JPWO2018029066A5

JPWO2018029066A5 -

Info

Publication number: JPWO2018029066A5
Application number: JP2019506644A
Authority: JP
Publication date: 2022-02-01
Anticipated expiration: 2037-08-03

Claims

A method (100) for detecting an LED light source within a series of frames for an environment.
The blinking characteristics of the LED light source are based on the frequency and rectification method of the AC signal supplied by the power grid.
Selecting at least one frame portion of interest in the series of frames (102) and
To detect fluctuations in the light intensity of at least one selected frame of interest (104) and
When the detected frequency of the fluctuation is within a predetermined frequency range, it is determined that one frame portion of interest is a candidate indicating an LED light source (106).
Including
Detecting the fluctuation of the light intensity (104) makes the characteristic value of the attention frame portion in one frame in a series of frames, and the characteristic value of the attention frame portion in a plurality of different frames in a series of frames. Including comparing with value
The characteristic value of the attention frame portion is a value indicating the light intensity of the attention frame portion.
The above comparison is based on the following mathematical formula.

T (i, j) indicates the comparison result, i indicates the characteristic value of the attention frame portion in one frame in a series of frames, and j indicates the characteristic value in one frame of a plurality of different frames in a series of frames. Indicates the characteristic value of the frame of interest of, where t indicates a constant,
When the characteristic value of the attention frame portion in one frame in a series of frames is smaller than the first threshold value jt, the first comparison result "0" is shown.
When the characteristic value of the attention frame portion in one frame in a series of frames is equal to or less than the second threshold value j + t, the second comparison result “1” is shown.
When the characteristic value of the attention frame portion in one frame in a series of frames is larger than the second threshold value j + t, the third comparison result “2” is shown.
The first threshold value jt and the second threshold value j + t are based on the characteristic values of the frame of interest in a plurality of other frames in a series of frames.
Generates a pattern from each of the results of comparing the characteristic values of the frame of interest in one frame in a series of frames with each of the characteristic values in one frame of multiple different frames in a series of frames. ,
Determining that one frame of interest portion is a candidate for an LED light source comprises comparing the generated pattern with a reference pattern indicating the presence of the LED light source within a series of frames for an environment. A method for detecting a light source.

The method of claim 1, wherein the reference pattern is obtained from a training dataset.

The method according to claim 1 or 2, wherein the characteristic value of the frame portion of interest is a pixel value of the frame portion of interest.

The method of any one of claims 1 to 3, wherein selecting the frame portion of interest (102) comprises repeatedly filtering at least a portion of a series of frames using a plurality of filter criteria.

Iteratively filtering at least a portion of a set of frames using multiple filter criteria involves determining a first set of frame portions, for each frame portion of the first set of frame portions. The method of claim 4, wherein the sum of the differences in pixel values between a frame portion within one frame of a frame and a frame portion of a plurality of other frames in a series of frames exceeds a threshold value.

Iteratively filtering at least a portion of a series of frames using multiple filter criteria further comprises determining from the first set of frame portions to the second set of frame portions, the second of the frame portions. 5. The method of claim 5, wherein the frame portion of the set indicates a predetermined color range.

Iteratively filtering a series of frames using multiple filter criteria further includes selecting one frame portion of the second set of frame portions as the frame portion of interest, and for the frame portion of interest a series of frames. The sixth aspect of the present invention, wherein the difference in the pixel value between the attention frame portion showing the highest pixel value in the frame and the attention frame portion showing the lowest pixel value in the series of frames exceeds the threshold value. Method.

Creating a connection frame part by combining the frame part of interest into multiple areas of a frame with at least one feature.
Determining multiple candidates for LED light sources and
When the connection frame portion includes one candidate representing the LED light source, it is determined that the connection frame portion represents the LED light source.
The method according to claim 6 or 7, further comprising.

The method according to any one of claims 1 to 8, wherein the frame portion of interest is one pixel.

The method according to any one of claims 1 to 9, wherein the series of frames comprises at least 5 frames.

The method according to any one of claims 1 to 10, wherein the frame rate of a series of frames is at least 100 frames per second.

The method according to any one of claims 1 to 11, wherein the predetermined frequency range is between 95 Hz and 105 Hz and between 115 Hz and 125 Hz.

A method of detecting a traffic signal lamp comprising at least one LED light source in a series of frames for an environment, comprising the method of any one of claims 1-12.

A vehicle (700) comprising a processing unit (710) that implements at least one of the method according to any one of claims 1 to 12 and the method for detecting a traffic signal lamp according to claim 13.

14. The vehicle of claim 14, further comprising a sensor (720) that produces a series of frames for the environment.

15. The vehicle of claim 15, wherein the sensor (720) produces a series of frames for the environment at a frame rate of at least 100 frames per second.

A machine containing a program having a program code that implements at least one of the method for detecting an LED light source according to any one of claims 1 to 12 and the method for detecting a traffic signal light according to claim 13. A machine-readable storage medium in which a program is executed by a computer unit or a processor.