an insurance claim or resulting in damage to a vehicle. vehicle and / or person. The terms accident event and traffic accident or accident are used synonymously in the course of this disclosure.

[0022] Generally, a road traffic accident report is understood to be a document that comprehensively details an accident and records the relevant information for insurance companies, the police, and legal purposes. It typically includes a description of the accident sequence, the vehicles and persons involved, the accident damage, witness statements, and, where applicable, photographs of the accident scene. Sketches or plans of the accident scene, as well as technical details such as directions of travel and speeds, are also frequently included. The terms "accident report" and "accident report" are used synonymously in this disclosure.

[0023] Generally, it is mentioned that the terminal device can be any portable computing device. Preferably, the terminal device can be a smartphone or tablet with an integrated camera and screen.

[0024] It is also generally mentioned that the creation of a rectified top view is achieved using an image rectification process, a technique that corrects distortions in images to normalize or improve their representation and is crucial in photogrammetry and mapping for precise measurements. The process converts perspective images into orthogonal images (orthoimages), using known quadrilaterals to make oblique images appear as if they were taken from above. This technique uses the real dimensions of the objects in the image, e.g., a square, for perspective correction.

[0025] The object of the invention is further achieved by a system for generating a report on an event in which at least one vehicle is involved, wherein the system comprises at least one terminal device and one server. The terminal device and / or the server are programmed to execute the computer-implemented method according to the invention.

[0026] Furthermore, the object of the invention is solved by a computer program product, wherein the computer program product comprises instructions that, when loaded and run on an end device and / or a server, execute the corresponding steps of the computer-implemented method according to the invention.

[0027] Preferred embodiments of the invention are described below using exemplary embodiments. Unless otherwise stated, the illustrated embodiments, or individual aspects thereof, can be combined with one another as desired.

[0028] According to one implementation variant, creating a rectified top-down view of the vehicle and its surroundings allows for the determination of the vehicle's final resting position within its environment. Establishing this resting position enables a precise analysis of the accident scene (in the vicinity of the vehicle) by documenting the vehicle's exact final position after the accident, which is essential for reconstructing the accident sequence. Determining the resting position is particularly helpful in understanding other parameters, such as the vehicle's direction of movement, speed, and points of impact, thus allowing for a more accurate determination of the accident's cause.

[0029] According to one implementation variant, the final rest position can be defined, in particular, by distances and dimensions between the vehicle and objects in the vehicle's vicinity. This enables a clear (quantitative) determination of the vehicle's final rest position in the rectified top view.

[0030] According to one implementation variant, the vehicle's coordinates can be captured in a single step and taken into account when creating the rectified top view of the vehicle and its surroundings. The rectified top view can thus include not only relative dimensions but also the actual location of the accident, thereby eliminating the need for manual documentation of the accident location by a user, which is usually based on a verbal description of the location and is therefore prone to errors.

[0031] According to one implementation variant, the vehicle's coordinates can be acquired by the user's device in a single process step. The coordinates can be acquired, in particular, via receivers for navigation satellite systems, such as GPS, GLONASS, Galileo, etc., built into the device. Specifically, instructions can be displayed to the user on the device during coordinate acquisition, for example, to guide the user to a specific defined position or reference point (such as a corner of the vehicle, etc.) to improve coordinate accuracy.

[0032] According to one implementation variant, the direction of travel and / or the route taken by the vehicle before the event can be recorded in a single process step. Recording the direction of travel and the route taken can provide crucial information for subsequent accident reconstruction. It helps to better understand the precise sequence of events and, in particular, to analyze how the vehicles involved became involved. The direction of travel and the route taken can also reveal whether traffic regulations were observed and whether lane changes, turns, or changes of direction were made. This can be especially helpful in clarifying responsibilities and liability issues following the accident and can serve as a basis for insurance claims.

[0033] According to another implementation variant, the user can record the direction of travel and / or the route using the device, with instructions displayed on the device during this process. This provides reliable support for the user in recording the direction of travel and the route, further reducing the potential for errors. Specifically, the instructions can include prompts to draw the direction of travel on a map or image displayed on the device. This allows the user to enter information intuitively and without complex considerations, which in turn improves the reliability of the process and ultimately the quality of the accident report.

