JP2012176641A - Detection apparatus for parking frame - Google Patents

Abstract

PROBLEM TO BE SOLVED: To extract a line of a parking frame painted in yellow or orange on a gray concrete road, in a detection apparatus for the parking frame, and to detect the parking frame using conventional edge detection without distinguishing from a process of a line of the parking frame painted in white.SOLUTION: A control device 14 includes: a yellowish extraction device 14B for extracting a yellowish part on an image on the basis of an RGB signal of the image which is imaged by an imaging device 12; and a brightness addition device 14C for adding a preset value to the brightness of a pixel of the yellowish part extracted by the yellowish extraction device 14B. The parking frame is detected based on the image to which the preset value is added by the brightness addition device 14C.

Description

  The present invention relates to a parking frame detection device, and more particularly to a parking frame detection device that detects a parking frame (parking space) from an image captured by an imaging means (camera) attached to a vehicle.

Some vehicles are equipped with a parking frame detection device in order to assist driving during parking. This parking frame detection device detects a parking frame (parking space) from an image captured by an imaging means (camera) attached to the rear portion of the vehicle during a parking retreat operation, and assists the driving operation of the driver.
In such a parking frame detection device, in order to detect a parking frame line in the image, a two-dimensional operator such as a Sobel filter or Laplacian is usually used when detecting the edge of the parking frame line in the image. After that, binarization using a threshold value is performed, but it is often difficult to uniquely determine this threshold value.
And in order to stop the own vehicle on a parking frame, when detecting a parking frame line, it is necessary to detect the edge of the parking frame line which appears in the vertical direction in an image. In detecting the edge of the parking frame line, it is necessary to eliminate the influence of road surface patterns, shadows, lighting, and the like as much as possible.

JP 2006-306224 A JP 2006-290051 A

The parking assistance device and the parking assistance method according to Patent Document 1 obtain the driving operation that should be performed first from the current position among the driving operation patterns determined to be parking as a driving operation necessary for parking the parking frame. Information indicating whether the parking frame can be parked and information for identifying the driving operation to be performed first among the driving operations necessary for parking the parking frame are displayed together with the parking frame.
The parking support device and the parking support method according to Patent Document 2 set a reverse start position when a vehicle is moved backward and parked in consideration of a psychological load when the host vehicle approaches an obstacle. .

By the way, in the conventional parking frame detection device, as shown in the input image of FIG. 27, in the parking frame line 103 drawn in yellow or orange on the concrete gray road surface 102 of the parking lot 101, When the line profile (luminance change) is measured on the measurement line L that is linear in the horizontal direction, as shown in the line profile (luminance change) in FIG. Since there is no difference in brightness (brightness exceeding a preset threshold) (in this example, the yellow or orange parking frame line 103 is lower in brightness than the gray road surface 102), the parking frame line 103 There was a problem that could not be extracted.
FIG. 29 shows a conventional example in which the edge of the parking frame line is detected, but it is understood that the edge is not sufficiently detected.
Furthermore, in the above-mentioned patent document 1, it is defined that the parking frame line is drawn with a white line. In the above-mentioned patent document 2, in the detection of the edge of the parking frame line, “the image processing unit uses camera image data. The edge strength and edge shape of the vehicle and the edge strength and edge shape indicating the parking frame line obtained in advance are compared to detect an edge. " Had the inconvenience that they could not respond.
Furthermore, the edge of the parking frame line could not be detected sufficiently with only the ordinary Sobel filter shown in FIG.

  Therefore, an object of the present invention is to eliminate the need for threshold value determination means (adaptive control type) when detecting the edge of the parking frame line, simplify the processing, and remove edges generated at both ends of the parking frame line on the road surface. A parking frame detection device that eliminates the effects of patterns, shadows, lighting, etc. as much as possible to increase reliability, and can stably detect parking frames even when there is little difference in brightness between the road surface and parking frame lines It is to provide.

