Two-step anti-fraud vehicle insurance image collecting and quality testing method, system and device

A two-step anti-fraud vehicle insurance image collecting and quality testing method, wherein, first, judging a position of the vehicle damage object on a vehicle in a first step, and then determining vehicle components closest to the vehicle damage object in a second step and abstracting them into tabular data, specifically comprising the following steps:

• step 1, collecting vehicle insurance scene images, and marking a pointing arrow through a vehicle insurance claim process terminal, a pointing arrow is a vehicle orientation;
• wherein, the vehicle insurance scene images comprise vehicle scene images and vehicle damage detail images, the vehicle scene images comprise images taken from six angles:
• left front, right front, left side, right side, left rear, and right rear;
• the images show that the whole vehicle is exposed, and the proportion of vehicle is not less than 80% of the entire image;
• the vehicle damage detail images show the detailed position of the vehicle damage;
• the pointing arrow points from the rear to the front of the vehicle, indicating the orientation of the vehicle in the vehicle scene images;
• step 2, performing object detection on the collected vehicle insurance scene images, obtaining a confidence score of vehicle damage objects, and screening to obtain object coordinates;
• step 3, according to the vehicle orientation marked in the step 1 and the object coordinates in the step 2, obtaining the specific position of the object coordinates located in the whole vehicle;
• wherein, the calculation method for the specific position of the object coordinates located in the whole vehicle in step 3 includes the following steps:
• 1) calculating the distance between the object coordinate and the center point of each part of the vehicle using the following formula:
• li(S)=∥S−si∥2
• wherein, ∥·∥2 is a 2-norm, and i can be taken as 0, 1, and 2, representing the specific positions of the vehicle, including the front, middle, and rear, s represents the object coordinate;
• when i is taken as 0, 1, and 2, si represents the coordinates of the center points of the front, middle, and rear of the vehicle, the calculation method for si is as follows:
• s
• i
• =
• {
• 5
• ⁢
• s
• h
• +
• s
• t
• 6
• ,
• i
• =
• 0
• s
• h
• +
• s
• t
• 2
• ,
• i
• =
• 1
• s
• h
• +
• 5
• ⁢
• s
• t
• 6
• ,
• i
• =
• 2
• among them, sh is the coordinate of the front of the vehicle, st is the rear of the vehicle;
• 2) based on the distance between the object coordinate and the center point of each part of the vehicle, obtain the position indication I of the object coordinate, the formula is as follows:
• I
• =
• f
• ⁡
• (
• s
• )
• =
• {
• 0
• ,
• if
• ⁢
• l
• 0
• (
• s
• )
• =
• l
• m
• (
• s
• )
• 2
• ,
• if
• ⁢
• l
• 2
• ⁢
• (
• s
• )
• =
• l
• m
• (
• s
• )
• 1
• ,
• else
• among them, I represents that the position indication of the object coordinate, 0 represents that the object coordinate is located in the front of the vehicle, 1 represents that the object coordinate is located in the middle of the vehicle, and 2 represents that the object coordinate is located in the rear of the vehicle;
• lm is the minimum distance between the object coordinate and the center point of each part of the vehicle, and the calculation formula is as follows:
• lm(s)=min(l0(S),l1(S),l2(s))
• step 4, according to the object coordinates screened in the step 2, performing vehicle component detection on the vehicle insurance scene images, obtaining the confidence score and component coordinates of the vehicle components, and screening to obtain the vehicle component closest to the object coordinates;
• wherein, the object detection is carried out on the vehicle insurance scene images, and all component coordinates are obtained, the object coordinates screened through the step 2 are used to calculate the distance between the object coordinates and all component coordinates, the component with the smallest distance is selected as the vehicle component closest to the object coordinates;
• step 5, according to the specific position of the object coordinates located in the whole vehicle obtained in the step 3 and the vehicle components closest to the object coordinates obtained in the step 4, obtaining the position of the vehicle components closest to the object coordinates that are located in the whole vehicle, and abstracting them into the tabular data.