A damage assessment system and method

Abstract

The invention relates to a method performed by a control unit (12) in a damage assessment system (10) comprising a camera (11) for taking at least one image of a vehicle for which damage assessment is desired and said control unit (12) for performing damage assessment for said vehicle according to said image taken by said camera (11), and for generating a damage report according to said damage assessment for viewing using a user interface (13).

Description

[0001] A DAMAGE ASSESSMENT SYSTEM AND METHOD

[0002] TECHNICAL FIELD

[0003] The invention relates to a damage assessment system and method enabling damage assessment of a vehicle for which damage assessment is desired.

[0004] PRIOR ART

[0005] Road vehicles are used to transport people or loads from one place to another. This generally includes cars, trucks, buses, motorcycles, and other motor vehicles. Road vehicles play an important role in meeting the transportation needs of modern societies. With technological developments, road vehicles are constantly being developed in terms of environmentally friendly features, more effective fuel use, and safety systems.

[0006] Traffic accidents may occur depending on the use of road vehicles. Traffic accidents are generally the result of a combination of a number of complex factors. These factors include various factors such as driver errors, road conditions, speed, vehicle technical condition, non-compliance with traffic signs and rules. For example, excessive speed, especially in bad weather conditions, increases the braking distance and increases the risk of accidents. Driver inattention or fatigue increases reaction times and reduces decision-making ability. Technical malfunctions in vehicles, for example brake system or tire problems, pave the way for accidents.

[0007] Various damage assessment systems are used in the present art to ensure that the damage to vehicles after a traffic accident is assessed. These systems enable the effects of the accident to be determined and the extent of the damage to be assessed, often using sensors, cameras, and various other technological components. Sensors measure the forces in the event of a collision and determine the severity of the impact to which the vehicle is subjected to. Cameras and image processing technology enable the assessment of damage to the vehicle's exterior. However, damage assessment systems have some shortcomings. Firstly, there is the problem that not all damages are identified by these systems. In particular, hidden damages, such as internal mechanical or electronic problems, cannot always be accurately assessed with such systems. Furthermore, in some cases, said systems are insufficient in determining the exact extent of the damage or may produce erroneous results. In addition, for damage assessment systems to be widely deployed, a large part of the vehicle fleet must have hardware and software integrated with this technology. This can create a situation where damage assessment becomes difficult for vehicles of older models or without such systems. Furthermore, it is required to wait for a long time in order to obtain results regarding the vehicle in the systems used in the present art. This situation results in a loss of time for the users and the individuals performing the work.

[0008] As a result, all the above-mentioned problems have made it imperative to make an innovation in the relevant technical field.

[0009] SUMMARY OF THE INVENTION

[0010] The present invention relates to a damage assessment system for eliminating the above-mentioned disadvantages and bringing new advantages to the relevant technical field.

[0011] An object of the invention is to introduce a damage assessment system and method enabling damage assessment on damaged vehicles to be carried out quickly and with increased accuracy.

[0012] Another object of the invention is to introduce a damage assessment system and method enabling calculation of the repair cost required for the repair of damage assessment in damaged vehicles and presentation thereof to the user.