  The present invention relates to a parking frame detection apparatus provided with an imaging means for imaging the periphery of a vehicle and a control means for detecting a parking frame from an image captured by the imaging means, wherein the control means is an image captured by the imaging means. A yellow color extracting means for extracting a yellow color portion on the image based on the RGB signal, and a luminance adding means for adding a preset value to the luminance of the pixel of the yellow color portion extracted by the yellow color extraction means And a parking frame is detected from an image added with a preset value by the luminance adding means.

  The parking frame detection device of the present invention eliminates the need for threshold value determination means (adaptive control type) when detecting the edge of the parking frame line, simplifies the processing, and detects the edges generated at both ends of the parking frame line on the road surface. Even when the influence of the pattern, shadow, lighting, etc. is eliminated as much as possible to increase reliability and the difference in brightness between the road surface and the parking frame line is small, stable parking frame detection can be performed.

FIG. 1 is a flowchart of parking frame detection. Example 1 FIG. 2 is a control block diagram of the parking frame detection device. Example 1 FIG. 3 is a plan view showing a parking frame and a vehicle in the parking lot. Example 1 FIG. 4 is a diagram showing a normal line profile (RGB) (luminance change). Example 1 FIG. 5 is a diagram illustrating an R component image among images separated for each of the R component, the G component, and the B component. Example 1 FIG. 6 is a diagram illustrating a B component image among images separated for each of the R component, the G component, and the B component. Example 1 FIG. 7 is a diagram illustrating an image of a difference between the R value and the B value. Example 1 FIG. 8 is a diagram showing the line profile (luminance change) of FIG. Example 1 FIG. 9 is a diagram illustrating an image of a detection result of a parking frame line (yellow color). Example 1 FIG. 10A is a diagram showing an input image when applied to a white parking frame line. 10 (B) is a diagram illustrating an image of a determination processing result when applied to a white parking frame line. Example 1 FIG. 11 is a diagram illustrating an image when a luminance value is added to a parking frame line (yellow color). Example 1 FIG. 12 is a diagram illustrating an image of the detection result of the edge of the parking frame line. Example 1 FIG. 13 is a diagram showing an image when applied to a white parking frame line. Example 1 FIG. 14A shows a difference operator. FIG. 14B is a diagram illustrating an operator of the white line double-ended differential value enhancement filter. Example 1 FIG. 15 is a diagram showing a flow of white line detection processing. Example 1 FIG. 16 is a diagram showing an image before the edge detection process. Example 1 FIG. 17 is a diagram illustrating an image when the threshold is set to “1” in edge detection. Example 1 FIG. 18 is a diagram showing an image when the threshold is set to “10” in edge detection. Example 1 FIG. 19 is a diagram showing an image when the threshold is set to “20” in edge detection. Example 1 FIG. 20A is a diagram illustrating an image before edge detection when “setting 1” is set to “brightness of wall (320.240) 180”. FIG. 20B is a diagram showing an image after edge detection when “setting 1” is set to “brightness of wall (320.240) 180”. Example 1 FIG. 21A is a diagram showing an image before edge detection when “setting 2” is “brightness of wall (320.240) 62”. FIG. 21B is a diagram showing an image after edge detection when “setting 2” is set to “brightness of wall (320.240) 62”. Example 1 FIG. 22A is a diagram illustrating an image before edge detection when “brightness 32 of the wall (320.240)” is set as “setting 3”. FIG. 22B is a diagram illustrating an image after edge detection when “setting 3” is set to “brightness of wall (320.240) 32”. Example 1 FIG. 23 is a diagram showing an image when the threshold value is “5” in “setting 3” of FIG. Example 1 FIG. 24 is a flowchart of parking frame detection. (Example 2) FIG. 25 is a control block diagram of the parking frame detection device. (Example 2) FIG. 26 is a diagram showing an image of edge detection when the detection method of this embodiment is applied to a Sobel filter. (Modification) FIG. 27 shows a conventional input image. (Conventional example) FIG. 28 is a diagram showing the line profile (luminance change) of FIG. (Conventional example) FIG. 29 shows a conventional edge detection image. (Conventional example) FIG. 30 is a diagram showing an edge image by a normal Sobel filter. (Conventional example)

