Heliostat surface shape detection system and method based on multi-view image recognition

1. A surface shape detection method of a heliostat based on multi-view image recognition, a 3D surface shape of the heliostat being reconstructed by measuring a pitch angle and a roll angle of each sub-mirror of the heliostat, the method comprising:

(1) determining a distance from the heliostat to a bracket and the number of image collectors in a multi-view image collector array according to an external dimension of the heliostat;

(2) mounting the multi-view image collector array on the bracket and adjusting each image collector so that main optical axes of the image collectors are parallel to each other and aligned with the heliostat;

(3) collecting, by the multi-view image collector array, heliostat images of a field of view corresponding to the multi-view image collector array and transmitting the heliostat images to a computer respectively;

(4) performing feature matching on the collected image data through a feature recognition technology in an image recognition technology to determine corresponding feature points in a common field of view of a plurality of image collectors, wherein a feature point of the heliostat has a real image on each image collector corresponding to the multi-view image collector array;

(5) wherein a deviation between the real image of the feature point in each image collector and an image center in an image coordinate system is (X i , Y i ), where i represents a number of the image collector; a center of the multi-view image collector array is an origin of a multi-view measurement coordinate system satisfying the right-hand rule, and a Z axis of the multi-view measurement coordinate system points to the heliostat; center coordinates of each image collector are (x i , y i , 0), and a distance between image collectors is L, coordinates of the multi-view measurement coordinate system of each real image point are (x i +X i ·Size Pixel , y i +Y i ·Size Pixel , 0);

(6) given a focal length of the multi-view image collector array is f, coordinates of an equivalent lens center of each image collector are (x i , y i , f); a 3D linear equation involving the equivalent lens center and the corresponding real image is

x

-

x

i

X

i

·

Size

Pixel

=

y

-

y

i

Y

i

·

Size

Pixel

=

z

-

f

-

f

(7) according to the equation

x

-

x

i

X

i

·

Size

Pixel

=

y

-

y

i

Y

i

·

Size

Pixel

=

z

-

f

-

f

,

a single feature point of the heliostat establishes a plurality of 3D linear equations in the multi-view image collector array, and relative coordinates (x i , y i , z i ) of linear intersection points in the multi-view measurement coordinate system are obtained by concatenating the plurality of 3D linear equations, the relative coordinates (x i , y i , z i ) of linear intersection points are the relative coordinates of the single feature point of the heliostat; and

(8) repeating the steps (3)-(7) to obtain a relative position information of a mirror surface of the heliostat in the multi-view measurement coordinate system, and calculating the surface shape of the heliostat.