On-vehicle image processing apparatus

1. An on-vehicle image processing apparatus comprising:

an image taking apparatus configured to generate image data by taking an image of a forward view of a vehicle;

an edge detection section configured to generate a detection image data including pixels of m rows and n columns (m and n are integers more than 2) based on the image data; and

a feature point detection section configured to detect at least one feature point based on the detection image data and output the at least one feature point to an output apparatus,

the feature point detection section categorizes the at least one feature point into a lane division line feature point indicating a point on a lane division line, a branch feature point indicating a branch of a lane, and a dead end and obstacle feature point indicating a dead end of a lane or an obstacle on a lane,

the detection image data includes a lane division line detection scanning region set for detecting the lane division line feature point, a branch detection scanning region set for detecting the branch feature point, and a dead end and obstacle detection scanning region set for detecting the dead end and obstacle feature point,

a distance between a vehicle loading the image taking apparatus and a position on a lane corresponding to a pixel in the detection image data becomes shorter from a first row to a m-th row of the detection image data, and

the dead end and obstacle detection scanning region is set in pixels positioned from the first row to a m1-th row of the detection image data, the branch detection scanning region is set in pixels positioned from the m1-th row to a m2-th row of the detection image data, and the lane division line detection scanning region is set in pixels positioned from the m2-th row to the m-th row of the detection image data, wherein 1<m1<m2<m.

2. The on-vehicle image processing apparatus according to claim 1 , wherein the feature point detection section includes:

a lane division line detection section configured to detect the lane division line by scanning the lane division line detection scanning region;

a branch detection section configured to detect the branch feature point by scanning the branch detection scanning region; and

a dead end and obstacle detection section configured to detect the dead end and obstacle feature point by scanning the dead end and obstacle detection scanning region.

3. The on-vehicle image processing apparatus according to claim 2 , wherein the lane division line detection section is configured to judge the lane division line feature point positioned in a left region of the detection image data as a left side lane division line feature point, and judge the lane division line feature point positioned in a right region of the detection image data as a right side lane division line feature point, and

the feature point detection section further includes:

a vehicle position estimation section configured to estimate a position of a vehicle loading the image taking apparatus based on the left side lane division line feature point and the right side lane division line feature point.

4. The on-vehicle image processing apparatus according to claim 3 , wherein the lane division line detection section is configured to scan pixels in the lane division line detection scanning region by one row at a time,

and when one of the left side lane division line feature point and the right side lane division line feature point on a certain row is detected and another one of the left side lane division line feature point and the right side lane division line feature point on the certain row is not detected, define a left or right half region in which the lane division line feature point is not detected as a feature point lacking half region, and

generate a complemented lane division line feature point positioned in a center of the feature point lacking half region in a scanning direction, and

the vehicle position estimation section is configured to estimate a position of a vehicle loading the image taking apparatus based on a left side or right side lane division line feature point detected on the certain row and the complemented lane division line feature point.

5. The on-vehicle image processing apparatus according to claim 1 , wherein the edge detection section is configured to:

calculate a weighted average of an intensity of a focused pixel and intensities of surrounding pixels surrounding the focused pixel by taking each of the pixels of m rows and n columns as the focused pixel;

perform a two-dimensional space filtering by generating a filtered image data including filtered pixels, each of the filtered pixels has an intensity determined based on the weighted average; and

generating a binarized filtered image data by binarizing each of pixels in the filtered image data by using a threshold intensity value.

6. The on-vehicle image processing apparatus according to claim 5 , wherein the edge detection section is configured to perform the two-dimensional space filtering by using a smoothing filter wherein the intensity of the focused pixel and the intensities of surrounding pixels are equal, or a laplacian filter.

7. The on-vehicle image processing apparatus according to claim 5 , wherein each of the pixels in the binarized filtered image data has a first pixel value or a second pixel value, and

the lane division line detection section is configured to detect a pixel having the first intensity as a candidate lane division line feature point, and judge the candidate lane division line feature point as the lane division line feature point when the candidate lane division line feature point is matched with a preset lane division line feature point pattern.

8. The on-vehicle image processing apparatus according to claim 5 , wherein each of the pixels in the binarized filtered image data has a first pixel value or a second pixel value, and

the branch detection section is configured to detect a pixel having the first intensity as a candidate branch feature point, and judge the candidate branch feature point as the branch feature point when the candidate branch feature point is matched with a preset branch feature point pattern.

9. The on-vehicle image processing apparatus according to claim 5 , each of the pixels in the binarized filtered image data has a first pixel value or a second pixel value, and

the dead end and obstacle detection section is configured to detect a pixel having the first intensity as a candidate dead end and obstacle feature point, and judges the candidate dead end and obstacle feature point as the dead end and obstacle feature point when the candidate dead end and obstacle feature point is matched with a preset dead end and obstacle feature point pattern.

10. The on-vehicle image processing apparatus according to claim 5 , wherein each of pixels of the binarized filtered image data has a binary value of 0 or 1, and

the on-vehicle image processing apparatus further comprises an edge correction and noise reduction section configured to:

count value 0 pixel having the binary value of 0 and value 1 pixel having the binary value of 1 among a focused binary pixel and surrounding binary pixels surrounding the focused binary pixel, by taking each of the pixels of the binarized filtered image data as a focused binary pixel;

generate a feature point detection image data by performing an edge correction and noise reduction process in which the binary value is replaced by a new binary value, wherein the new binary value is determined by a proportion of the counted number of the value 0 pixel to the counted number of the value 1 pixel, and a predetermined condition of an arrangement of the value 0 pixel and the value 1 pixel of the surrounding binary pixels for each of the focused binary pixel; and

output the feature point detection image data to the feature point detection section as the detection image data.

11. The on-vehicle image processing apparatus according to claim 1 , further comprising:

a memory; and

A/D converter configured to convert the image data generated by the image taking apparatus which is an analogue data into a digital data and stores the digital data in the memory,

wherein the edge detection section is configured to generate the detection image data from the digital data stored in the memory.

12. The on-vehicle image processing apparatus according to claim 1 , wherein the output apparatus is an automatic driving system for controlling a motion of a vehicle loading the image taking apparatus in response to the at least one feature point.

13. An image processing apparatus comprising:

an edge detection section configured to generate a detection image data including pixels of m rows and n columns (m and n are integers more than 2) by applying an image processing to an image data which is an image of a forward view of a vehicle;

a feature point detection section configured to detect at least one feature point based on the detection image data and output the at least one feature point to an output apparatus,

the feature point detection section categorizes the at least one feature point into a lane division line feature point indicating a point on a lane division line, a branch feature point indicating a branch of a lane, and a dead end and obstacle feature point indicating a dead end of a lane or an obstacle on a lane,

the detection image data includes a lane division line detection scanning region set for detecting the lane division line feature point, a branch detection scanning region set for detecting the branch feature point, and a dead end and obstacle detection scanning region set for detecting the dead end and obstacle feature point,

a distance between a vehicle loading the image taking apparatus and a position on a lane corresponding to a pixel in the detection image data becomes shorter from a first row to a m-th row of the detection image data, and

the dead end and obstacle detection scanning region is set in pixels positioned from the first row to a m1-th row of the detection image data, the branch detection scanning region is set in pixels positioned from the m1-th row to a m2-th row of the detection image data, and the lane division line detection scanning region is set in pixels positioned from the m2-th row to the m-th row of the detection image data, wherein 1<m1<m2<m.