  The present invention eliminates the need for threshold determination means (adaptive control type) when detecting the edge of the parking frame line, simplifies the processing, and removes the edges generated at both ends of the parking frame line from the road surface pattern, shadow, The luminance signal and color difference signal of the captured image are used for the purpose of stable parking frame detection even when the luminance difference between the road surface and the parking frame line is small by eliminating the influence of lighting etc. as much as possible. Are converted into RGB signals, a yellow portion on the image is extracted based on the RGB signals, a preset value is added to the luminance of the pixels of the yellow portion, and a parking frame is added from the added image. It is realized by detecting.

1 to 23 show Embodiment 1 of the present invention.
In FIG. 3, 1 is a vehicle and 2 is a parking lot.
The parking lot 2 is made of concrete and has a gray road surface 3.
A parking frame (parking space) 4 is defined on the road surface 3 of the parking lot 2.
The parking frame 4 includes a parking frame line 5 that divides the parking range of the vehicle 1, and a left parking frame line parallel to the longitudinal direction (yellow line or orange line) 6 and a right parking frame line (yellow line or orange line). ) 7 and the left parking frame line 6 / right parking frame line 7 are connected to the left front side edge / right front side edge of the left parking frame line 6 / right parking frame line 7 in the short direction perpendicular to the left parking frame line 6 / right parking frame line 7. The front parking frame line (yellow line or orange line) 8 is formed.
In addition, the parking frame 4 has left and right wheel stops 9 and 10 arranged in series in the lateral direction as a wheel stopper at the rear portion.

A parking frame detection device 11 is mounted on the vehicle 1.
As shown in FIG. 2, the parking frame detection device 11 includes an ultra-wide-angle imaging unit (camera) 12 that images the periphery of the vehicle, an image recording unit 13 that communicates with the imaging unit 12, and an image recording unit 13. The control means 14 which contacted is provided.
The imaging means 12 is composed of an imaging device such as a super-wide-angle back camera attached to the rear part of the vehicle 1 and images the periphery of the vehicle.
The image recording unit 13 includes a recording device such as a video decoder, records the image captured by the imaging unit 12 and outputs the image to the control unit 14.

The control unit 14 extracts the parking frame line 5 from the image captured by the imaging unit 12 and detects the parking frame 4. The control unit 14 detects RGB (red) from the luminance signal and the color difference signal of the image captured by the imaging unit 12. RGB conversion means 14A for converting into a green / blue signal, yellow system extraction means 14B for extracting a yellow system portion on the image based on the RGB signal of the image captured by the imaging system 12, and this yellow system extraction means 14B. Luminance adding means 14C for adding a preset value to the brightness of the extracted yellow-based pixel, and detecting the parking frame 4 from the image added with the preset value by the brightness adding means 14C. .
Further, the yellow color extracting means 14B is such that the R value of the RGB signal converted by the RGB converting means 14A is larger than the B value, the G value is larger than the B value, and the B value is subtracted from the R value. Pixels having a larger value than the preset value are extracted as yellow parts.

