JPH1196367A - Method and device for detecting parked vehicle - Google Patents

Abstract

PROBLEM TO BE SOLVED: To suppress the halation of a high luminance part by finding the differential image of luminance between an initial image and the other respective images and preparing an integrated image while weighing an image digitized corresponding to that luminance difference. SOLUTION: A color video signal from a color camera 1 is converted to three primary color signals in R, G and B by an RGB decoder 201 for the unit of a frame, and these R, G and B signals are converted to digital signals by an A/D converter 202. That digitized image is converted to the luminance image by a 1st luminance converting processing part 204 for the unit of each frame, the initial luminance image to be a reference is selected out of provided plural luminance images, the differential images of luminance between the initial luminance image and the other respective luminance images are respectively found by a differential processing part 206, and the integrated image can be provided by an RGB selective integrating processing part 208 while weighing the image digitized corresponding to that found luminance difference.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and an apparatus for detecting a parked vehicle which is used, for example, in detecting a parking situation of a vehicle by photographing a parking lot. The present invention relates to a method and an apparatus for detecting a parked vehicle in which halation due to a headlight or the like having a high luminance is suppressed, and a moving vehicle or a headlight does not cause a blur in an image due to a trail in a screen.

[0002]

2. Description of the Related Art Conventionally, when monitoring a parked vehicle at night using a camera, simply increasing the amount of incident light lowers the S / N ratio and makes it impossible to detect the vehicle due to halation. Many methods for detecting a vehicle using information have been adopted. However, information such as headlights and tail lamps is spot-like information, and the amount of information is small,
There is a problem that the detection accuracy is lower than that in daytime vehicle detection.

[0003] The inventors of the present invention have conducted intensive studies,
A method and an apparatus for detecting a parked vehicle that solves the above-mentioned problem have been proposed in Japanese Patent Application No. 8-072573. This prior invention is
A vehicle such as a parking lot is photographed, a digital image obtained by digitizing the video signal is integrated in the time axis direction to create an integrated image, and a brightness converted image is created from the integrated image by brightness conversion by histogram equalization processing. It is configured to detect a parked vehicle by performing image processing on the luminance conversion image, and it is possible to create an image in which halation in a high luminance portion is suppressed and contrast in a low luminance region is emphasized. The detection accuracy of the vehicle in was significantly improved.

[0004]

However, the prior invention has the following problems. That is,
In the prior invention, since the integrated image is obtained by simply integrating the captured image in the time axis direction, a moving object,
For example, if there is a moving vehicle, the image of the vehicle body or the headlights will be blurred due to the tail, and the vehicle may be erroneously detected.

The present invention has been made to solve such a problem. As in the prior application, it is possible to suppress halation in a high-luminance portion and to enhance the contrast of a vehicle in a low-luminance region. Moreover, an object of the present invention is to provide a method and a device for detecting a parked vehicle, in which a moving vehicle, a headlight or the like hardly trails and causes image blurring.

[0006]

In general, simply integrating an image requires moving objects (particularly vehicles and vehicle lights).
Are simply added, and the image is blurred so as to leave a trail as described above. Therefore, the present inventor has noticed that there is a luminance difference between the vehicle and the road surface in order to reduce this phenomenon, and first calculates the luminance difference between the next image and the initial image based on the captured initial image. If the difference is small, the weight of the next image at the time of integration is increased, while if the difference is large, the weight of the next image at the time of integration is reduced and the integration is performed. By performing this processing in the direction of the time axis, a moving object having a large dependence on the initial image remains at the same place, and at the same time, it is possible to improve the S / N ratio in other areas as in the case of the prior application. The present invention has solved the above-mentioned problem based on such a concept.

That is, in the parked vehicle detection method according to the present invention, an integrated image is created by integrating a digitized image of a video signal of a vehicle detection target area in the time axis direction, and a luminance conversion process is performed from the integrated image. In the parked vehicle detection method of detecting a parked vehicle by creating a brightness conversion image in which the contrast of the low brightness region is emphasized and performing image processing on the brightness conversion image, a brightness conversion is performed from a plurality of images of the digitized image. It is characterized in that a reference initial image is selected, a luminance difference image between the initial image and each of the other images is obtained, and an integrated image is created while weighting the digitized image according to the luminance difference. It is assumed that.