[0034] According to one embodiment of the invention, the user can at least partially capture the dimensions of the vehicle and its surroundings using the terminal device, with instructions for recording the dimensions being displayed to the user on the terminal device during the measurement process. This provides reliable support for the user in capturing the dimensions of the vehicle and its surroundings and reduces the potential for errors during measurement. In particular, the instructions can include prompts for the correct recording of measurements or prompts regarding which specific dimensions are to be recorded. This reduces the likelihood that the user will record incorrect dimensions or forget to record certain or all dimensions altogether.

[0035] According to one embodiment of the invention, capturing the dimensions of the vehicle and its surroundings can include retrieving the vehicle's dimensions from a database. For example, if the model and make of the vehicle are already known, capturing the vehicle's dimensions can be done automatically. The vehicle's own dimensions are no longer needed, as they can be retrieved from a database. Furthermore, the dimensions retrieved in this way are exact and not subject to any measurement errors that may occur when the user enters the dimensions.

[0036] According to one embodiment of the invention, the capture of a series of images of the vehicle and its surroundings can take place at the terminal device in an augmented reality (AR) environment.

[0037] According to one embodiment of the invention, the dimensions of the vehicle and its surroundings can be detected in an augmented reality (AR) environment.

[0038] According to one embodiment of the invention, the detection of the direction of travel and / or the route traveled at the terminal device can take place in an augmented reality environment.

[0039] An AR environment can be understood as the visual representation of information in the real environment, i.e., the supplementation of images or videos with computer-generated additional information or virtual objects by means of overlay / superimposition.

[0040] According to one implementation variant, the AR environment can be created electronically on the end device (e.g., a smartphone) by combining / overlaying the reproduction of sensory perceptions from the real world with virtual elements. For this purpose, the real world can be captured via sensor data (e.g., from the smartphone's camera) and overlaid with information, such as instructions or input options, displayed on the end device's screen. The user can thus "see through the end device" the real environment, with the information then being displayed at corresponding locations within that real-world environment.

[0041] When capturing the series of images of the vehicle and its surroundings, according to one implementation variant, a video of the surroundings can be continuously recorded by the camera of the terminal device in the augmented reality environment and simultaneously overlaid on a screen of the terminal device with information about the movement of the terminal device.

[0042] Furthermore, according to one implementation variant, the series of images can be recorded as a video. This can further simplify image capture for the user, as the user only needs to move the device according to the displayed directional instructions while it records the video. The necessary images for creating the 3D model and the rectified top view can then be extracted from the recorded video.

[0043] When capturing the dimensions of the vehicle and its surroundings, one implementation allows the device's camera to continuously record a video of the environment within an augmented reality environment. This video is simultaneously overlaid on the device's screen with information about the device's movement. Simultaneously, data from the device's accelerometer can be read and processed. The video and accelerometer data can then be used to calculate actual dimensions in the real world. Furthermore, instructions for the user can include, for example, prompts to mark specific points in the environment between which the dimensions should be measured. The user can mark these points directly on the device within the augmented reality environment.

[0044] According to one implementation variant, at least one of the following features can be recognized in a single process step from the captured images of the vehicle and its surroundings: wheel position; speedometer; odometer reading; position and angle of any trailer (if present); traffic signs; signal devices; traffic lights; tire marks; skid marks on the road; tire wear; debris; broken glass; lighting conditions; time of day; weather conditions. The automated recognition of features can provide information necessary for accident reporting, information that is often overlooked by those involved in the accident or users, but is crucial for reconstructing the event. Since the process already captures images of the vehicle and its surroundings, these images can be used directly for feature recognition, and the user does not need to perform any further input or steps.This allows for the creation of a particularly simple and reliable process.