Next, a method of using color information together for extracting the yellow or orange parking frame 5 according to this embodiment will be described.
In general, when there is no significant difference in luminance information between the gray road surface 102 and the yellow parking frame line 103 as shown in FIG. 27 and only the color information is different, a difference in color information with adjacent pixels is detected. There is known a method for extracting a region having the same color.
However, in an actual outdoor environment, depending on conditions such as sunlight and lighting, for example, in a dark place, it looks yellow or even a color close to brown, and when it is bright, it becomes a color close to white. For this reason, it is difficult to extract the parking frame line neatly even if the video signal is confirmed after being separated into an R component, a G component, and a B component from the image captured by the imaging unit.
For example, when a part of a parking frame line is illuminated at night when the surroundings are dark, the R component may be saturated due to the influence of auto gain control of the imaging means, etc. The ratio of the component and the component of B is not constant.
Also, yellow is ideally (R, G, B) = (255, 212, 0), but in an actual video, the B value is extremely smaller than the R value and the G value. There are few. For this reason, it is difficult to extract the edge of the yellow parking frame line using only color information set in advance.
As an example, FIG. 4 shows the line profile (luminance change) of FIG. 27 measured for the R value, G value, and B value.
In FIG. 4, the R value, the G value, and the B value are substantially the same except for the left parking frame line and the right parking frame line. However, in the left parking frame line and the right parking frame line, the R value and the G value are The B value magnitude relationship (R value is largest, then G value, B value is smallest) is the same, but the ratio and value are not constant, and R value, G value, B value By providing a range for each, it can be confirmed that it is difficult to extract a yellowish color. Therefore, it is difficult to extract a yellow parking frame line using conditions such as a certain range of R value, G value, and B value.
Therefore, since the parking frame line is yellow or orange, the condition that the R value and the G value are larger than the B value, and the condition that the difference between the R value and the B value is greater than or equal to the first threshold value. We propose a correction method for extracting the parking frame line.

Here, as shown in FIG. 4, it is understood that R> B, R> G, and G> B in the orange parking frame line.
Based on these conditions, first, it is considered to extract a parking frame line on condition that R> B and G> B.
R> G is not used. Ideally, yellow is (R, G, B) = (255, 212, 0) and R> G, but R = G or R <G This area is excluded here because it may appear yellow even in the area of.
For example, when the R value, the G value, and the B value on the left and right parking frame lines in FIG. 4 are confirmed, the B value is not close to zero (0) and is a high value of about “170”. I understand that. This is because, in FIG. 25, since the entire screen is bright and whitish, the R value, G value, and B value are biased on the white side.
Also, under the conditions of R> B and G> B only, an erroneous region is extracted in a region of white, gray, black, etc., with colors having substantially the same R value, G value, and B value. is there.
Therefore, among the images separated for each of the R value, G value, and B value in FIG. 27, those relating to the R value and the B value and the difference image thereof are shown in FIGS.
In FIG. 7, it can be seen that the portion of the parking frame line is whitish compared to the surrounding area.
Therefore, when the line profile (luminance change) on the measurement line L in the straight line (lateral direction) in FIG. 7 is measured, it can be seen that FIG. 8 is obtained.

In FIG. 8, it can be seen that peaks P and P having a luminance value of about 50 appear in the left and right parking frame portions.
Therefore, a condition that | R−B | (absolute value) as a difference result is larger than the first threshold is added, and a parking frame line is extracted.
Since there is a condition of R> B first, | R−B |> (first threshold value) or | R−B | ≧ (first threshold value) can be set.
Here, when the first threshold value is “15”, the determination processing with R> B, G> B, and | R−B |> (first threshold value) is applied to FIG. 25. The extraction result is shown in FIG.
In FIG. 9, a part that is partially different is also extracted, but is not linear, and is eliminated by subsequent straight line recognition processing such as Hough transform or isolated point removal.
Here, when the above conditions, R> B, G> B, | R−B |> (first threshold value) are applied to a case where a white line is drawn on a general concrete road surface, FIG. (A) and FIG. 10 (B) are obtained.