Further, a color video signal obtained by photographing the vehicle detection target area is converted into R, G, B signals, and the R, G, B images composed of the R, G, B signals are converted into luminance images in frame units. Selecting an initial luminance image as a reference from the obtained plurality of luminance images to obtain a difference image between the initial luminance image and each of the other luminance images, based on each of the obtained difference images. The weighted R, G, B images are integrated in the time axis direction while weighting the R, G, B images to generate weighted R, G, B integrated images, and the obtained weighted R, G, B images are obtained. After converting the G, B integral image into a luminance image, the luminance image is subjected to a predetermined luminance conversion process, and the luminance image subjected to the luminance conversion is multiplied by the weighted R, G, B integral image to thereby obtain luminance. Generating a corrected weighted R, G, B integral image, Is characterized in that to detect the vehicle using time corrected weighted R, G, and B integral image.

A color camera for taking a bird's eye view of an area to be detected by a vehicle, and an A for converting R, G, B signals output through the color camera or the RGB decoder into digital signals.
D conversion means, first brightness conversion processing means for converting the digitized R, G, B images into brightness images for each frame and storing the brightness images, and a reference from among the obtained plurality of brightness images. Difference processing means for selecting and storing a difference image between the initial luminance image and each of the other luminance images, and digitizing based on the difference image for each of the obtained frames. R, G, B selective integration for generating and storing weighted R, G, B integrated images by integrating each R, G, B image in the time axis direction while weighting each R, G, B image Processing means; second luminance conversion processing means for converting the obtained weighted R, G, B integral image into a luminance image and storing the luminance image; and performing predetermined luminance conversion processing on the converted luminance image and storing the processed luminance image Brightness conversion processing means,
Multiplying means for generating and outputting a weighted R, G, B integrated image whose brightness has been corrected by multiplying the brightness-converted brightness image and the weighted R, G, B integrated image; Image processing means for detecting a vehicle by performing predetermined image processing on the weighted R, G, B integral images.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention will be described below with reference to the drawings when a color image is used.
FIG. 1 is a block diagram showing a first embodiment of a parked vehicle detection device according to the present invention configured by applying the method of the present invention. The illustrated parking vehicle detection device detects a vehicle in a parking lot day and night, and includes a color camera 1 that captures a bird's-eye view of the parking lot from above, a vehicle detection image generation unit 2, and an image for recognizing the parked vehicle. It is roughly divided into a processing unit 3.

The vehicle detection image generation unit 2 converts the color video signal sent from the color camera 1 into three primary color signals of R, G and B in frame units and outputs the converted signals. , B signal into a digital signal, an AD converter 202, and digitized R, G, B
An R, G, B image memory 203 for storing images;
A first luminance conversion processing unit 204 that converts the R, G, and B digital images stored in the G and B image memories 203 into luminance images on a frame-by-frame basis, and a first luminance image that stores the converted luminance images A memory 205 and a difference processing unit 2 that calculates a difference image between the reference initial luminance image and another luminance image
06, and a difference image memory 207 for storing the obtained difference image.

Further, the R, G, and B images stored in the R, G, and B image memories 203 are weighted and integrated based on the result of the difference image stored in the difference image memory 207. R, G, B selection integration processing section 208 for obtaining weighted R, G, B integral images, and R, G, B for storing the obtained weighted R, G, B integral images
A B-integrated image memory 209, a second luminance conversion processing unit 210 that performs luminance conversion processing on the weighted R, G, and B integrated images to convert the luminance image into a luminance image, and a second luminance storage unit that stores the converted luminance image A luminance image memory 211, a luminance conversion processing unit 212 that performs predetermined luminance conversion processing such as histogram equalization processing and local histogram equalization processing on the converted luminance image to correct the contrast, A converted image memory 213 for storing the converted luminance image, the converted luminance image and the R, G, B integrated image memory 2
A multiplier 214 for generating a final weighted R, G, B integral image corrected by performing a multiplication of the weighted R, G, B integral images stored in the image data 09; And a final R, G, B image memory 215 for storing final weighted R, G, B integral images.
The above-described vehicle detection image generation unit 2 and image processing unit 3
Can be configured by a discrete hardware circuit, or can be configured by a software circuit using a microcomputer and a processing program.