[0045] According to one implementation variant, the features can also be detected using an object recognition method, in particular using an AI object recognition model. Such an object recognition method can reliably and automatically detect and classify a large number of different objects, thus enabling a nearly complete classification of the vehicle and its entire environment.

[0046] An AI object recognition model is an object recognition method that uses artificial intelligence (AI) to identify and classify objects in images or videos. It analyzes... Visual data is processed, and certain features and patterns are recognized and assigned specific labels or categories. For example, it can recognize vehicles, people, animals, traffic signs, and other objects, and mark their positions within the image. Such models are trained through machine learning, often using large datasets to improve the accuracy and reliability of recognition.

[0047] According to one implementation variant, the captured images of the vehicle and its surroundings can be transmitted from the device to a server, which then executes the object recognition process. Object recognition processes, in particular, usually require significant computing power and can only be executed to a limited extent on mobile devices without blocking them from further input. However, if the data is transmitted to a server for object recognition, the object recognition process can be executed in parallel with other processing steps, thus preventing any blocking of user input.

[0048] According to one implementation variant, an audio signal or a video including audio of at least one witness statement can be captured in a single process step using a user's terminal device. Witness statements can be an important piece of evidence in assessing legal and liability issues. Furthermore, witness statements from individuals other than those directly involved in the accident, such as third parties who witnessed the accident, can also be recorded. Video recordings of witness statements offer an additional layer of authenticity, as the identity of the testifying individuals can be verified.

[0049] According to one implementation variant, multiple witness statements, particularly consecutive ones, can be recorded using a user's terminal device. The following description always refers to a single witness statement. However, it should be noted that the description applies equally to multiple witness statements, to all of which the described implementation variants are applicable.

[0050] According to one implementation variant, the content of the witness statement can be transcribed from the audio signal using a speech recognition method, in particular using an AI speech recognition model. This significantly facilitates the subsequent evaluation of the accident report, as the witness statement no longer necessarily needs to be listened to. Furthermore, the conversion of the Witness statements should be converted into text form in a format that is easily exchangeable and readable by all parties involved.

[0051] In particular, the recorded audio signals can be appended to the accident report for security reasons, even if the witness statement is transcribed as text. This ensures that a verifiable and traceable original source exists in case of errors in the transcription. This creates a process for generating a particularly reliable accident report.

[0052] An AI speech recognition model is a speech recognition method that uses artificial intelligence to transcribe spoken language into text. It analyzes acoustic signals, recognizes speech patterns, and converts them into their written form. Such models use machine learning and large amounts of speech data to understand and accurately transcribe different accents, dialects, and speaking rates.

[0053] According to one embodiment, the transcribed witness statement can be summarized in a single process step using an AI language model. Recipients of the accident report no longer need to listen to or read the complete witness statements, but can directly access the generated summary, resulting in a significant time saving. A concise accident report, from which the relevant information can be quickly extracted, can thus be created using the method according to the invention.

[0054] According to one embodiment, essential information can also be extracted from the transcribed witness statement in a single process step using an AI language model. This allows the key information to be extracted from the witness statements and processed in a structured manner. This enables a quick and easy understanding of the content of the witness statements in the accident report. A concise accident report, from which the relevant information can be quickly extracted, can thus be created using the method according to the invention.

[0055] According to one implementation variant, a structured list or comparison of the essential information from all witness statements can be created in a single process step using an AI language model from several transcribed witness statements. Such a comparison allows for a quick and easy identification of the essential content of each witness statement and identifies any contradictions.

[0056] According to one implementation variant, differences or contradictions between various transcribed witness statements can be identified and written down in text form using an AI language model in one procedural step.

[0057] The AI ​​language model can be, in particular, a Large Language Model (LLM). An AI language model is a machine learning model capable of understanding and generating human-like language. For example, it can generate texts, answer questions, perform translations, create summaries, and participate in conversations. Through training with extensive datasets, it learns language patterns, grammar, and context, enabling it to provide coherent and relevant answers. Such models are used in a variety of applications, including chatbots, digital assistants, word processing, and information retrieval.

