Patent Issued for Undamaged/damaged determination (USPTO 11636581): Tractable Limited
2023 MAY 15 (NewsRx) -- By a
The patent’s inventors are Aktas, Rusen (
This patent was filed on
From the background information supplied by the inventors, news correspondents obtained the following quote: “Typically, and as shown in FIG. 1, when a vehicle is involved in an accident (or is damaged) 105, the vehicle or its driver will be insured, and the driver will contact the relevant insurance company 110 to make a claim following a typical claim procedure 100.
“The insurance company’s estimation team 135 will then need to assess the damage to the vehicle and approve any claim, and the driver or insurer will then arrange for the vehicle to be repaired 145. Alternatively, the insurance company may make a cash settlement 150 in place of arranging or paying for repairs or may make a decision that the vehicle is a total loss 140 and compensate the insured party accordingly or arrange for a replacement vehicle to be procured.
“As shown in FIG. 1, the claim procedure 100 following an accident 105 requires the driver or insured party to call their insurer 110, and personnel at the insurer will follow a script 115 to receive and process the claim.
“As part of the script 115, the insurer will obtain from the driver or insured party some information about the accident 105. Typically, the insurer will be provided with information about the insured person 120 (which may also include details of the vehicle and its condition etc. that are provided during the call, or which are stored in the insurer’s database and retrieved following receipt of the details of the insured person); details of the crash or accident 125, for example the circumstances and extent of the damage; and photos of the damage 130.
“The photos of the damage 130 are typically taken by the driver or the insured party and can be of varying quality and comprehensiveness. Typically, photos are taken using phones equipped with cameras. Various problems can arise from this approach, including that too few photos are taken and provided to the insurer. Also, the photos taken may not be sufficiently well composed or may be of low quality due to the quality of the camera used to take the photos or the skill of the user.
“The photos of the damage 130 can be provided to the insurer either via e-mail, facsimile or post, for example. This means there is typically a delay in the receipt of the photos 130 by the insurer, thus delaying the processing of the claim by the insurer and slowing down the decision-making process as to whether the damage is a total loss 140, or whether a cash settlement 150 can be offered, or whether to arrange or allow the driver or insured part to arrange for repairs to the vehicle 145.
“As part of the claim procedure, and more specifically the claim review procedure which is carried out by the insurer to verify the costs of the proposed repair work by manually assessing data provided by the client and any proposed repairer, the insurer may request further information or claim data to be provided from the driver or insured party regarding the accident. This may include details of the vehicle and its condition prior to any damage etc. These are typically provided during a telephone call or are obtained having been stored in the insurer’s database, but sometimes requires the insurer to contact the insured party in a follow up telephone call, letter or e-mail requesting the further details. Further, the insurer will require sufficient details of the accident to be provided, along with sufficient photographs of the damage for example, so this must be obtained during the first and any subsequent contact with the insured party. The process of obtaining sufficient information can be slow, especially if further requests for information are made in separate subsequent contacts with the insured party, and thus can significantly delay the processing of an insurance claim. Further, the proposed repairer may be required to send details of the proposed repairs, including for example the labour tasks as well as any parts or materials costs, to the insurer for approval prior to commencing work. The insurer can then assess whether the claim is covered by the relevant policy under which the claim is made and determine whether the estimated costs of repair can be verified and/or approved as may be appropriate.
“Various tools and processes have been developed to assist vehicle repair businesses and vehicle insurers respectively to prepare and approve repair proposals for damaged vehicles, for example as a result of the vehicle being involved in an accident.
“Vehicle repair businesses need to be able to itemise both the labour required and the specific parts required in order to repair the vehicle, and then submit this for approval to an insurer where the repair is covered by an insurance policy. Due to the large number of different possible makes and models that might require repair, and the optional extras that might have been fitted to the vehicle to be repaired, vehicle repair businesses typically have to use a commercial database to identify the correct make, model, year of manufacture and options fitted in order to correctly identify the parts that would need to be ordered if any need replacement.
“Insurers typically require vehicle repair businesses to submit evidence of the damage to each vehicle and a detailed repair proposal that itemises the parts to be ordered and the respective costs of each part along with detailed itemisation of the labour tasks and time that will be required to carry out any repairs or replacement of parts. Preparing such detailed repair proposals manually typically takes vehicle repair businesses a significant amount of time.
“In different jurisdictions, different approaches are taken by both vehicle repair businesses (in respect of how repairs are carried out, what labour is deemed to be required, and preferences as to whether to repair or replace parts, for example) and insurers (in respect of what policies are applied when approving or rejecting proposed repairs, for example), so depending on a variety of factors such as commercial pressures, regulation, consumer preference and typical insurance coverage. Thus, detailed repair proposals will differ between jurisdictions and what insurers are prepared to approve in a detailed repair proposal will also differ between jurisdictions.
“Insurers, however, typically perform manual reviews on proposed repairs that are submitted for approval by vehicle repair businesses. As a result, the manual review process either requires a large workforce to perform the task of reviewing each submitted repair proposal or becomes a bottleneck in the repair approval process. For vehicle repair businesses, manual review can result in several disadvantages including delay in being able to begin repair work; further delays if the repair proposal is rejected by the insurer; and having to store customer vehicles for longer periods than necessary resulting in both higher storage space requirements and a higher probability of dissatisfied customers.
