Vehicle insurance underwriting vehicle inspection video acquisition method and system based on deep learning

Abstract

The invention discloses a vehicle insurance underwriting vehicle inspection video acquisition method and system based on deep learning, and the method comprises the steps: checking whether the posture of a user holding a mobile phone and the surrounding light environment meet conditions or not, and if not, adjusting the posture and turning on light supplement equipment; a vehicle certificate is shot and scanned by using a handheld mobile phone, wherein the vehicle certificate comprises an insurance applicant identity card, a driving license, a VIN code, a vehicle four-corner photo, a person-vehicle group photo and an odometer of a driving seat in a vehicle; the handheld mobile phone is internally provided with a lightweight target detection algorithm module, and the target detection algorithm module monitors the category of image content shot by a user in real time. According to the classification method based on the deep neural network, extremely high accuracy can be obtained, and the guide prompt accuracy is high.

Description

Technical Field

[0001] The present invention relates to the field of vehicle insurance verification technology, and in particular to a method and system for collecting vehicle insurance underwriting and vehicle inspection videos based on deep learning. Background Art

[0002] The insurance underwriting and vehicle inspection process is a crucial step for insurance companies to mitigate underwriting risks and ensure the effective implementation of insurance contracts. However, in practice, this process has also exposed some significant drawbacks, affecting the efficiency and fairness of insurance business.

[0003] 1. Long time cycle and low efficiency

[0004] The auto insurance survey and vehicle inspection process often lacks clear timelines, resulting in lengthy work. Surveyors are required to inspect the accident scene, communicate with relevant personnel, and collect evidence. However, due to the lack of clear time limits, surveys are often delayed, resulting in lengthy claims processing times and inconvenience for the insured. The vehicle inspection process also faces similar challenges. Especially with the explosive growth of the auto insurance market, inspectors struggle to keep up with the demand for inspections of all vehicles, resulting in low inspection efficiency.

[0005] 2. Lack of Professionalism and Objectivity

[0006] During implementation, the legal expertise and sense of responsibility of surveyors and vehicle inspectors have been inconsistent, compromising the accuracy and objectivity of survey reports and vehicle inspection results. Some surveyors and vehicle inspectors may lack sufficient expertise to accurately determine accident liability or vehicle condition, thus impacting the fairness of auto insurance claims. Furthermore, some surveyors and vehicle inspectors may intentionally conceal facts for profit, distorting underwriting results.

[0007] 3. Incomplete data collection

[0008] The data collection process during the auto insurance survey and vehicle inspection process is often flawed. Surveyors sometimes fail to collect relevant evidence, such as photos and videos, during the on-site inspection, compromising the accuracy of the accident reconstruction. Furthermore, during the vehicle inspection process, inspectors may fail to fully examine the vehicle's condition, leading to potential risks being overlooked. The supporting documentation provided by the insured may also be incomplete, complicating the work of surveyors and inspectors and, in turn, impacting the efficiency of claims processing.

[0009] IV. Lack of effective supervision and feedback mechanisms

[0010] The auto insurance survey and vehicle inspection process lacks an effective oversight and feedback mechanism. Because surveys and vehicle inspections involve multiple steps, without effective oversight and feedback mechanisms, some irresponsible surveyors and inspectors may shirk their duties, leaving work quality unchecked and uncorrected. This not only harms the rights of the insured but also tarnishes the reputation and image of the insurance company.

[0011] 5. Information asymmetry is a prominent issue

[0012] Insureds often lack a comprehensive understanding of insurance companies' vehicle inspection and survey processes and requirements, leading to uncertainty about how to respond and cooperate with the inspection process when an accident occurs. Furthermore, some criminals exploit information asymmetry to commit insurance fraud, such as fabricating insurance benefits, creating insurance incidents, and falsifying accident scenes, resulting in significant financial losses for insurance companies.

[0013] In summary, the insurance underwriting and vehicle inspection process suffers from shortcomings in practice, including long processing times, insufficient professionalism and objectivity, incomplete data collection, a lack of effective oversight and feedback mechanisms, and information asymmetry. To improve the efficiency and fairness of vehicle inspections and surveys, it is necessary to optimize and improve these processes. For example, advanced technologies can be introduced to enhance the efficiency and quality of inspections and surveys; training and management of surveyors and inspectors can be strengthened to enhance their professionalism and sense of responsibility; a comprehensive oversight and feedback mechanism can be established to promptly identify and address operational issues; and enhanced communication with insureds to improve their insurance awareness and cooperation. These measures can effectively reduce underwriting risks and enhance the competitiveness and sustainable development of insurance companies. Summary of the Invention

[0014] In response to the shortcomings of the above problems, the present invention provides a method and system for collecting video for automobile insurance underwriting and vehicle inspection based on deep learning.