[0058] According to one implementation variant, data relating to the vehicle and / or the user can be retrieved from the end device in a single process step. For example, basic data can be pre-stored on a user's end device or in a memory accessible from the end device, and then retrieved in a single process step. This creates a particularly efficient process, as certain basic data relating to the vehicle and / or the user no longer needs to be re-entered after an accident. This saves time and prevents situations where certain data cannot be recorded because the necessary documents (e.g., driver's license, vehicle registration, insurance card, etc.) are not readily available.

[0059] According to one implementation variant, the stored data relating to the vehicle and / or the user can include at least one of the following parameters: name; driver's license; contact address; insurance number; license plate number; vehicle type; vehicle manufacturer; vehicle model; date of purchase; condition of the vehicle at the time of purchase; date of first registration; previous damage to the vehicle. This way, all relevant data required in an accident report can already be stored and does not need to be entered again.

[0060] According to one implementation variant, the stored data relating to the vehicle can include at least an image of the vehicle in its condition before the accident. This makes it particularly easier to correctly assess the extent of the damage to the vehicle that occurred as a result of the accident. This is also relevant with regard to potential insurance claims, as damages from previous events may be excluded.

[0061] According to one implementation variant, the stored data relating to the vehicle and / or the user can be stored in a blockchain. The data can thus be stored immutably and securely in a decentralized block of the blockchain. In particular, the data can also be encrypted in the blockchain to ensure data security and prevent unauthorized access by third parties.

[0062] According to one embodiment, the method for generating a report can be started by a user via an end device. This can be done, for example, by launching a program or app on the end device. After the method is started, the steps according to the invention are executed, and finally the accident report is generated.

[0063] According to another implementation variant, the process for generating a report can also be started automatically on a user's device. This can be done, for example, by the device detecting an accident or by receiving a signal to start the process from a connected device, which in turn detects an accident.

[0064] According to one implementation variant, the stored data relating to the vehicles and users involved in the incident can be exchanged between the users' devices. This eliminates the need for manual data exchange between the individual parties involved, and all users involved in the accident automatically receive the data of the other users (subject to consent, if necessary). This provides a particularly efficient method for generating a report.

[0065] According to one implementation variant, the exchange of stored data between users can be automated, with a handshake procedure between the end devices being performed at the beginning of the data exchange. Such a handshake procedure can establish a secure connection between the end devices and prepare the data transmission.

[0066] A handshake procedure is understood to be a communication method used in the initial phase of a network connection to establish a secure and reliable connection between two end devices. It comprises several steps in which the participating parties verify their identities. Confirm, negotiate encryption algorithms and keys, and ensure that both sides are ready for data transfer. This exchange lays the foundation for secure and trustworthy communication between the end devices.

[0067] According to one implementation variant, further data, such as captured images, captured dimensions or recorded witness statements, can also be exchanged between the end devices.

[0068] If the data relating to the vehicle and / or the user is stored in a blockchain, the exchange of this data can occur via the exchange of keys or references. Users' devices can retrieve the stored data from the blockchain using the corresponding reference and, if necessary, decrypt it with a matching key (provided the data is stored encrypted in the blockchain). This creates a particularly simple and efficient process in which the data exchange does not cause any delays or blocking of the devices.

[0069] According to one implementation variant, the incident report can further include at least one of the following pieces of information: the vehicle's coordinates; the vehicle's direction of travel and / or trajectory prior to the incident; a feature in the vehicle's vicinity; a transcribed witness statement; one or more summarized transcribed witness statements or information extracted from one or more transcribed witness statements; data relating to a vehicle and / or user involved in the incident. A comprehensive and detailed accident report can thus be generated in a particularly simple, efficient, and reliable manner using this method.

[0070] According to one implementation variant, a summary can be generated in a single process step from the information contained in the report using an AI language model (LLM - Large Language Model). The data contained in the report can thus be processed in a quickly understandable manner.