[0013] Another object of the invention is to introduce a damage assessment system and method enabling the assessment of damaged sub-parts affected by the damaged part in the damaged vehicle. In order to accomplish all the objects mentioned above and to be derived from the detailed description below, the present invention is a method performed by a control unit in a damage assessment system comprising a camera for taking at least one image of a vehicle for which damage assessment is desired and said control unit for performing damage assessment for said vehicle according to said image taken by said camera, and for generating a damage report according to said damage assessment for viewing using a user interface. Accordingly, it comprises the process steps of taking at least one image from the camera, inputting the vehicle image taken from the camera to a first mathematical model that has been trained with vehicle images and the orientation information of the vehicle parts in these vehicle images and that, upon receiving a vehicle image as input, outputs the orientation information corresponding to this vehicle image and outputting an orientation information, inputting the vehicle image taken from the camera to a second mathematical model that has been trained with vehicle type images and that, upon receiving a vehicle type image as input, outputs the type information of the vehicle corresponding to this vehicle type image, and outputting the vehicle type information, inputting the vehicle image taken from the camera to a third mathematical model that has been trained with vehicle images and information on vehicle parts in said vehicle images and that, upon receiving a vehicle image as input, outputs at least one vehicle part information corresponding to this vehicle image, and outputting at least one vehicle part information, inputting the vehicle image taken from the camera to a fourth mathematical model that has been trained with vehicle damage images and information on the type and pixel location of said damage and that, upon receiving a vehicle image as input, outputs damage information including the type and pixel locations of damage in this vehicle image, and outputting a damage information, obtaining data containing information on the damaged part of the vehicle and the area affected by the damage to the vehicle by applying the detected orientation information, vehicle type information, vehicle part information, and damage information to a first mathematical equation, inputting the vehicle image taken from the camera to a fifth mathematical model that has been trained with a vehicle damage image and depth dimension information of this vehicle part and that, upon receiving a vehicle damage image as input, outputs the depth information of the vehicle damage corresponding to this vehicle damage image, and outputting the depth information of the vehicle damage, inputting the vehicle image taken from the camera to a sixth mathematical model that has been trained with vehicle type information, information on damaged part of the vehicle, information on the area affected by the damage on the vehicle, and depth information and that, upon receiving a vehicle image as input, outputs the information on the lower mechanical part that is under said damaged vehicle part and is not visible in the two-dimensional image, and outputting the information on the lower mechanical parts that are not visible under the damaged vehicle part, generating price information for the assessed damaged vehicle parts by accessing the information in a memory where the workload to be incurred for the repair of damaged vehicle parts and the cost information to be incurred for this workload are stored, generating a report to be forwarded to the user interface. In this way, the damaged part in a damaged vehicle and the sub-parts affecting this damaged part are detected, a price is determined for the detected parts and presented to the user. This reduces the time loss and workload of the user.

[0014] A possible embodiment of the invention is characterized in that the first mathematical model is a VGG16 artificial intelligence model.

[0015] Another possible embodiment of the invention is characterized in that the second mathematical model is a VGG16 artificial intelligence model.

[0016] Another possible embodiment of the invention is characterized in that the third mathematical model is a UNET artificial intelligence model.

[0017] Another possible embodiment of the invention is characterized in that the fourth mathematical model is a ResNet18 artificial intelligence model.

[0018] Another possible embodiment of the invention is characterized in that the fifth mathematical model is a ResNet18 artificial intelligence model.

[0019] Another possible embodiment of the invention is characterized in that the sixth mathematical model is a Gradient Boosting and random forest artificial intelligence model.

[0020] Another possible embodiment of the invention is characterized in that it comprises the step of inputting the images taken from the camera to a seventh mathematical model that has been trained with vehicle images, images of documents used for vehicles, and images of the environment in which a vehicle may have an accident, and that, upon receiving the images taken from the camera as input, outputs an image information including the information indicating that the corresponding image is related to at least one of the vehicle image, image of documents used for vehicles, and image of the environment in which a vehicle may have an accident, and identifying the image according to the image information obtained from said seventh mathematical model.

[0021] Another possible embodiment of the invention is characterized in that the seventh mathematical model is a ResNet18 artificial intelligence model.

[0022] Another possible embodiment of the invention is characterized in that it comprises the step of inputting the vehicle image taken from the camera to an eighth mathematical model that has been trained with vehicle interior images and vehicle exterior images and that, upon receiving a vehicle image as input, outputs a vehicle image information indicating that this vehicle image is the vehicle interior image or the vehicle exterior image, and outputting a vehicle image information.

[0023] Another possible embodiment of the invention is characterized in that the eighth mathematical model is a VGG16 artificial intelligence model.

[0024] Another possible embodiment of the invention is characterized in that it comprises the step of inputting the vehicle image taken from the camera to a ninth mathematical model that has been trained with distance information comprising the distance of the location from which an image is taken to the objects in the image, and that, upon receiving a vehicle image as an input, outputs a distance information regarding the distance of the location from which this vehicle image is taken to the vehicle, and outputting a distance information.

[0025] Another possible embodiment of the invention is characterized in that the ninth mathematical model is a VGG16 artificial intelligence model.

[0026] Another possible embodiment of the invention is characterized in that it comprises the step of inputting the vehicle image taken from the camera to a tenth mathematical model that has been trained with the airbag location images and that, upon receiving a vehicle image as input, outputs the airbag usage information corresponding to this vehicle image, and receiving the airbag usage information. Another possible embodiment of the invention is characterized in that the tenth mathematical model is a YOLOvS artificial intelligence model.

[0027] Another possible embodiment of the invention is characterized in that it comprises the step of inputting a vehicle image taken from the camera to an eleventh mathematical model that has been trained with heavily damaged vehicle images, and that, upon receiving a vehicle image as input, outputs a heavy damage information indicating that the vehicle is heavily damaged, and outputting the heavy damage information.