[0015] To achieve the above objectives, the present invention provides a deep learning-based vehicle insurance underwriting and vehicle inspection video acquisition method, which includes checking whether the user's mobile phone holding posture and the surrounding lighting environment meet the conditions. If not, the posture is adjusted and the fill light device is turned on;

[0016] Use the handheld mobile phone to take a photo and scan the vehicle documents, which include the insured's ID card, driving license, VIN code, four-corner photos of the vehicle, a photo of the person and the vehicle, and the odometer in the driver's seat of the vehicle;

[0017] The handheld mobile phone has a built-in lightweight target detection algorithm module, and the target detection algorithm module monitors the category of the content of the image taken by the user in real time.

[0018] Preferably, using the handheld mobile phone to photograph and scan the vehicle certificate includes:

[0019] The key words prompted to the user during the scanning process are: Please place the insured's ID card in the frame and keep still, please place the insured's driving license in the frame and keep still;

[0020] Please move to the vehicle's VIN code location and place the VIN code in the box and remain still. Please move to the front left of the vehicle and remain still.

[0021] Please move to the right front of the vehicle and remain still;

[0022] Please move to the right rear of the vehicle and remain still;

[0023] Please move to the left rear of the vehicle in question and remain still;

[0024] Ask the insured to take a photo with the vehicle and remain still;

[0025] Please move to the cab to take a picture of the odometer and remain still.

[0026] Preferably, after the prompt to keep still is given, a countdown of 3 seconds is started, and frame difference algorithm 2 is performed on the video frame. When the 3s time is established and the frame difference trigger condition of the video frame within 3s is established, the video recording is considered successful and the recording is ended; if not, rescan and start keeping still.

[0027] Preferably, the timeout period of the scanning step is set to 15 seconds. If the timeout period expires, the recording is terminated and a recording failure is returned.

[0028] Preferably, the target detection algorithm module is Faster RCNN, SSD, YOLO, YOLO-v5, Nanodet or Rtmdet.

[0029] The present invention also provides a deep learning-based vehicle insurance underwriting and vehicle inspection video acquisition system, comprising:

[0030] The checking module is used to check whether the user's holding posture of the phone and the surrounding lighting environment meet the requirements. If not, the posture is adjusted and the fill light device is turned on;

[0031] A scanning module is used to use the handheld mobile phone to photograph and scan vehicle documents, wherein the vehicle documents include the insured's ID card, driving license, VIN code, four-corner photos of the vehicle, a photo of the person and the vehicle, and the odometer in the driver's seat of the vehicle;

[0032] The handheld mobile phone has a built-in lightweight target detection algorithm module, and the target detection algorithm module monitors the category of the content of the image taken by the user in real time.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] The classification method based on deep neural networks in this invention can achieve extremely high accuracy, and the guidance and prompt accuracy is high; in the vehicle inspection process, real-time voice and prompt text guide the vehicle inspectors to collect each link, which is convenient for the standardization and standardization of image data; the efficiency of the vehicle inspection process is improved, and the deep learning-based solution can significantly shorten the vehicle inspection operation time compared with the traditional operation process; video-based collection also has good anti-fraud and tampering protection compared with picture collection, reducing company leakage. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a flowchart of the deep learning-based vehicle insurance underwriting and vehicle inspection video collection method. DETAILED DESCRIPTION

[0036] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0037] Reference Figure 1 The present invention provides a method for collecting video for vehicle underwriting and vehicle inspection based on deep learning, which includes checking whether the user's mobile phone holding posture and the surrounding lighting environment meet the conditions. If not, the posture is adjusted and the fill light device is turned on;

[0038] Use a handheld mobile phone to take a photo and scan the vehicle documents, which include the insured's ID card, driving license, VIN number, four-corner photos of the vehicle, a photo of the insured and the vehicle, and the odometer in the driver's seat of the vehicle;

[0039] Among them, the handheld mobile phone has a built-in lightweight target detection algorithm module, which monitors the category of the image content taken by the user in real time.

[0040] In this embodiment, the target detection algorithm module is Faster RCNN, SSD, YOLO, YOLO-v5, Nanodet or Rtmdet.