“Across all jurisdictions, a variety of the above-described problems can result from manual preparation of proposed vehicle repairs and manual review of the proposed vehicle repairs by insurers.
“Improvements to the claim procedure would enable repairs to be completed sooner and for insurers to reach decisions faster and more efficiently.”
Supplementing the background information on this patent, NewsRx reporters also obtained the inventors’ summary information for this patent: “Aspects and/or embodiments seek to provide a computer-implemented method for determining damage states of each part of a damaged vehicle, indicating whether each part of the vehicle is damaged or undamaged and optionally the severity of the damage to each part of the damaged vehicle, using images of the damage to the vehicle and trained models to assess the damage indicated in the images of the damaged vehicle.
“According to a first aspect, there is provided a computer-implemented method for determining one or more damage states to one or more parts of a vehicle, comprising the steps of: receiving a plurality of images of the vehicle; determining, per part, one or more feature representations for each of the plurality of images using a first trained model, wherein each feature representation comprises an indication of damage to at least one part; concatenating, per part, the one or more feature representations for each of the plurality of images; determining a damage state of each of the parts of the vehicle using the concatenated feature representation per part using a second trained model; and outputting the one or more damage states of the parts of the vehicle.
“Providing a method for determining the damage states of parts of a vehicle can allow for a substantially accurate estimate of what repairs are required to a damaged vehicle using images of the vehicle and/or damage to the vehicle. Optionally, the models and/or classifiers can be trained using supervised or unsupervised techniques.
“Optionally, the step of concatenating comprises one or more pooling operations performed on the one or more feature representations, wherein the one more pooling operations comprise any one or any combination of: mean function; max function; attentive pooling function; region of interest pooling; and/or softmax layer.
“Various concatenation methods can provide options to combine the results per image into a combined result (i.e. classification of damage) per part across multiple images in which the part is shown.”
The claims supplied by the inventors are:
“1. A computer-implemented method for determining one or more damage states to one or more parts of a vehicle, comprising: receiving a plurality of images of the vehicle; determining one or more feature representations for each of the plurality of images using a first trained model, wherein each feature representation comprises an indication of damage to at least one part, wherein the first trained model comprises a plurality of image classifiers, each image classifier trained to detect damage to one of a plurality of normalized parts of the vehicle and each of the plurality of images of the vehicle are processed by each of the plurality of image classifiers, wherein each of the plurality of image classifiers are generic with respect to a make and model of the vehicle; concatenating, per part, the one or more feature representations for each of the plurality of images; determining a damage state of each of the parts of the vehicle using the concatenated feature representation per part using a second trained model; and outputting the one or more damage states of the parts of the vehicle.
“2. The method of claim 1, wherein the concatenating comprises one or more pooling operations performed on the one or more feature representations, wherein the one more pooling operations comprise any one or any combination of: mean function; max function; attentive pooling function; region of interest pooling; or softmax layer.
“3. The method of claim 1, wherein determining a damage state of each of the parts of the vehicle is only performed when each of the received plurality of images of the vehicle are analyzed using the first trained model.
“4. The method of claim 1, further comprising determining a plurality of parts of the damaged vehicle that are represented in the plurality of images of the damaged vehicle comprises the use of a plurality of classifiers.
“5. The method of claim 4, wherein each one of the plurality of classifiers is operable to detect each of the parts of the damaged vehicle.
“6. The method of claim 1, wherein the damage states of parts of the vehicle are determined as one or more quantitative values.
“7. The method of claim 1, wherein the first trained model comprises a multi-instance model, further wherein the step of determining, per part, one or more feature representations is performed substantially simultaneously per part.
“8. The method of claim 1, wherein the first trained model comprises a graph neural network.
“9. The method of claim 8, wherein each of the one or more feature representations for each of the plurality of images is represented as a node in the graph neural network.
“10. The method of claim 9, wherein the concatenating, per part, the one or more feature representations comprises determining a relationship between the nodes on the graph neural network.
“11. The method of claim 10, wherein the connection between the nodes on the graph neural network comprises one or more learned function.
“12. The method of claim 1, wherein the first trained model and the second trained model are substantially the same.
“13. The method of claim 1, wherein the first or second trained model comprises any one or any combination of: a neural network; a convolutional neural network; or a recurrent neural network.
“14. The method of claim 1, wherein the image resolution of each of the received plurality of images is substantially maintained.
“15. The method of claim 1, further comprising: cropping each of the plurality of images into a plurality of cropped images, wherein the determining one or more feature representations comprises determining, per part, one or more feature representations for each of the plurality of cropped images using a first trained model, wherein each feature representation comprises an indication of damage to at least one part.
“16. The method of claim 1, further comprising: segmenting each of the plurality of images to produce one or more segmentations, wherein the determining one or more feature representations uses the one or more segmentations determined for each of the plurality of images.”
For the URL and additional information on this patent, see: Aktas, Rusen. Undamaged/damaged determination.
(Our reports deliver fact-based news of research and discoveries from around the world.)
Patent Issued for Processing system having a machine learning engine for providing a selectable item availability output (USPTO 11635877): Allstate Insurance Company
New Risk Management Research from Peter the Great St. Petersburg Polytechnic University Discussed (Developing a System for Monitoring Human Resource Risks in a Digital Economy): Insurance – Risk Management
Advisor News
Annuity News
Health/Employee Benefits News
Life Insurance News