Plane Position Measurement for $\mu \text{LED}$ Based on Single Camera

${\mu} \mathbf{LED}$
, whose size is less than
$\mathbf{100}\ {\mu}\mathbf{m}$
, is the LED technology of miniaturization and matrixing. The prospect of
${\mu} \mathbf{LED}$
technology is widely used, such as micro sensors and display screen. But in actual application, it needs to be transferred to a large number of circuit substrates. This subject uses monocular vision to identify and locate
${\mu} \mathbf{LED}$
. The focus is on sub-pixel edge detection technology. Compared with traditional edge detection technology, sub-pixel edge detection technology can improve the accuracy by
$\mathbf{2\sim 10}$
times, and the accuracy is more higher. Hough transform, least squares and corner detection methods were compared in detail in straight line fitting. Weighted least squares were selected according to the characteristics of the
${\mu} \mathbf{LED}$
chip image. By image preprocessing, image segmentation, feature recognition, target positioning and other operations are performed on the circuit board and
${\mu} \mathbf{LED}$
chip pictures, respectively, to obtain micron-level position information. The magnification of the system can be determined by resolution board with a group of bars. The uncertainty of the results is analyzed, such as the tilt angle between camera and
${\mu} \mathbf{LED}$
, lens distortion and the edge of the image. The results show that the uncertainty is less than
$5{\mu}\mathbf{m}$
.