[0041] Example

[0042] Check whether the user's phone holding posture and surrounding lighting environment meet the requirements. If not, prompt the user to adjust the posture and turn on the fill light device;

[0043] The mobile phone has a built-in lightweight object detection algorithm module that supports ID cards, driving licenses, VIN codes, vehicles at 45 degrees to the left front, 45 degrees to the right front, 45 degrees to the right rear, 45 degrees to the left rear, photos of people and vehicles together, and in-vehicle odometer detection. This module is used to detect the category of the content of the user's captured image in real time;

[0044] The scanning process begins with scanning documents, including the ID card, driving license, and VIN code, followed by photos of the four corners of the vehicle, a photo of the driver and the vehicle, and finally a scan of the odometer in the driver's seat. The key words prompted to the user during the scanning process are: Please place the insured's ID card in the frame and remain still, please place the insured's driving license in the frame and remain still, please move to the vehicle's VIN code position and place the VIN code in the frame and remain still, please move to the front left of the vehicle and remain still, please move to the front right of the vehicle and remain still, please move to the rear right of the vehicle and remain still, please move to the rear left of the vehicle and remain still, please take a photo of the insured with the vehicle and remain still, please move to the driver's cab to take a photo of the odometer and remain still;

[0045] The last scan step in the "Stay Still" function is explained as an example. When the video frame detects the "Odometer in the Cab," a "Stay Still" prompt is displayed, and a 3-second countdown begins. Simultaneously, the video frame is subjected to frame difference algorithm 3. When the 3-second countdown is met and the frame difference trigger condition within 3 seconds is met, the video recording is considered successful and ends. If the above conditions are not met, the "Odometer in the Cab" is scanned again and the "Stay Still" function is initiated. The timeout for this scan step is set to 15 seconds. If the timeout expires, the recording ends and a "Recording Failure" message is displayed.

[0046] The present invention also provides a deep learning-based vehicle insurance underwriting and vehicle inspection video acquisition system, comprising:

[0047] The checking module is used to check whether the user's holding posture of the phone and the surrounding lighting environment meet the requirements. If not, the posture is adjusted and the fill light device is turned on;

[0048] The scanning module is used to use a handheld mobile phone to take photos and scan vehicle documents. Vehicle documents include the insured's ID card, driving license, VIN code, four-corner photos of the vehicle, a photo of the person and the vehicle, and the odometer in the driver's seat of the vehicle;

[0049] Among them, the handheld mobile phone has a built-in lightweight target detection algorithm module, which monitors the category of the image content taken by the user in real time.

[0050] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A method for collecting video for vehicle insurance underwriting and vehicle inspection based on deep learning, characterized in that: include: Check whether the user's phone holding posture and surrounding lighting environment meet the requirements. If not, adjust the posture and turn on the fill light device; Use the handheld mobile phone to take a photo and scan the vehicle documents, which include the insured's ID card, driving license, VIN code, four-corner photos of the vehicle, a photo of the person and the vehicle, and the odometer in the driver's seat of the vehicle; The handheld mobile phone has a built-in lightweight target detection algorithm module, and the target detection algorithm module monitors the category of the content of the image taken by the user in real time.

2. The method for collecting video of automobile insurance underwriting and vehicle inspection based on deep learning according to claim 1 is characterized in that: Using the handheld mobile phone to photograph and scan vehicle documents includes: The key words prompted to the user during the scanning process are: Please place the insured's ID card in the frame and keep still, please place the insured's driving license in the frame and keep still; Please move to the vehicle's VIN code location and place the VIN code in the box and remain still. Please move to the front left of the vehicle and remain still. Please move to the right front of the vehicle and remain still; Please move to the right rear of the vehicle and remain still; Please move to the left rear of the vehicle in question and remain still; Ask the insured to take a photo with the vehicle and remain still; Please move to the cab to take a picture of the odometer and remain still.

3. The method for collecting video of automobile insurance underwriting and vehicle inspection based on deep learning according to claim 2 is characterized in that: After the "Keep Still" prompt is displayed, a 3s countdown begins and the frame difference algorithm 1 is performed on the video frames. When the 3s time is met and the frame difference trigger condition of the video frames within 3s is met, the video recording is considered successful and the recording ends. If not, the scan is restarted and Keep Still is started.

4. The method for collecting video of automobile insurance underwriting and vehicle inspection based on deep learning according to claim 3 is characterized in that: The timeout period of the scanning step is set to 15 seconds. If it times out, the recording ends and a recording failure message is returned.

5. The method for collecting video of automobile insurance underwriting and vehicle inspection based on deep learning according to claim 4 is characterized in that: The target detection algorithm module is Faster RCNN, SSD, YOLO, YOLO-v5, Nanodet or Rtmdet.

6. A deep learning-based vehicle insurance underwriting and vehicle inspection video acquisition system, characterized by: include: The checking module is used to check whether the user's holding posture of the phone and the surrounding lighting environment meet the requirements. If not, the posture is adjusted and the fill light device is turned on; A scanning module is used to use the handheld mobile phone to photograph and scan vehicle documents, wherein the vehicle documents include the insured's ID card, driving license, VIN code, four-corner photos of the vehicle, a photo of the person and the vehicle, and the odometer in the driver's seat of the vehicle; The handheld mobile phone has a built-in lightweight target detection algorithm module, and the target detection algorithm module monitors the category of the content of the image taken by the user in real time.

Images