[0071] According to one implementation variant, the report and / or the information contained in the report can be stored in a block of a blockchain in a single process step. Storing the report data or the information contained in the report in a blockchain ensures the immutability and security of the data, as every transaction in the blockchain is cryptographically secured and stored decentrally. This protects the information from manipulation and unauthorized access. This ensures the integrity and reliability of accident reports. Furthermore, it improves the transparency and verifiability of the data history, thereby increasing credibility and traceability for all parties involved, such as insurance companies, police, authorities, courts, etc.

[0072] According to one implementation, the various AI models can communicate via vector-like structures known as embeddings. Each model processes its input data and transforms it into an embedding, a high-dimensional numerical representation that captures the essential features and semantics of the data. For effective communication, these embeddings must reside in a common or compatible embedding space, which is achieved through joint training or alignment. Once the embeddings are created, they are exchanged between the models. For example, a text-processing model transforms a sentence into an embedding and sends it to an image-processing model, which interprets it to find relevant images.The receiving model uses its neural network layers to interpret the incoming embedding based on the shared space, enabling tasks such as classification, retrieval, or generation. An example is text-to-image generation, where a language model converts a sentence into an embedding and sends it to an image generation model, which then creates an image based on the description. By using embeddings as a common language, AI models can exchange and interpret information, thereby improving the capabilities of integrated AI systems. Brief description of the characters

[0073] Preferred embodiments of the invention are described in more detail below with reference to the drawing.

[0074] Fig. 1 shows a schematic flowchart according to one implementation variant of the computer-implemented method. Ways to implement the invention

[0075] The invention is described below by way of example using the embodiment of the computer-implemented method 100 shown in Fig. 1.

[0076] The computer-implemented method 100 for creating an accident report 1 according to the illustrated embodiment is partially executed by an app on a smartphone as the user device (some of the method steps are performed by the app itself). Of course, the method can also be executed on other (mobile) devices that have a camera, a screen, and input capabilities. For the sake of simplicity, however, the invention will be described below using a smartphone as an example.

[0077] The app also communicates in the background with a server to process and / or retrieve data, with some of the process steps then being executed on the server. The app can be used by many different users via their devices to execute the process steps (or parts thereof) in the event of involvement in an accident.

[0078] Procedure step 10 is executed before an accident event 20 occurs and includes the recording of basic data 2. This may include, for example, the following data of the vehicle and / or the user: - Name, - Driving licence number and / or photo of the driving licence. - Contact address (address, email address, etc.) - Insurance number, - License plate number and / or photo of the license plate number, - Vehicle type, - Vehicle manufacturers, - Vehicle model, - Purchase date, - Condition of the vehicle at the time of purchase, - Date of first registration, - Previous damage to the vehicle, - Photos of the vehicle in its current condition.

[0079] The collection of data 2 can be carried out on the end device and, for example, by the user when setting up or initializing the associated app.

[0080] Regardless of the recording of the basic data 2, an accident event 20 occurs at a later time. Either a user involved in the accident event 20 or another user at the accident site initiates the further procedural steps using the app on their device.

[0081] After initiation, in a further step 30, the basic data of all participating vehicles and users is retrieved using a handshake function and collected for further processing. This data can be retrieved from a central storage location (a server) or exchanged directly between the users' end devices.

[0082] In the following steps, three essential process modules 40, 50, 60 are executed, which serve the central data collection and data processing.

[0083] The process module 40 for obtaining visual data 4 comprises several steps 41 , 42, 43, 44, 45, 46, 47, all of which are continuously checked by a KL model to ensure accuracy and reliability.

[0084] In the first step (41), the driving behavior, i.e., the direction of travel and / or the route taken before the accident, is determined. In this first step (41), a map in an augmented reality (AR) environment is used, where the user points the smartphone camera at the direction of travel. The user can then mark and record the desired route directly on the AR surface with their finger.

[0085] In a further step, visual data is captured in the form of a video or a series of images of the vehicle and its surroundings. In this phase, the AR system plays a central role by guiding the user in an augmented reality environment to hold the smartphone camera at the correct angle and follow the corresponding steps. The system provides real-time feedback to ensure the entered data is correct and prompts the user for corrections if necessary.