[0028] Another possible embodiment of the invention is characterized in that the eleventh mathematical model is a ResNet18 artificial intelligence model.

[0029] Another possible embodiment of the invention is characterized in that it comprises the step of inputting the vehicle image taken from the camera to a twelfth mathematical model that has been trained with the vehicle image and the scratch and dent information in said vehicle images and that, upon receiving a vehicle image as input, outputs a first vehicle damage information including the information on scratch and dents on said vehicle, and outputting the first vehicle damage information. Thus, it is possible to detect the parts of the vehicle with small damage size.

[0030] Another possible embodiment of the invention is characterized in that the twelfth mathematical model is a UNET artificial intelligence model.

[0031] Another possible embodiment of the invention is characterized in that it comprises the step of inputting the vehicle image taken from the camera to a thirteenth mathematical model that has been trained with the vehicle image and the broken vehicle part, damaged tire, and damaged glass information in said vehicle images, and that, upon receiving a vehicle image as input, outputs a second vehicle damage information including the information on broken vehicle part, damaged tire, and damaged glass in said vehicle, and outputting the second vehicle damage information. Thus, it is possible to detect the parts of the vehicle with large damage size.

[0032] Another possible embodiment of the invention is characterized in that the thirteenth mathematical model is a YOLOv8 artificial intelligence model. Another possible embodiment of the invention is characterized in that, after the step of generating price information for the assessed damaged vehicle parts by accessing the information in a memory where the workload to be incurred for the repair of damaged vehicle parts and the cost information to be incurred for this workload are stored, it comprises the step of generating a total price information for the repair of damaged vehicle parts, and determining a damage type according to this total price information.

[0033] BRIEF DESCRIPTION OF THE DRAWING

[0034] Figure 1 shows a representative view of the working scenario of a damage assessment system.

[0035] DETAILED DESCRIPTION OF THE INVENTION

[0036] In this detailed description, the subject of the invention is explained by way of example only for a better understanding of the subject, which shall not create any limiting effect.

[0037] The invention relates to a damage assessment system (10) and method enabling damage assessment of a vehicle for which damage assessment is desired. Referring to Figure 1 ; the damage assessment system (10) comprises a camera (11 ) to enable an image of a vehicle to be taken for which damage assessment is desired. Said camera (11) is provided in an image capturing device. In a possible embodiment of the invention, the image capturing device can be a mobile device that allows an image to be taken such as a phone, a computer, etc. The damage assessment system (10) includes a control unit (12) for performing damage assessment for said vehicle according to said image taken by said camera (11), and for generating a damage report according to said damage assessment for viewing using a user interface (13). In a possible embodiment of the invention, the user interface (13) may be a mobile application, a web, etc., which is accessed via a mobile device.

[0038] The control unit (12) enables the assessment of damage to the damaged vehicle by reading predefined computer-based software commands. A possible embodiment of the invention comprises a memory unit in which computer-based software commands are stored. The control unit (12) is configured to allow data to be written to and read from the memory unit.

[0039] The control unit (12) enables taking at least one image from the camera (11). The control unit (12) enables inputting the vehicle image taken from the camera (11) to a first mathematical model that has been trained with vehicle images and the orientation information of the vehicle parts in these vehicle images and that, upon receiving a vehicle image as input, outputs the orientation information corresponding to this vehicle image and outputting an orientation information. Therefore, it is possible to detect from which side of the vehicle the image from camera (11) was taken. This ensures that the damaged part is detected accurately in the assessment of the damaged parts of the vehicle. For example, if the image contains the right side image of the vehicle and the right fender herein is damaged, it is ensured that the damage assessment is made according to the right fender. For this reason, it is necessary to assess the orientation information of the vehicle correctly. In a possible embodiment of the invention, it is preferred to use the VGG16 artificial intelligence model as said first mathematical model.

[0040] The control unit (12) enables inputting the vehicle image taken from the camera (11) to a second mathematical model that has been trained with vehicle type images and that, upon receiving a vehicle type image as input, outputs the type information of the vehicle corresponding to this vehicle type image, and outputting the vehicle type information. In a possible embodiment of the invention, it is preferred to use the VGG16 artificial intelligence model as said second mathematical model. In a possible embodiment of the invention, it is preferred to use the ResNet18 artificial intelligence model as said second mathematical model. Thus, it is possible to identify the brand and model of the vehicle in the image taken from the camera (11). Each vehicle has a brand and model. Vehicles may contain different vehicle parts and locations thereof depending on the brand and model. Therefore, detecting the brand and model ensures that the damaged part is correctly detected.

