초록 ▼

The present disclosure relates generally to optical flow algorithms. Section 1 of the present disclosure describes an optical flow algorithm with real-time performance and adequate accuracy for embedded vision applications. This optical flow algorithm is based on a ridge estimator. Sections 2 and 3

The present disclosure relates generally to optical flow algorithms. Section 1 of the present disclosure describes an optical flow algorithm with real-time performance and adequate accuracy for embedded vision applications. This optical flow algorithm is based on a ridge estimator. Sections 2 and 3 describe an obstacle detection algorithm that utilizes the motion field that is output from the optical flow algorithm. Section 2 is focused on unmanned ground vehicles, whereas section 3 is focused on unmanned aerial vehicles.

대표청구항 ▼

1. An apparatus, comprising: an optical flow sensor comprising a camera interface and an optical flow calculation module that calculates a sequence of optical flow fields using a ridge estimator; andan obstacle detection module comprising: a de-rotation sub-module processor comprising a first line b

1. An apparatus, comprising: an optical flow sensor comprising a camera interface and an optical flow calculation module that calculates a sequence of optical flow fields using a ridge estimator; andan obstacle detection module comprising: a de-rotation sub-module processor comprising a first line buffer, the de-rotation sub-module being configured to de-rotate the sequence of optical flow fields to obtain de-rotated optical flow fields;a de-translation sub-module processor comprising a second line buffer, the de-translation sub-module being configured to de-translate the de-rotated optical flow fields to obtain de-translated optical flow fields; anda post-processing sub-module processor that is configured to identify pixels with an inconsistent motion pattern as obstacles, wherein the inconsistent motion pattern is determined based on a residual motion component of a pixel compared to residual motion components of a majority of pixels of an image after de-translation;wherein the optical flow fields comprise νx and νy components, wherein the νy components of the optical flow fields are used for the de-rotation, the de-translation, and the identification, and wherein the νx components of the optical flow fields are not used for the de-rotation, the de-translation, or the identification. 2. The apparatus of claim 1, wherein the obstacle detection module is implemented in a field-programmable gate array (FPGA) or an application specific integrated circuit (ASIC). 3. The apparatus of claim 1, wherein the de-rotation of the sequence of optical flow fields comprises: using least squares estimation to calculate slope ratios, wherein a slope ratio is calculated for each row of the optical flow fields;calculating confidence indices, wherein a confidence index is calculated for each row of the optical flow fields; andaveraging the slope ratios over multiple rows, using the confidence indices as weights for the averaging. 4. The apparatus of claim 3, wherein: the calculation of the slope ratios is performed in accordance with βj=∑xi⁢vyi,j∑xi2;the calculation of the confidence indices is performed in accordance with cj=(∑xi⁢vyi,j)2∑xi2⁢∑vyi,j2; and the averaging is performed in accordance with β_=∑j=1N⁢βj⁢cj∑j=1N⁢cj, wherein (νxi,j,νyi,j) is an optical flow field at pixel (i,j), xi=i−n−1−68 [−n,n], 1=1,2, . . . , 2n+1, and N is the number of rows being averaged. 5. The apparatus of claim 1, wherein the de-translation of the sequence of optical flow fields comprises calculating, for each pixel in the de-rotated optical flow fields, a de-translated νy component. 6. The apparatus of claim 5, wherein the calculation of the de-translated νy component for pixel (i,j) in the de-rotated optical flow fields is performed in accordance with vyi,jD=vyi,jR-v_yiR=vyi,jR-∑k=j-mk=j+m-1⁢∑i=-ni=n-1⁢(vyi,kR)4⁢m⁢⁢n, wherein νyi,jR is the de-rotated vy component of optical flow field for pixel (i,j), and νyiR is a mean of de-rotated vy component of optical flow fields over 4 mn pixels. 7. The apparatus of claim 1, wherein the identification of the pixels with the inconsistent motion pattern comprises binarizing the de-rotated and de-translated νy components by applying a threshold. 8. The apparatus of claim 7, wherein the binarizing of the de-rotated and de-translated νy components is performed in accordance with binit⁡(x,y)={1,if⁢⁢vyΔ⁡(x,y)≥δ;0,if⁢⁢vyΔ⁡(x,y)<δ. 9. The apparatus of claim 1, wherein a slope ratio is estimated for each row, and a de-rotated νy component of the optical flow fields is obtained based on an averaged slope ratio over a predetermined number of rows. 10. The apparatus of claim 9, wherein a mean of de-rotated νy components of the optical flow fields over the predetermined number of rows is used as a translational component for each pixel. 11. The apparatus of claim 9, wherein the predetermined number of rows is eight.