[0086] Once the visual data (4) has been captured in step 42, it is processed in step 43 using a 3D photogrammetry method to create a detailed 3D model of the damaged vehicle. This conversion of 2D images or videos into a 3D model is essential for an accurate assessment of the vehicle's condition. Furthermore, the AR guide continues to assist the user in capturing images of the vehicle's surroundings, if needed.

[0087] In a subsequent step (44), an AI image recognition model identifies important features in the visual data, such as skid marks, skid marks on the ground, and debris, and highlights these features in the images. These features can be crucial for the accurate reconstruction of the accident.

[0088] After capturing the general scene around the vehicle, the vehicle's absolute final position must be determined. In this step (45), an AR measuring device is used to precisely measure the area, including its perpendicular and diagonal dimensions.

[0089] These measurements are crucial for the subsequent rectification process 46, in which the exact final resting position of the vehicle is determined from a top view. Combining these measurements with the previously recorded photographs or videos ensures comprehensive documentation of the scene, including a rectified orthographic image (rectified top view) showing the final resting positions.

[0090] Finally, in a further step (47), the AI ​​guides the user, if necessary, to take additional images to expand the dataset. These photos contain specific details such as the positions of the wheels, the speedometer reading, the position and angle of any trailers involved (if any), and any traffic signs in the vicinity. AI-based image recognition is then used to find relevant elements in the images.

[0091] After completing all steps 41 to 47 of module 40, visual data 4 are obtained, comprising the following information: the recorded direction of travel or the recorded route before the accident, captured images or videos of the vehicle and its surroundings, a 3D model of the vehicle, a rectified top view of the vehicle and its surroundings, features recognized in the images, and supplementary images. This visual data 4 is then transferred to the central AI unit 70 for further processing after completion of module 40.

[0092] Steps 43 and 44 can be executed on a server according to one implementation variant, after the visual data captured on the smartphone in steps 41 and 42 has been transferred to the server. In step 45, visual data can again be captured on the smartphone, which is then processed on the server in step 46. Step 47 can then be executed on the smartphone.

[0093] In another version, all steps 41 to 47 can also be performed on the smartphone.

[0094] The procedural module 50 for the recording and processing of witness statement data 5 outlines the corresponding steps for comprehensively recording, transcribing, and analyzing witness statements. This module 50 is of central importance for ensuring accurate documentation and identifying potential conflicts within the statements of different witnesses.

[0095] In a first step, witness statements are recorded either through audio recordings or video recordings. Video recordings offer the additional possibility of authentication and visual context compared to audio recordings alone.

[0096] Once the statements have been recorded, the next step is to transcribe the speech data. This transcription is performed by an AI speech recognition model specifically designed for converting speech to text (speech-to-text model).

[0097] Finally, in step 53, the transcribed statements are analyzed and examined by a large AI language model (LLM). The LLM plays a crucial role in this phase by extracting the most important parts, passages, or information from the transcribed statements. It efficiently scans the text to highlight key information, helping to filter out the most important details of each statement. Furthermore, by comparing and contrasting the various transcribed witness statements, the LLM can also identify inconsistencies in the statements of different witnesses.

[0098] The recorded, transcribed and summarized or analyzed witness statements are transferred as witness statement data 5 to the central clerical unit 70 for further processing after completion of module 50.

[0099] Steps 52 and 53 can be performed on a server according to one implementation variant, after the audio or video data of the witness statements recorded on the smartphone in step 51 have been transferred to the server.

[0100] In another version, steps 52 and 53 can also be performed on the smartphone.

[0101] According to Module 60, a data acquisition system is provided, which includes steps 61 to 63 for retrieving and processing data from external sources.

[0102] This can be achieved, for example, by connecting to multiple external application programming interfaces (APIs). This module 60 ensures that comprehensive and up-to-date information is collected to support various applications and decision-making processes.

[0103] In a first step, 61, important data from the current environment, such as current weather and date information, are retrieved for the accident report.