From FIG. 10B, it can be seen that such a method cannot extract a white line (does not affect the detection of the white line).
However, in the case of a white line, an edge can be extracted by a process such as comparison of luminance values with adjacent pixels by a conventional method or the like. The yellow parking frame line is also devised so that the edge can be extracted by the same process as the white line.
The reason why the yellow parking frame line cannot be extracted by the edge extraction method applied to the conventional white line is that the luminance difference between the parking frame line and the background (road surface) is as shown in FIG. This is due to the fact that there are few.
That is, if the yellow parking frame line is sufficiently brighter (the brightness value is higher) than the surroundings, edge detection using the same brightness value as in the past can be applied.
Therefore, in this embodiment, as shown in FIG. 11, a certain value is added to the luminance value (Y value in YCbCr) of the extracted yellow parking frame line. Here, the edge detection process is performed based on the difference from the adjacent pixel value.
As an example, “30”, which is about three times the threshold value “10”, is added to the luminance value. This is because, as shown in FIG. 8, the yellow parking frame line extracted by this embodiment has a peak value P / P rather than a constant luminance value, so that the region is equal to or higher than the first threshold value. This is to make everything bright enough for the surroundings (road surface).
The result of adding the luminance values is shown in FIG. FIG. 12 shows an edge extraction result using a difference from an adjacent pixel value using FIG.
From FIG. 12, it can be confirmed that the edge of the parking frame line can be detected as compared with FIG. In this case, the specific threshold values vary depending on the characteristics of the imaging means (camera) to be used, the image recording device (video decoder), etc., and need to be adjusted accordingly.

FIG. 13 shows an edge detection result when the detection process according to this embodiment is applied to the white parking frame line shown in FIGS. 10 (A) and 10 (B). In FIG. 13, it can be confirmed that there is no adverse effect on edge detection.
Here, the conditions of R> B, G> B, and | R−B |> (first threshold value) result in the extraction of a region including colors from yellow to red. Usually, since it is rare that a red line is drawn on the road surface as a parking frame line, the purpose can be achieved under the above conditions, but if a problem occurs due to the extraction of red, | R−G | < A condition (second threshold) may be provided. This uses the fact that the R value and the G value are substantially equal in yellow, and that the R value increases with respect to the G value as it approaches red.

Next, the detection of the parking frame according to this embodiment will be described more specifically below.
The NTSC signal output from the imaging means 12 is transmitted to the ITU-R BT. 601 (International Telecommunication Union Radiocommunication Sector), etc. The output signal from the decoder of the image recording device 13 is generally YCbCr. This YCbCr signal is composed of a luminance signal Y and color difference signals Cb and Cr.
Here, in order to apply the above-described conditions, R> B, G> B, | R−B |> (first threshold value), conversion from the YCbCr signal to the RGB signal is performed.
The conversion formula to this RGB signal is
ITU-R BT. From 601
R = Y + 1.40200 * Cr
G = Y−0.34414 * Cb−0.71414 * Cr
B = Y + 1.77200 * Cb
It is.
Also, ITU-R BT. From 601, the value of the YCbCr signal is represented by digital 8 bits, but it is not 0-255, but Y is 16-235, CbCr is 16-240, while the RGB signal is represented by 0-255. Conversion to scale is required.
In this embodiment, accurate RGB values are not necessarily required, and it is only necessary to detect the magnitude relationship of RGB values and the difference between the R value and the B value. Therefore, it is not necessary to strictly convert the YCbCr signal to the RGB signal. A simpler conversion can be used.
For example,
R = Y + 1.5Cr
G = Y-0.25Cb-0.75Cr
B = Y + 1.5Cb
Even if the first threshold is set to about “15”, the same result can be obtained.
As described above, by simplifying the equation, the coefficient can be set to “0.5”, “0.25”, “0.75”, “1.5”, etc. Therefore, there is an advantage that it is not necessary to use division that takes time for processing.
Unlike the RGB signal, the YCbCr signal is output as Cb0, Y0, Cr0, Y1, Cb1, and Y2 because the color differences Cb and Cr are handled in common with the adjacent pixels with respect to the luminance value Y of each pixel. Therefore, it is necessary to handle in units of Cb0, Y0, Cr0, Y1. Y1 uses Cr0 which is common to Y0, but Cb is not common to Y0 and uses Cb1 output next to Y1. For this reason, conversion to RGB signals and yellow parking frame line extraction processing are performed for Cb0, Y0, and Cr0, and in response to the result, luminance value addition processing is performed in common for Y0 and Y1.
If luminance values are not added to both Y0 and Y1, there is a problem that an edge that is not originally between Y0 and Y1 is extracted in the subsequent edge detection processing.
However, by adding luminance values for both Y0 and Y1, the luminance values of all yellow areas are increased, so that an edge is detected in the contour as shown in FIG.
In addition, you may perform said process exactly like Cb0, Y0, Cr0 and Cr0, Y1, Cb1.