[0041] The control unit (12) enables inputting the vehicle image taken from the camera (11) to a third mathematical model that has been trained with vehicle images and information on vehicle parts in said vehicle images and that, upon receiving a vehicle image as input, outputs vehicle part information corresponding to this vehicle image, and outputting vehicle part information. In a possible embodiment of the invention, it is preferred to use the UNET artificial intelligence model as said second mathematical model. In this way, it is possible to detect which vehicle parts are visible in what proportion in the image taken from the camera (11).

[0042] The control unit (12) enables inputting the vehicle image taken from the camera to a fourth mathematical model that has been trained with vehicle damage images and information on the type and pixel location of said damage and that, upon receiving a vehicle image as input, outputs damage information including the type and pixel locations of damage in this vehicle image, and outputting a damage information. In a possible embodiment of the invention, it is preferred to use the ResNet18 artificial intelligence model as said second mathematical model. In this way, the damage type of the vehicle parts in the image taken from the camera and the information on pixel location affected by the damage according to said damage type are detected. Thus, it is possible to assess the extent of the damage to the vehicle part.

[0043] The control unit (12) enables obtaining data containing information on the damaged part of the vehicle and the area affected by the damage to the vehicle by applying the detected orientation information, vehicle type information, vehicle part information, and damage information to a first mathematical equation. It is ensured that said first mathematical equation is stored in a memory. The control unit (12) provides access to the first mathematical equation in said memory, enabling the calculation of the information on the damaged part of the vehicle and the area affected by the damage to the vehicle. For example; in the image of a vehicle taken from a rear angle in which the taillight is damaged, it is possible to first detect from which orientation (direction) the image was taken. In this case, it is detected that it was taken from the rear angle. Afterwards, the vehicle type is determined. As a vehicle type, the vehicle is detected as, for example, Renault Clio. Afterwards, the vehicle parts in the image are detected as taillights and luggage. According to this information, the vehicle in the image, the information on the damaged parts of the vehicle and the area affected by the damage to the vehicle are calculated. Thus, it is ensured that the taillight of the vehicle is damaged and the extent of this damage is detected.

[0044] The control unit (12) enables inputting the vehicle image taken from the camera (11) to a fifth mathematical model that has been trained with a vehicle damage image and depth dimension information of this vehicle part and that, upon receiving a vehicle damage image as input, outputs the depth information of the vehicle damage corresponding to this vehicle damage image, and outputting the depth information of the vehicle damage. In a possible embodiment of the invention, it is preferred to use the ResNet18 artificial intelligence model as the fifth mathematical model.

[0045] The control unit (12) enables inputting the vehicle image taken from the camera (11) to a sixth mathematical model that has been trained with vehicle type information, information on damaged part of the vehicle, information on the area affected by the damage on the vehicle, and depth information and that, upon receiving a vehicle image as input, outputs the information on the lower mechanical part that is under said damaged vehicle part and is not visible in the two-dimensional image, and outputting the information on the lower mechanical parts that are not visible under the damaged vehicle part. In a possible embodiment of the invention, it is preferred to use the Gradient Boosting and Random Forest artificial intelligence model as the sixth mathematical model. Thus, it is possible to determine an estimated damage state for the damaged parts that are under the damaged part in the image. For example; it is detected that the radiator may be damaged due to a severe dent in the front bumper, or the door mechanics may be damaged due to a deep dent or fracture in the door.

[0046] The control unit (12) enables generating price information for the assessed damaged vehicle parts by accessing the information in a memory where the workload to be incurred for the repair of damaged vehicle parts and the cost information to be incurred for this workload are stored. During the repair of vehicle parts according to the brand and model, the cost to be incurred for each part is determined and stored in memory. The information in the memory includes which methods will be used for the repair of the vehicle part, which man / hour, labor, electricity, paint, disassembly and replacement process will be applied, and the cost of the operations.

[0047] The control unit (12) enables inputting the images taken from the camera (11) to a seventh mathematical model that has been trained with vehicle images, images of documents used for vehicles, and images of the environment in which a vehicle may have an accident, and that, upon receiving the images taken from the camera (11) as input, outputs an image information including the information indicating that the corresponding image is related to at least one of the vehicle image, image of documents used for vehicles, and image of the environment in which a vehicle may have an accident, and identifying the image according to the image information obtained from said seventh mathematical model. Thus, it is possible to detect information on which of the images taken from the camera (11) belongs to the vehicle images, the images of the documents used for the vehicle and the images of the environment in which the vehicle may have an accident. In a possible embodiment of the invention, it is preferred to use the ResNet18 artificial intelligence model as the seventh mathematical model. This process is carried out after the step of taking the images from the camera (11).

