Research on the Application of Visual SLAM in Embedded GPU

Algorithm 1

Optimization of Gaussian pyramid construction and feature detection algorithm processes.

Input: An image I_m, Gaussian pyramid layers levelNum

Output: A set of feature points of the first layer image

1 Image is transferred to the GPU memory

2 cudaMemCopyHostToDevice(I_m);

3 Construct Gaussian pyramid of the image on GPU

4 for i 0 to levelNum

5  d_ I_p[i] gpuBuildScalePyramid(I_m);

6 end

7 Calculate the feature points of the first layer image

8 gpuFeatureDetect(d_I_p[0]);

9 P_t[0][0…N_t1] gpuGetFeaturePoint();

10 Synchronize the instructions of constructing Gaussian pyramid

11 synchronization();

Note: The beginning of d refers to the data allocated in the video memory, N_t1 is the temporary number of feature points