Next, the operation of the parked vehicle detection device having the above configuration will be described with reference to the flowchart of FIG. First, a predetermined parking area of a parking lot to be detected is overhead-viewed by the color camera 1 (step S1 in FIG. 2). The captured color video signal (for example, NTSC color composite signal) is color-separated by the RGB decoder 201 into R, G, and B signals of three primary colors (step S2).
It is converted into a digital signal by the D converter 202, and R, G, B
It is stored in the image memory 203 (step S3).

Next, the first luminance conversion processing section 204 processes each of the R, G, and B frames (for example, 128 frames) stored in the R, G, and B image memories 203.
The image is subjected to luminance conversion processing, and each of R, G,
The B image is converted into a luminance image in frame units (step S
4), and store it in the first luminance image memory 205 (step S5).

The conversion process to the luminance image can be realized by using, for example, any one of the following equations (1) to (4). y = 0.30r + 0.59g + 0.11b (1) y = max (r, g, b) (2) y = (r + g + b) / 3 (3) y = ｛max (r, g, b) + min (r, g, b)｝ / 2 (4) where y: gray-scale luminance value r: Red luminance value g: Green luminance value b: Blue luminance value

Next, the difference processing unit 206 selects a reference initial luminance image from a plurality of luminance images stored in the first luminance image memory 205, and selects the selected initial luminance image and another initial luminance image. Difference processing is performed one after another with respect to the luminance image, a difference image based on the initial image is generated for each frame (step S6), and stored in the difference image memory 207 (step S7).

Next, an R, G, B selection integration processing section 208
Are stored in the R, G, B image memory 203 as R, G, B images corresponding to the initial luminance image and R, G, B images corresponding to the luminance image obtained by performing the first difference processing between the initial luminance image and the R, G, B image.
Read the G and B images. The R, G, and B images corresponding to the luminance image on which the first difference processing has been performed are weighted based on the difference image, and the weighted R, G, and B images are used as a reference R image. , G, and B images to perform integration processing to obtain R, G, and B integrated images. Note that such integration is referred to as “selective integration” in the present invention.

When the first selective integration process is completed,
The R, G, B selection integration processing unit 208 reads the next R, G, B image from the R, G, B image memory 203, and performs weighting using a difference image corresponding to the R, G, B image. The weighted R, G, and B images are added to the R, G, and B integrated images obtained by the selective integration process to obtain new R, G, and B integrated images.

The R, G, B selection integration processing unit 208
The same selective integration processing as described above is successively executed for all of the plurality of R, G, and B images to be processed stored in the G and B image memories 203, and all the R, G to be processed are processed. , B, and R, G,
A B integral image is obtained (Step S8). Then, one weighted R, G, B integral image finally obtained in this manner is stored in the R, G, B integral image memory 209 (step S9).

The R, G, B selective integration processing section 20
The selective integration processing in 8 can be realized by executing the following equations (5) to (8).

[0021]

(Equation 1)

When the respective R, G, and B images are integrated while performing weighting with the difference image as described above, image blurring due to a moving object on the screen, which is a problem in the prior application, can be almost eliminated. . In other words, in the case of a still image, the difference between images separated in the time direction is zero, but when a moving object is present on the screen, the difference is not zero and the initial luminance image And a value corresponding to the difference between

In the case where the integration processing is performed while weighting with the difference image as described above, since the same image is superimposed one after another at the same position on a stationary object, the image becomes clearer as the integration proceeds. Meanwhile, moving objects are superimposed as attenuated images,
Only the initial image of the moving object appearing in the initial luminance image remains on the screen, and the moving object does not become trailing and blurred. For example, to explain the R signal,
From the condition of equation (5), if the difference image _St = 0, equation (6) indicates that R
= R ₀ + r ₂ + r + ─ + r ₀ , and a predetermined number of images are integrated from the initial luminance image. On the other hand, the difference image _St
In the case of> _Th (threshold), R = r ₀ (initial luminance image).