Further, the reason for extracting the edge based on the difference from the adjacent pixel will be described together with the idea process as follows.
As a processing example of the conventional method, a method using a white line double-ended differential value enhancement filter is shown below. In this method, the differentiation of the image is obtained, and a white line having the same width as the white line width of the road is extracted using a white line both-end differential value enhancement filter.
Here, since differentiation can be replaced with a difference, a difference image is obtained using the difference operator of FIG.
The white line both end differential value enhancement filter is an operator having the same width as the road white line width (number of pixels), and the sign is inverted at the center of the operator width as shown in FIG.
With the above processing, a result as shown in FIG. 15 is obtained. Therefore, by selecting an appropriate threshold, only features having the same width as the white line on the road can be extracted. A discriminant analysis method or the like is used for selection of the threshold value.
Similarly, the differential (difference) processing used in the conventional technology is the differential processing in the scanning line direction (screen horizontal direction).
However, in order to use the white line both-end differential value emphasis filter, it is necessary to once record an image (scanning line) in a storage means (memory), so that resource is required. Further, when the white line both-end differential value emphasis filter is used, the threshold level fluctuates as shown in FIG. 15. Therefore, it is necessary to adaptively control the threshold according to each position of the image. For this purpose, processing such as discriminant analysis is required. However, these processing are very heavy, and enormous resources are required even in parallel processing.
Therefore, “the difference image is the difference between the adjacent pixels, and it is rare that the brightness (illumination condition) of the adjacent pixels and the pixel is greatly different” (exception: edge by shadow), and generally Since the parking frame line such as the white line drawn on the parking frame is considered to have a clear contrast with the road surface so that the driver can easily identify it, the threshold value is set to a constant value.
16 to 19 show the edge detection results when the images before processing and the threshold values are “1”, “10”, and “20”. In this case, all are displayed (white → black, black → white) regardless of the polarity of the edge.
The threshold “1” in FIG. 17 shows the influence of the edge due to the fine texture (the light is reflected on the unevenness). However, the threshold “10” in FIG. 18 and the threshold “20” in FIG. It has been.
Here, in order to confirm the influence of the luminance change, the results when the lens aperture is changed at the threshold value “10” in FIG. 18 are shown in FIGS. In FIGS. 20 to 22, it can be seen that it is possible to cope with darkness of about ¼ of the dynamic range.
Here, in the case of “Setting 3” in FIG. 22, the edge cannot be detected as it approaches the wall at the back of the image. However, if the threshold value is lowered to “5”, the edge can be detected, but the noise can be detected. Will increase the influence (see FIG. 23). From this, it can be confirmed that even if the discriminant analysis method is used to determine the threshold value of this method, the effect is small.

Next, parking frame detection control according to this embodiment will be described based on the flowchart of FIG.
As shown in FIG. 1, when the program is started by the control means 14 (step A01), the image pickup means 12 picks up an image behind the vehicle, records the image picked up by the image pickup means 12, and A video signal (YCbCr) as a signal is input (step A02) and converted into an RGB signal (step A03).
Then, it is determined whether the R value of the RGB signal is larger than the B value and the G value is larger than the B value (step A04).
If this step A04 is YES, it is determined whether or not | R−B |> first threshold value (step A05).
If this step A05 is YES, a predetermined value is added to the luminance value (step A06).
After step A06, if step A04 is NO, or if step A05 is NO, the program is ended (step A07).

