Abstract:
An attitude estimation is an important process
for auto-piloting unmanned aerial vehicles (UAVs)
control. We implemented the attitude estimation
system from the visible horizon in the images. We use
images from the camera to compensate IMU (Inertial
measurement units). We used FPGA since the image
processing needs hard computation and UAV’s
control needs hard real-time processing. The horizon
is detected through the use of morphological
smoothing, Sobel filter and Hough transform. Runlength
encoder is used before Hough transform to
reduce memory bandwidth. The system is
implemented by high-level synthesis and on Zynq
UltraScale+MPSoC ZCU102. The result shows the
system is faster than the implementation using
desktop computer four times.
