Systems and methods for estimating monetary loss to an accident damaged vehicle

Patent metadata

Field	Value
Publication number	US-12148043-B2
Application number	US-202117521821-A
Country	US
Kind code	B2
Filing date	Nov 8, 2021
Priority date	Nov 8, 2021
Publication date	Nov 19, 2024
Grant date	Nov 19, 2024

Abstract

Embodiments of the disclosure provide systems and methods for estimating an amount of monetary loss to an accident damaged vehicle. An exemplary system includes a communication interface configured to receive one or more accident images taken of an accident damaged vehicle. It also includes a database for storing loss data of one or more past vehicles, and each past vehicle is associated with a historical accident. It further includes a processor coupled to the communication interface and the database. The processor is configured to detect the accident damaged vehicle in the one or more accident images, identify one or more most similar past vehicles, and estimate the amount of monetary loss to the accident damaged vehicle based on the loss data of the one or more past vehicles.

First claim

What is claimed is:

1. A system for estimating an amount of monetary loss to an accident damaged vehicle, comprising:
a communication interface configured to receive one or more accident images taken of the accident damaged vehicle;
a database for storing loss data of past vehicles, each past vehicle being associated with a historical accident; and
a processor coupled to the communication interface and the database and configured to:
detect the accident damaged vehicle in the one or more accident images;
apply a trained model to identify one or more most similar past vehicles from the past vehicles; and
estimate the amount of monetary loss to the accident damaged vehicle based on the loss data of the one or more most similar past vehicles,

wherein the processor is further configured to:
select one or more past vehicles for evaluation from the past vehicles; and
evaluate the trained model by calculating at least one of an average amount of monetary losses to the one or more past vehicles for evaluation or a relative similarity.

2. The system of claim 1 , wherein the processor is further configured to:
detect a body area of the accident damaged vehicle in each of the accident images; and
determine a view angle of the accident damaged vehicle in each of the accident images.

3. The system of claim 2 , wherein the processor is further configured to:
detect one or more vehicle parts of the accident damaged vehicle; and
merge the detected one or more vehicle parts to define a body area of the accident damaged vehicle.

4. The system of claim 3 , wherein the processor is further configured to:
prior to or during the merging of the detected one or more vehicle parts, filter out the vehicle part from a result of the detected vehicle parts if a dimension of the vehicle part is less than a first preset threshold or a distance between a centroid of the detected vehicle part and a boundary of the accident image is less than a second preset threshold.

5. The system of claim 3 , wherein the processor is further configured to:
determine a view angle of the accident damaged vehicle based on a spatial relationship of the detected vehicle parts or one or more bounding boxes surrounding the body area.

6. The system of claim 1 , wherein the processor is further configured to:
apply the trained model to the accident images to generate a first batch of embedding vectors associated with the accident damaged vehicle; and
compare the first of embedding vectors with a second batch of embedding vectors associated with the past vehicles to identify the one or more most similar past vehicles.

7. The system of claim 6 , wherein the processor is further configured to:
retrieve, from the database, the second batch of embedding vectors associated with the past vehicles, which have been generated by the trained model.

8. The system of claim 6 , wherein the processor is further configured to:
calculate a weighted average of similarity scores between the first batch of embedding vectors and the second batch of embedding vectors to identify the one or more most similar past vehicles.

9. The system of claim 1 , wherein the loss data contain a total amount of monetary loss to each of the past vehicles, and
wherein the processor is further configured to:
calculate a weighted similarity between the accident damaged vehicle and the one or more most similar past vehicles;
retrieve, from the database, the total amount of monetary loss for each of the one or more most similar past vehicles; and
estimate the amount of monetary loss to the accident damaged vehicle based on the calculated weighted similarity and the retrieved total amount of monetary losses for the one or more most similar past vehicles.

10. The system of claim 9 , wherein the processor is further configured to:
determine whether the accident damaged vehicle suffered a total loss.

11. The system of claim 9 , wherein the processor is further configured to:
determine a confidence level of the one or more most similar past vehicles by evaluating a variance among similarity weights calculated between the accident damaged vehicle and the one or more most similar past vehicles; and
manually determine the monetary loss to the accident damaged vehicle if the confidence level is below a predetermined threshold.

12. The system of claim 1 , wherein the loss data contain cost of a replacement part and labor cost corresponding to the replacement part, and
wherein the processor is further configured to:
calculate a weighted similarity between the accident damaged vehicle and the one or more most similar past vehicles;
retrieve, from the database, the cost of a replacement part and the labor cost corresponding to the replacement part for each of the one or more most similar past vehicles; and
estimate cost for replacing a damaged vehicle part of the accident damaged vehicle and labor cost corresponding to the damaged vehicle part based on the calculated weighted similarity and the retrieved cost of a replacement part and the retrieved labor cost.

13. The system of claim 1 , wherein the processor is further configured to:
generate a similarity label for each pair of the past vehicles; and
compile a similarity correlation matrix using the generated similarity label for each pair of the past vehicles.

14. The system of claim 13 , wherein the processor is further configured to:
divide the loss data of the past vehicles into one or more sub-datasets;
sample data for each sub-dataset;
use the sampled data as baseline cases;
select supporting cases based on the similarity correlation matrix; and
obtain the trained model by using the baseline cases and the supporting cases.

15. The system of claim 14 , wherein to select the one or more past vehicles for evaluation, the processor is further configured to:
apply the trained model to a validation case to generate a validation embedding vector associated with the validation case; and
calculate a similarity between the validation embedding vector and each embedding vector of the past vehicles to select the one or more past vehicles for evaluation.

16. A method for estimating an amount of monetary loss to an accident damaged vehicle, comprising:
receiving one or more accident images taken of the accident damaged vehicle;
detecting the accident damaged vehicle in the one or more accident images;
receiving loss data of past vehicles;
applying a trained model to identify one or more most similar past vehicles from the past vehicles; and
estimating the amount of monetary loss to the accident damaged vehicle based on the loss data of the one or more most similar past vehicles,
wherein the method further comprises:
selecting one or more past vehicles for evaluation from the past vehicles; and
evaluating the trained model by calculating at least one of an average amount of monetary losses to the one or more past vehicles for evaluation or a relative similarity.

17. The method of claim 16 , further comprising:
converting each of the one or more accident images to a preprocessed images by:
detecting parts of the damaged vehicle in the accident image;
determining a view angle of the damaged vehicle in the accident image based on the detected parts;
determining a bounding box of the damaged vehicle based on the detected parts; and
generating the preprocessed image based on the bounding box.

18. The method of claim 16 , further comprising:
generating a similarity correlation matrix from the loss data by:
calculating similarities of

Assignees

• Ping An Tech Shenzhen Co Ltd

Inventors

• Yao Xi
• Lin Ruei-Sung
• Li Kun
• Chen Zijia

Classifications

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G06F18/213

  Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods · CPC title

• G06V20/20

  in augmented reality scenes · CPC title

• G06Q30/0278

  Product appraisal · CPC title

• G06V10/82

  using neural networks · CPC title