Next, the second luminance conversion processing section 210 applies any one of the above equations (1) to (4) to the weighted R, G, B integrated image obtained by the selective integration processing. Is performed to convert the image into a luminance image (step S10), and stored in the second luminance image memory 211 (step S11).

Next, the luminance conversion processing unit 212
The luminance image stored in the second luminance image memory 211 is subjected to luminance conversion by the histogram equalization processing (step S12), and the luminance image whose contrast has been corrected by the luminance conversion processing is stored in the converted image memory 213. (Step S13).

FIG. 3 shows a luminance conversion method by the above-mentioned histogram equalization processing. FIG. 3A is a luminance histogram of a luminance image before luminance conversion stored in the second luminance image memory 211. The luminance value was represented by 8 bits and 256 gradations. FIG. 3B is a luminance conversion characteristic curve.

As is clear from the luminance histogram of FIG. 3A, in the parking lot image photographed at night, many pixels are concentrated in a dark gradation part having a low luminance level, so that the vehicle, the road surface, and the like can be clearly seen. Cannot be distinguished from each other, and the image is dark and has low contrast. Therefore, if such a dark and low-contrast image is subjected to luminance conversion according to a luminance conversion characteristic curve as shown in FIG.
Low-brightness areas such as road surfaces and vehicle bodies are stretched,
The luminance of high-luminance parts such as headlights and tail lamps is compressed. As a result, a high-luminance portion does not cause halation, and can be converted to a clear luminance image with contrast and a gradation difference.

Next, the multiplier 214 reads out the weighted R, G, B integral images from the R, G, B integral image memory 209, and reads out the luminance image after the luminance conversion from the conversion image memory 213, and The images are multiplied to obtain a final weighted R, G, B integrated image whose luminance has been corrected (step S14). The final weighted R, G, B integrated image obtained by the luminance correction obtained in this manner is stored in the final R, G, B image memory 215 (step S15).

As described above, the luminance image after the luminance conversion is multiplied by the weighted R, G, and B integral images to correct the luminance, so that the original color ratio of R, G, and B is not lost. Can be improved, a halation does not occur, and a clear color image with contrast and a gradation difference can be obtained.

Finally, the image processing unit 3 outputs the luminance-corrected final weighted R, G, B obtained as described above.
A predetermined image process is performed on the integrated image to detect a parked vehicle existing in the vehicle detection target area (step S1).
6).

In the above embodiment, for simplicity of explanation, the case where the entire image is subjected to the histogram equalization processing collectively is illustrated. However, the image is divided into a plurality of sections, and the histogram is divided for each section. Brightness conversion can also be performed by a so-called local histogram equalization process that performs an equalization process. When this local histogram equalization processing is used, even if a vehicle is parked on a dark road surface as a background, the brightness of a vehicle portion brighter than the road surface can be further improved by performing optimal partitioning. It is possible to effectively reduce the brightness of high-luminance areas such as headlights, which cause the vehicle to appear more clearly than when histogram equalization is performed on the entire screen at once. A clear image can be obtained.

Further, the luminance conversion processing is not limited to the above-described histogram equalization processing and local histogram equalization processing. For example, a luminance conversion characteristic curve obtained by averaging a large number of photographed images in advance is stored in a table. Other known processing methods, such as a table conversion process for converting the brightness by referring to this table, can be adopted.

A color camera having the function of an RGB decoder is commercially available.
It is not always necessary to provide a GB decoder.

In the above embodiment, the multiplier 214
Final weighted R, G, B obtained by
The integrated image is temporarily stored in the final R, G, B image memory 215. However, the final R, G, B image memory 215 is not necessarily required. If the processing speed permits, the integrated image is directly sent to the image processing unit 3. It may be sent.

In the above embodiment, the case where a parked vehicle in a parking lot is detected has been described as an example. However, the scope of the present invention is not limited to this, and is applied to the detection of a vehicle stopped on a road. You can do it.

