Method and system for detecting vehicle position by employing polarization image

1. A method of detecting a vehicle position by employing a polarization image, comprising:

a polarization image obtaining step of capturing a polarization image by using a polarization camera;

a road surface detecting step of acquiring two road shoulders in the polarization image based on a difference between a road surface in the polarization image and each of the two road shoulders in the polarization image, and determining a part between the two road shoulders as the road surface;

a vehicle bottom detecting step of detecting at least one vehicle bottom from the road surface based on a pixel value difference between each wheel in the polarization image and the road surface in the polarization image; and

a vehicle position generating step of generating a vehicle position from the vehicle bottom based on a pixel value difference between a vehicle outline corresponding to the vehicle bottom in the polarization image and background in the polarization image,

wherein, the vehicle bottom detecting step includes,

a step of extracting candidate wheel pixels by carrying out, based on a predetermined pixel threshold value, binarization with regard to the road surface in the polarization image; and

a step of adopting scan windows, whose size gradually increase in proportion from top to bottom in the polarization image, to scan the whole road surface in the polarization image at least from top to bottom, and calculating elements of a co-occurrence matrix corresponding to each of the scan windows so as to determine at least one scan window, whose corresponding co-occurrence matrix has an element whose value is greater than a predetermined threshold value, as the vehicle bottom.

2. The method according to claim 1 , wherein, the vehicle bottom detecting step further comprises:

a step of performing a smoothing process with regard to the candidate wheel pixels so as to reduce noise.

3. The method according to claim 1 , wherein, the acquiring two road shoulders in the polarization image comprises:

a step of obtaining edge pixels by utilizing Canny edge detection; and

a step of acquiring two lines serving as the two road shoulders by using Hough transform based on the edge pixels.

4. The method according to claim 3 , wherein, the vehicle position generating step comprises:

a step of obtaining, based on a left side and a right side of the vehicle bottom as well as prior knowledge of a vehicle, a left side and a right side of the vehicle by extending the left side and the right side of the vehicle bottom upwards; and

a step of acquiring, based on a difference between a polarized nature of a top of the vehicle and a polarized nature of the background, a weak edge serving as a vehicle top so as to form a rectangle of the vehicle outline by connecting the left and right sides of the vehicle, a bottom side of the vehicle bottom, and the vehicle top.

5. The method according to claim 3 , wherein,

the detecting at least one vehicle bottom comprises:

calculating elements of co-occurrence matrixes, and

carrying out binarization based on a predetermined threshold value.

6. A system comprising:

a memory unit;

and a processor,

the memory unit storing instructions that, when executed by the processor, cause the processor to perform operations for detecting a vehicle position by employing a polarization image, the operations including, a polarization image obtaining step of capturing a polarization image by using a polarization camera;

a road surface detecting step of acquiring two road shoulders in the polarization image based on a difference between a road surface in the polarization image and each of the two road shoulders in the polarization image, and determining a part between the two road shoulders as the road surface;

a vehicle bottom detecting step of detecting at least one vehicle bottom from the road surface based on a pixel value difference between each wheel in the polarization image and the road surface in the polarization image;

and a vehicle position generating step of generating a vehicle position from the vehicle bottom based on a pixel value difference between a vehicle outline corresponding to the vehicle bottom in the polarization image and background in the polarization image,

wherein, the vehicle bottom detecting step includes,

a step of extracting candidate wheel pixels by carrying out, based on a predetermined pixel threshold value, binarization with regard to the road surface in the polarization image; and

a step of adopting scan windows, whose, size gradually increase in proportion from top to bottom in the polarization image, to scan the whole road surface in the polarization image at least from top to bottom, and calculating elements of a co-occurrence matrix corresponding to each of the scan windows so as to determine at least one scan window, whose corresponding co-occurrence matrix has an element whose value is greater than a predetermined threshold value as the vehicle bottom.

7. A method of detecting a vehicle position by employing a polarization image, comprising:

a polarization image obtaining step of capturing a polarization image by using a polarization camera;

a road surface detecting step of acquiring two road shoulders in the polarization image based on a difference between a road surface in the polarization image and each of the two road shoulders in the polarization image, and determining a part between the two road shoulders as the road surface;

a vehicle bottom detecting step of detecting at least one vehicle bottom from the road surface based on a significant pixel value difference between each wheel in the polarization image and the road surface in the polarization image; and

a vehicle position generating step of generating a vehicle position from the vehicle bottom based on a pixel value difference between a vehicle outline corresponding to the vehicle bottom in the polarization image and background in the polarization image,

wherein, the acquiring two road shoulders in the polarization image includes,

a step of obtaining edge pixels by utilizing Canny edge detection; and

a step of acquiring two lines serving as the two road shoulders by using Hough transform based on the edge pixels, and

wherein the detecting at least one vehicle bottom includes,

calculating elements of co-occurrence matrixes, and

carrying out binarization based on a predetermined threshold value.