[0104] In addition to weather data, precise location information is also recorded in a further step. This is achieved by using or recording GPS, GLONASS, or Galileo coordinates, which enable accurate geolocation (e.g., via a corresponding API of the smartphone to use the GPS receiver).

[0105] Finally, in step 63, local traffic conditions, restrictions, or traffic rules are recorded as an important component of the data retrieved by module 30. This can be done, for example, by accessing traffic APIs, which may contain information about local traffic restrictions, construction sites, speed limits, etc.

[0106] The collected external data 6 are in turn made available to the central AI unit 70 for further processing after completion of steps 61 to 63.

[0107] Steps 61 and 63 can be executed on a server according to one implementation variant, optionally using the location information collected on the smartphone in step 62, with the server being programmed to retrieve data from the relevant APIs.

[0108] In another version, steps 61 and 63 can also be performed on the smartphone.

[0109] The central Kl unit 70 connects and integrates various data input streams into a comprehensive accident report. This includes the data recorded and determined in the previously described modules 40, 50, and 60: basic data 2 and 3, visual data 4, witness statement data 5, and external data 6.

[0110] In a first step, the central AI unit (71) synthesizes various data sources, including voice recordings, video footage, weather conditions, precise locations, and traffic details. Communication between the various data input streams or modules (40, 50, 60) and the central AI unit (70) is based on a communication vector format that enables seamless data exchange and integration.

[0111] In a second step 72, the data are analyzed contextually and analyzed for potential conflicts or contradictions to ensure that the final report 1 is correct and reliable.

[0112] Finally, in step 73, a comprehensive accident report 1 is created from all the data using a Classroom Language Model (LLM), which is the final result of combining the collected data sources and is divided into several sections, including: 1. Basic data: This section contains the user's basic data as well as the basic data of other vehicles involved in the incident, if a handshake was performed; 2. Witness statements: This section summarizes the various witness statements, including a list of possible conflicts of testimony; 3. Orthographic map: This section contains a detailed orthographic map of the location, including the direction of approach of the participants, the planned route, the final position of the vehicles, the location of tracks and debris, and the traffic signs; 4. Photographs: This section contains a structured list of all photographs taken; 5. 3D Photogrammetric Comparison: This section contains a photogrammetric 3D comparison of the vehicles in a state before and after the accident; 6. Raw data elements: This appendix contains a list of the raw data elements; 7. Blockchain Report: This appendix documents when the data was entered into the report.

[0113] Each component of Report 1 is intended to provide a comprehensive overview of the incident so that the recipients (insurance companies, police, authorities, courts, etc.) can better understand the events that led to the accident and make informed decisions.

Claims

Patent claims 1. Computer-implemented method (100) for producing a report (1 ) on a road traffic accident involving at least one vehicle, comprising the steps: (42) Capture of a series of images of the vehicle and its surroundings by a user using an end device, wherein instructions for recording the images are displayed to the user on the end device during the capture of the series of images, Creating (43) a 3d model of the vehicle from the series of captured images, Acquisition (45) of dimensions of the vehicle and its surroundings, creation (46) of a rectified top view of the vehicle and its surroundings from the series of captured images and the captured dimensions, and creation (73) of a report of the event using the 3d model of the vehicle and the rectified top view of the vehicle and its surroundings.

2. Computer-implemented method according to claim 1, characterized in that when creating (46) a rectified top view of the vehicle and its surroundings, a final rest position of the vehicle in its surroundings is determined.

3. Computer-implemented method according to claim 1 or 2, characterized in that in a step (62) the coordinates of the vehicle are recorded and that when creating (46) the rectified top view of the vehicle and its surroundings the coordinates of the vehicle are taken into account.

4. Computer-implemented method according to one of claims 1 to 3, characterized in that in a step (41) the direction of travel and / or the route of travel of the vehicle is detected before the event, wherein the detection (41) of the direction of travel and / or the route is carried out in particular by the user by means of the terminal device, wherein instructions are displayed to the user on the terminal device during the detection (41) of the direction of travel and / or the route.