Further, the present invention is not limited to the above-described color image, but can be applied to a monochrome image shown in FIG. In the color image shown in FIG. 3, the same blocks as those in the first embodiment are denoted by the same reference numerals with dashes.
The operation of the block diagram of FIG. 4 is as shown in the flowchart of FIG. 5, and the operations similar to those of the flowchart of the first embodiment shown in FIG. The description of the operation is omitted.

[0037]

As described above, according to the method and apparatus for detecting a vehicle according to the present invention, it is possible to suppress halation in a high-luminance portion such as a headlight or a tail lamp, and to reduce contrast of a vehicle body in a low-luminance region. Can be emphasized, and it is also possible to prevent a moving object existing in the screen from trailing and causing image blurring. For this reason, more accurate vehicle detection becomes possible.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a first embodiment of the present invention.

FIG. 2 is a flowchart of a processing operation of the embodiment.

FIG. 3 is an explanatory diagram of luminance conversion by histogram equalization processing, where (A) is a histogram before luminance conversion;
(B) is a diagram showing a luminance conversion characteristic curve.

FIG. 4 is a block diagram showing a second embodiment of the present invention.

FIG. 5 is a flowchart of a processing operation according to the second embodiment.

[Explanation of symbols]

 Reference Signs List 1 color camera 2 vehicle detection image generation unit 3 image processing unit 201 RGB decoder 202 AD converter 203 R, G, B image memory 204 first luminance conversion processing unit 205 first luminance image memory 206 difference processing unit 207 difference Image memory 208 R, G, B selection integration processing unit 209 R, G, B integration image memory 210 Second luminance conversion processing unit 211 Second luminance image memory 212 Luminance conversion processing unit 213 Converted image memory 214 Multiplier 215 Final R, G, B image memory

Claims (3)

[Claims] An integrated image is created by integrating a digitized image of a video signal obtained by capturing a vehicle detection target area in the time axis direction, and a luminance in which a contrast of a low luminance area is enhanced by a luminance conversion process from the integrated image. Create a converted image, in the parked vehicle detection method of detecting a parked vehicle by performing image processing of the brightness conversion image, by selecting an initial image as a brightness reference from a plurality of images of the digitized image, A method for detecting a parked vehicle, comprising: obtaining a difference image of luminance between the initial image and each of the other images; and creating an integrated image while weighting the digitized image in accordance with the luminance difference. 2. A color video signal obtained by photographing a vehicle detection target area is converted into R, G, B signals, and an R, G, B image composed of the R, G, B signals is converted into a luminance image in frame units. Selecting a reference initial luminance image from the plurality of obtained luminance images to obtain a difference image between the initial luminance image and each of the other luminance images, based on each of the obtained difference images. A weighted R, G, B integrated image is generated by integrating the R, G, B images in the time axis direction while weighting the R, G, B images, and the obtained weighted R, G, B images are obtained. After converting the G, B integral image into a luminance image, the luminance image is subjected to a predetermined luminance conversion process, and the luminance converted luminance image is multiplied by the weighted R, G, B integral image to obtain a luminance. Generating a corrected weighted R, G, B integral image; A vehicle detection method comprising: detecting a vehicle using a weighted R, G, B integral image whose degree has been corrected. 3. A color camera for taking a bird's-eye view of a vehicle detection target area; A / D conversion means for converting R, G, B signals output via the color camera or RGB decoder into digital signals; First luminance conversion processing means for converting the R, G, and B images into luminance images for each frame and storing the luminance images; and selecting a reference initial luminance image from the obtained plurality of luminance images, Difference processing means for obtaining and storing a difference image between the initial luminance image and each of the other luminance images; and weighting the digitized R, G, and B images based on the obtained difference image for each frame. R, G, B selective integration processing means for generating and storing weighted R, G, B integrated images by integrating each of the R, G, B images in the time axis direction while performing R, G, A second luminance conversion processing unit that converts the integrated image into a luminance image and stores the luminance image; a luminance conversion processing unit that performs predetermined luminance conversion processing on the converted luminance image and stores the luminance image; Multiplying means for generating and outputting weighted R, G, and B integrated images corrected by multiplying the weighted R, G, and B integral images; and outputting the weighted R, G, and B integrated images. An image processing means for detecting a vehicle by performing predetermined image processing on the B integral image.