[0048] The control unit (12) enables inputting the vehicle image taken from the camera (11) to an eighth mathematical model that has been trained with vehicle interior images and vehicle exterior images and that, upon receiving a vehicle image as input, outputs a vehicle image information indicating that this vehicle image is the vehicle interior image or the vehicle exterior image, and outputting a vehicle image information. In a possible embodiment of the invention, it is preferred to use the VGG16 artificial intelligence model as said eighth mathematical model. This allows the image to be classified as belonging to an interior or an exterior of the vehicle. If it is detected that the image belongs interior of the vehicle; the control unit enables inputting the vehicle image taken from the camera (11) to a tenth mathematical model that has been trained with the airbag location images and that, upon receiving a vehicle image as input, outputs the airbag usage information corresponding to this vehicle image, and receiving the airbag usage information. In a possible embodiment of the invention, it is preferred to use a YOLOv8 artificial intelligence model as the tenth mathematical model. Thus, it is possible to detect the usage status of the airbag in the vehicle. Because the control unit (12) is provided to output the heavy damage information directly without applying other identification processes for vehicles using airbags. The control unit (12) enables determining a market value according to the information kept in a database according to the brand and model of the vehicle in which the airbag is used. As is well known in the art, the market value includes the market price of the vehicle according to the brand and model. The market value of heavily damaged vehicles is calculated and this value is paid to the vehicle user by insurance or car insurance companies.

[0049] The control unit enables inputting a vehicle image taken from the camera to an eleventh mathematical model that has been trained with heavily damaged vehicle images, and that, upon receiving a vehicle image as input, outputs a heavy damage information indicating that the vehicle is heavily damaged, and outputting the heavy damage information. In a possible embodiment of the invention, it is preferred to use a ResNet18 artificial intelligence model as the eleventh mathematical model. Thus, it is possible to detect that the vehicle is heavily damaged from the vehicle image. In this case, it is possible to determine the market value of the vehicle.

[0050] The control unit (12) enables inputting the vehicle image taken from the camera to a ninth mathematical model that has been trained with distance information comprising the distance of the location from which an image is taken to the objects in the image, and that, upon receiving a vehicle image as an input, outputs a distance information regarding the distance of the location from which this vehicle image is taken to the vehicle, and outputting a distance information. In a possible embodiment of the invention, it is preferred to use a VGG16 artificial intelligence model as the ninth mathematical model.

[0051] The control unit (12) enables inputting the vehicle image taken from the camera to a twelfth mathematical model that has been trained with the vehicle image and the scratch and dent information in said vehicle images and that, upon receiving a vehicle image as input, outputs a first vehicle damage information including the information on scratch and dents on said vehicle, and outputting the first vehicle damage information. In a possible embodiment of the invention, it is preferred to use the a UNET artificial intelligence model as the twelfth mathematical model. Thus, it is possible to detect the slightly damaged parts in the vehicle.

[0052] The control unit (12) enables inputting the vehicle image taken from the camera to a thirteenth mathematical model that has been trained with the vehicle image and the broken vehicle part, damaged tire, and damaged glass information in said vehicle images, and that, upon receiving a vehicle image as input, outputs a second vehicle damage information including the information on broken vehicle part, damaged tire, and damaged glass in said vehicle, and outputting the second vehicle damage information. In a possible embodiment of the invention, it is preferred to use a YOLOv8 artificial intelligence model as said thirteenth mathematical model. Thus, it is possible to detect the heavily damaged parts in the vehicle. If the control unit (12) detects that the image belongs to the interior of the vehicle, it also provides panel detection. The control unit (12) enables the detection of the mileage information by detecting the panel from the image. The control unit (12) enables inputting the vehicle image taken from the camera (11) to a fourteenth mathematical model that has been trained with panel images of vehicles and that, upon receiving a vehicle image as input, outputs a panel information, and receiving the panel information. Panel information is used to detect the mileage information of the vehicle. In a possible embodiment of the invention, it is preferred to use the a VGG artificial intelligence model as the fourteenth mathematical model. In an alternative embodiment of the invention, it is preferred to use a ResNet artificial intelligence model as the fourteenth mathematical model. The control unit (12), when the image taken from the camera (11) is input to the Object Detection Y0I0V8 artificial intelligence model that has been trained with panel information, enables the detection of the area in which the mileage information on the panel is located. The control unit (12) enables outputting a mileage information in response to inputting the vehicle image to a fifteenth mathematical model that has been trained with the information on the area in which the mileage information on the panel is located. In a possible embodiment of the invention, it is preferred to use the Tesseract artificial intelligence model as the fifteenth artificial intelligence model. In an alternative embodiment of the invention, it is preferred to use the Azure OCR On-Prem artificial intelligence model as the fifteenth artificial intelligence model.