24 and 25 show Embodiment 2 of the present invention.
In the second embodiment, portions that perform the same functions as those of the first embodiment will be described with the same reference numerals.
In the parking frame detection device 1 of the first embodiment described above, the output signal from the video decoder of the image recording means 13 is YCbCr. However, some video decoders have RGB output.
As shown in FIG. 25, in the parking frame detection device 1 of the second embodiment, the output signal from the video decoder of the image recording means 13 is RGB.
Therefore, the control means 14 extracts a yellowish-colored extraction means 14B for extracting a yellowish-colored portion on the image based on the RGB signals of the image taken by the imaging means 12, and a YCbCr signal from the RGB signals of the image taken by the imaging means 12. YCbCr converting means 14D for converting to YCbCr, and luminance adding means for adding a preset value to the luminance of the yellow portion pixel extracted by the yellow extracting means 14B to the image converted by the YCbCr converting means 14D 14C, and the parking frame 4 is detected from an image added with a value set in advance by the luminance adding means 14C.

Next, parking frame detection control according to the second embodiment will be described based on the flowchart of FIG.
As shown in FIG. 24, when the program is started in the control means 14 (step B01), the image pickup means 12 picks up an image behind the vehicle, records the image picked up by the image pickup means 12, and A video signal (RGB) as a signal is input (step B02).
Then, it is determined whether or not the R value of the RGB signal is larger than the B value and the G value is larger than the B value (step B03).
If this step B03 is YES, it is determined whether or not | R−B |> first threshold value (step B04).
If this step B04 is YES, it is converted into a YCbCr signal (step B05), and a predetermined value is added to the luminance value (step B06).
After step B06, if the step B03 is NO or if the step B04 is NO, the program is ended (step B07).

Although the embodiments of the present invention have been described above, the configuration of the above-described embodiments will be described for each claim.
First, in the first aspect of the present invention, the control unit 14 includes a yellow system extraction unit 14B that extracts a yellow system part on the image based on the RGB signal of the image captured by the imaging unit 12, and the yellow system extraction. A luminance adding means 14C for adding a preset value to the luminance of the yellow portion pixel extracted by the means 14B, and a parking frame is obtained from the image added with the preset value by the luminance adding means 14C. To detect.
Thereby, the parking frame line 5 drawn in yellow or orange on the concrete road surface 3 can be extracted. Furthermore, the parking frame 4 can be detected by the conventional edge detection without distinguishing from the processing for the parking frame line drawn in white.
In the invention according to claim 2, the yellow color extracting means 14B is a value obtained by subtracting the B value from the R value and the R value of the RGB signal is larger than the B value, the G value is larger than the B value. A pixel having a value larger than a preset value is extracted as a yellowish portion.
Thereby, the yellow or orange parking frame line 5 drawn on the concrete road surface 3 can be extracted with high accuracy.

FIG. 26 shows a modification of the present invention.
This modification can be applied not only to edge detection based on a difference but also to a normal Sobel file. That is, the edge cannot be confirmed only with the normal Sobel file in FIG. 30, but in FIG. 26, it can be seen that the edge of the parking frame line can be detected by using the detection method according to this embodiment in the Sobel file. .

  The parking frame detection device according to the present invention allows a driver to provide an appropriate image and can be applied to various vehicles.

1 Vehicle 2 Parking 3 Road surface of parking 4 Parking frame (parking space)
DESCRIPTION OF SYMBOLS 5 Parking frame line 6 Left parking frame line 7 Right parking frame line 8 Front parking frame line 11 Parking frame detection apparatus 12 Imaging means 13 Image recording equipment 14 Control means 14A RGB conversion means 14B Yellow type extraction means 14C Brightness addition means

Claims (2)

Imaging means for imaging the surroundings of the vehicle;
In the parking frame detection apparatus provided with a control unit for detecting a parking frame from an image captured by the imaging unit, the control unit is configured to display a yellowish portion on the image based on an RGB signal of the image captured by the imaging unit. And a luminance adding means for adding a preset value to the luminance of the pixels of the yellow portion extracted by the yellow extracting means. The luminance adding means A parking frame detection device for detecting a parking frame from an image to which a value is added.   The yellow color extracting means is configured such that the R value of the RGB signal is larger than the B value, the G value is larger than the B value, and a value obtained by subtracting the B value from the R value is larger than a preset value. The parking frame detection device according to claim 1, wherein a large pixel is extracted as a yellow portion.

Images