5. Computer-implemented method according to one of claims 1 to 4, characterized in that the acquisition (45) of dimensions of the vehicle and its surroundings is at least partially carried out by the user by means of the terminal device, wherein Instructions for recording the dimensions are displayed to the user during the recording (45) of the dimensions on the terminal device.

6. Computer-implemented method according to one of claims 1 to 5, characterized in that the acquisition (45) of dimensions of the vehicle and its environment comprises retrieving dimensions of the vehicle from a database.

7. Computer-implemented method according to one of claims 1 to 6, characterized in that the acquisition (42) of a series of images of the vehicle and its surroundings and / or the acquisition (45) of dimensions of the vehicle and its surroundings and / or the acquisition (41) of the direction of travel and / or the route traveled takes place at the terminal device in an augmented reality environment.

8. Computer-implemented method according to one of claims 1 to 7, characterized in that in a step (44) at least one of the following features is recognized in the captured images of the vehicle and its surroundings: position of the wheels; speedometer; mileage; position and angle of a trailer (if present); traffic signs; signal transmitters; traffic lights; tracks; skid marks on the road; tire abrasion; debris; broken glass; lighting conditions; time of day; weather conditions.

9. Computer-implemented method according to claim 8, characterized in that the recognition (44) of the features is carried out using an object recognition method, in particular using an AI object recognition model.

10. Computer-implemented method according to claim 9, characterized in that the captured images of the vehicle and its surroundings are transferred to a server which performs the object recognition method.

11. Computer-implemented method according to one of claims 1 to 10, characterized in that in a step (51) an audio signal or a video including an audio signal of a witness statement is captured by means of a user's terminal device and that in a further step (52) in particular the content of the witness statement is transcribed from the audio signal by means of a speech recognition method, in particular using an AI speech recognition model.

12. Computer-implemented method according to claim 11, characterized in that in a further step (53) the transcribed witness statement is summarized using an AI language model, or that from the Essential information is extracted from the transcribed witness statement using an Al language model.

13. Computer-implemented method according to one of claims 1 to 12, characterized in that in a step (10) stored data (2) relating to the vehicle and / or the user are retrieved from the terminal device.

14. Computer-implemented method according to claim 13, characterized in that the stored data (2) relating to the vehicle and / or the user includes at least one of the following parameters: name; driver's license; contact address; insurance number; license plate number; vehicle type; vehicle manufacturer; vehicle model; date of purchase; condition of the vehicle at the time of purchase; date of first registration; previous damage to the vehicle.

15. Computer-implemented method according to claim 13 or 14, characterized in that the stored data (2) relating to the vehicle includes at least one image of the vehicle in the state before the event.

16. Computer-implemented method according to one of claims 1 to 15, characterized in that the method for creating a report is started by a user via an end device.

17. Computer-implemented method according to one of claims 1 to 16, characterized in that in a step (30) the stored data relating to the vehicles and users involved in the event are exchanged between the users' terminal devices.

18. Computer-implemented method according to one of claims 1 to 17, characterized in that the exchange (30) of the stored data between the users takes place automatically after a user has started the method.

19. Computer-implemented method according to any one of claims 1 to 18, characterized in that the report (1) of the event further comprises at least one of the following information: the coordinates of the vehicle; the direction of travel and / or the path of travel of the vehicle before the event; a feature in the vicinity of the vehicle; a transcribed witness statement; a summarized transcribed witness statement or information extracted from a transcribed witness statement; data relating to a vehicle and / or a user involved in the event.

20. Computer-implemented method according to claim 19, characterized in that a summary is generated from the information contained in the report (1) using an AI language model (Large Language Model).

21. Computer-implemented method according to one of claims 1 to 20, characterized in that in one step the report and / or the information contained in the report are stored in a block of a blockchain.

22. System for generating a report on an event involving at least one vehicle, comprising at least one terminal device and one server, wherein the terminal device is programmed to execute the method according to any one of claims 1 to 21.

23. Computer program product comprising instructions which, when loaded and run on an end device, performs the corresponding steps according to one of claims 1 to 21.

Images