[0053] If the control unit (12) detects that the image belongs to the exterior of the vehicle, it enables the detection of the orientation information of the vehicle. If the control unit (12) detects that the image belongs to the exterior of the vehicle, it provides the color detection of the vehicle. The control unit (12), when the vehicle image is input to a sixteenth artificial intelligence model that has been trained with color information, enables outputting the color information of the vehicle. In a possible embodiment of the invention, it is preferred to use the VGG artificial intelligence model as the sixteenth artificial intelligence model. In an alternative embodiment of the invention, it is preferred to use the ResNet artificial intelligence model as the sixteenth artificial intelligence model. If the control unit (12) detects that the image belongs to the exterior of the vehicle, it enables the detection of the vehicle type. If the control unit (12) detects that the image belongs to the exterior of the vehicle, it enables the detection of the license plate of the vehicle. The control unit (12) enables inputting the vehicle image to an Object Detection Y0I0V8 artificial intelligence model that has been trained with license plate area images, and outputting the license plate area information. The control unit (12) enables outputting a license plate information in response to inputting the vehicle image to a seventeenth mathematical model that has been trained with the license plate are information. In a possible embodiment of the invention, it is preferred to use the Tesseract artificial intelligence model as the seventeenth artificial intelligence model. In an alternative embodiment of the invention, it is preferred to use the Azure OCR On-Prem artificial intelligence model as the seventeenth artificial intelligence model.

[0054] The control unit (12) enables checking the values obtained from the image with the license data. The control unit (12) enables the comparison of information such as vehicle brand, model, color, year, address in the license data with the values coming from the EXIF values in artificial intelligence models and images. It is ensured that the license plate information can be checked in case of abuse. Furthermore, it is possible to detect the incoming image has been taken before. For these operations, it is preferred to use the pre-trained efficientnet liteO model. The control unit (12), with the efficientnet liteO model, enables the image to be converted into hash values (vectors of 128) and the control is ensured through identical or similar hash values in the database.

[0055] The control unit (12) enables the detection of the repair status of damaged vehicle parts. For example, repair suggestions such as replacing the rear fender, repairing the rear headlight, painting the right door, etc. are provided.

[0056] The control unit (12), after the step of generating price information for the assessed damaged vehicle parts by accessing the information in a memory where the workload to be incurred for the repair of damaged vehicle parts and the cost information to be incurred for this workload are stored, enables generating a total price information for the repair of damaged vehicle parts, and determining a damage type according to this total price information.

[0057] The control unit (12) enables the classification of the damage according to price information. The control unit (12) ensures that the total damage price is classified as mini repair if it complies with the mini repair standard, as module damage if the total damage price is up to 30.000 TRY (this figure can be updated by sector), as normal damage if the total damage price is above the module damage price, and as heavy damage if the price ratio between the market value of the automobile and the total damage price is above 30-35%. Also, there are different methods and rules to detect that there is heavy damage. For example, if it is detected that the airbag is used in the vehicle interior images of the vehicle, it can be evaluated that the vehicle is heavily damaged.

[0058] The control unit (12) enables the generation of a report to be transmitted to the user interface (13). The said report includes an image showing the damaged part of the vehicle on a representative vehicle model, images taken from the camera (11) and the vehicle type, vehicle mileage, model year, vehicle color, vehicle orientation, sector market value, angle information, distance information, damaged part information, subdamaged part information, work required to replace the damaged parts, bodywork, dismantling, painting, electrical, labor hours, total price information obtained from these images.

[0059] An example working scenario of the invention is described as follows.

[0060] The control unit (12) enables taking vehicle images from the camera (11). The control unit (12) primarily ensures that the images are classified as vehicle images, document images, and accident environment images. The control unit (12) enables the detection of the vehicle image as belonging to the interior or exterior of the vehicle. If the control unit (12) detects that the image belongs to the interior of the vehicle, it provides panel detection. The control unit (12) enables the detection of the mileage information of the vehicle from the panel detection. The control unit (12), if it is detected that the airbag is used from the vehicle interior images, enables the generation of the information that the vehicle is heavily damaged. If it is detected that the vehicle is heavily damaged, the market value is calculated. If the control unit (12) detects that the image taken from the camera (11) belongs to the exterior of the vehicle, it provides the detection of the distance of the vehicle to the camera (11). The control unit (12) detects the distance of the camera (11 ) and evaluates whether the shot is close-up or far shot. The control unit (12) enables the detection of the orientation information of the vehicle in the image. The control unit (12) enables the detection of the color of the vehicle. The control unit (12) enables the detection of the vehicle type of the vehicle. The control unit (12) enables the detection of the license plate of the vehicle. The control unit (12) enables the detection of the accuracy of the images. The control unit (12) enables the detection of the vehicle parts in the image. The control unit (12) enables the calculation of the size and depth of damage to vehicle parts, including scratches, dents, fractures, glass damage, tire damage. The control unit (12) enables the detection of sub-damaged parts according to the size and depth of the damage. The control unit (12) enables the determination of the repair status of the damaged vehicle part according to vehicle type, vehicle part, damage type, damage size / part size. The control unit (12) enables the calculation of the marker value of the vehicle. The control unit (12) enables the generation of price information for the repair of damaged parts. According to said price information, it enables the generation of total repair price information. The control unit (12) enables the classification of damage according to the total price information as mini repair, module damage, normal damage, heavy damage. The control unit (12) enables the presentation of all detected and calculated information to the user with a report. Said report is transmitted to the user interface (13) via the control unit (12).

[0061] The scope of protection of the invention is specified in the appended claims and cannot be limited to what is described for illustrative purposes in this detailed description. It is clear that a person skilled in the art can produce similar embodiments in the light of what is explained above, without deviating from the main theme of the invention.

[0062] REFERENCE NUMERALS GIVEN IN DRAWING

[0063] 10 Damage assessment system

[0064] 11 Camera

[0065] 12 Control unit

[0066] 13 User interface

Claims

CLAIMS1 . A method performed by a control unit (12) in a damage assessment system (10) comprising a camera (11 ) for taking at least one image of a vehicle for which damage assessment is desired and said control unit (12) for performing damage assessment for said vehicle according to said image taken by said camera (11), and for generating a damage report according to said damage assessment for viewing using a user interface (13), characterized in that it comprises the process steps of:- taking at least one image from the camera (11),- inputting the vehicle image taken from the camera (11) to a first mathematical model that has been trained with vehicle images and the orientation information of the vehicle parts in these vehicle images and that, upon receiving a vehicle image as input, outputs the orientation information corresponding to this vehicle image and outputting an orientation information,- inputting the vehicle image taken from the camera (11) to a second mathematical model that has been trained with vehicle type images and that, upon receiving a vehicle type image as input, outputs the type information of the vehicle corresponding to this vehicle type image, and outputting the vehicle type information,- inputting the vehicle image taken from the camera (11) to a third mathematical model that has been trained with vehicle images and information on vehicle parts in said vehicle images and that, upon receiving a vehicle image as input, outputs at least one vehicle part information corresponding to this vehicle image, and outputting at least one vehicle part information,- inputting the vehicle image taken from the camera to a fourth mathematical model that has been trained with vehicle damage images and information on the type and pixel location of said damage and that, upon receiving a vehicle image as input, outputs damage information including the type and pixel locations of damage in this vehicle image, and outputting a damage information,- obtaining data containing information on the damaged part of the vehicle and the area affected by the damage to the vehicle by applying thedetected orientation information, vehicle type information, vehicle part information, and damage information to a first mathematical equation,- inputting the vehicle image taken from the camera (11) to a fifth mathematical model that has been trained with a vehicle damage image and depth dimension information of this vehicle part and that, upon receiving a vehicle damage image as input, outputs the depth information of the vehicle damage corresponding to this vehicle damage image, and outputting the depth information of the vehicle damage;- inputting the vehicle image taken from the camera (11) to a sixth mathematical model that has been trained with vehicle type information, information on damaged part of the vehicle, information on the area affected by the damage on the vehicle, and depth information and that, upon receiving a vehicle image as input, outputs the information on the lower mechanical part that is under said damaged vehicle part and is not visible in the two-dimensional image, and outputting the information on the lower mechanical parts that are not visible under the damaged vehicle part,- generating price information for the assessed damaged vehicle parts by accessing the information in a memory where the workload to be incurred for the repair of damaged vehicle parts and the cost information to be incurred for this workload are stored,- generating a report to be forwarded to the user interface (13).

2. A method according to claim 1 , characterized in that the first mathematical model is a VGG16 artificial intelligence model.

3. A method according to claim 1 , characterized in that the second mathematical model is a VGG16 artificial intelligence model.

4. A method according to claim 1 , characterized in that the third mathematical model is a UNET artificial intelligence model.

5. A method according to claim 1 , characterized in that the fourth mathematical model is a ResNet18 artificial intelligence model.

6. A method according to claim 1 , characterized in that the fifth mathematical model is a ResNet18 artificial intelligence model.

7. A method according to claim 1 , characterized in that the sixth mathematical model is a Gradient Boosting and random forest artificial intelligence model.

8. A method according to claim 1 , characterized in that it comprises the step of inputting the images taken from the camera (11) to a seventh mathematical model that has been trained with vehicle images, images of documents used for vehicles, and images of the environment in which a vehicle may have an accident, and that, upon receiving the images taken from the camera (11) as input, outputs an image information including the information indicating that the corresponding image is related to at least one of the vehicle image, image of documents used for vehicles, and image of the environment in which a vehicle may have an accident, and identifying the image according to the image information obtained from said seventh mathematical model.

9. A method according to claim 8, characterized in that the seventh mathematical model is a ResNet18 artificial intelligence model.

10. A method according to claim 1 , characterized in that it comprises the step of inputting the vehicle image taken from the camera (11) to an eighth mathematical model that has been trained with vehicle interior images and vehicle exterior images and that, upon receiving a vehicle image as input, outputs a vehicle image information indicating that this vehicle image is the vehicle interior image or the vehicle exterior image, and outputting a vehicle image information.11 . A method according to claim 10, characterized in that the eighth mathematical model is a VGG16 artificial intelligence model.

12. A method according to claim 1 , characterized in that it comprises the step of inputting the vehicle image taken from the camera to a ninth mathematical model that has been trained with distance information comprising the distance of the location from which an image is taken to the objects in the image, andthat, upon receiving a vehicle image as an input, outputs a distance information regarding the distance of the location from which this vehicle image is taken to the vehicle, and outputting a distance information.

13. A method according to claim 12, characterized in that t e ninth mathematical model is a VGG16 artificial intelligence model.

14. A method according to claim 1 , characterized in that it comprises the step of inputting the vehicle image taken from the camera (11 ) to a tenth mathematical model that has been trained with the airbag location images and that, upon receiving a vehicle image as input, outputs the airbag usage information corresponding to this vehicle image, and receiving the airbag usage information.

15. A method according to claim 14, characterized in that the tenth mathematical model is a YOLOv8 artificial intelligence model.

16. A method according to claim 1 , characterized in that it comprises the step of inputting a vehicle image taken from the camera to an eleventh mathematical model that has been trained with heavily damaged vehicle images, and that, upon receiving a vehicle image as input, outputs a heavy damage information indicating that the vehicle is heavily damaged, and outputting the heavy damage information.

17. A method according to claim 16, characterized in that the eleventh mathematical model is a ResNet18 artificial intelligence model.

18. A method according to claim 1 , characterized in that it comprises the step of inputting the vehicle image taken from the camera to a twelfth mathematical model that has been trained with the vehicle image and the scratch and dent information in said vehicle images and that, upon receiving a vehicle image as input, outputs a first vehicle damage information including the information on scratch and dents on said vehicle, and outputting the first vehicle damage information.

19. A method according to claim 18, characterized in that t e twelfth mathematical model is a UNET artificial intelligence model.

20. A method according to claim 1 , characterized in that it comprises the step of inputting the vehicle image taken from the camera to a thirteenth mathematical model that has been trained with the vehicle image and the broken vehicle part, damaged tire, and damaged glass information in said vehicle images, and that, upon receiving a vehicle image as input, outputs a second vehicle damage information including the information on broken vehicle part, damaged tire, and damaged glass in said vehicle, and outputting the second vehicle damage information.21 . A method according to claim 20, characterized in that the thirteenth mathematical model is a YOLOv8 artificial intelligence model.

22. A method according to claim 1 , characterized in that, after the step of generating price information for the assessed damaged vehicle parts by accessing the information in a memory where the workload to be incurred for the repair of damaged vehicle parts and the cost information to be incurred for this workload are stored, it comprises the step of generating a total price information for the repair of damaged vehicle parts, and determining a damage type according to this total price information.

Images