JP2672863B2 - Vehicle detection method - Google Patents

Description

The present invention relates to a vehicle detection method for detecting a vehicle existing in a predetermined monitoring target area such as a road or a parking lot at night or in a dark place.

[Conventional technology]

As a method of detecting a night-time traveling vehicle, a method of detecting a vehicle by extracting a vehicle light source such as a headlight or a tail lamp of the vehicle from an image captured by a television camera has been proposed. For example, in JP-A-63-184898, a vehicle running at night is captured as a multi-valued image by a television camera or the like,
The light and dark densities of each pixel are projected and added in the vertical direction, and the vertically added image is binarized by a threshold value determined based on the maximum value of this addition result, and a predetermined vehicle determination condition is set for this binarized image. There has been proposed a device adapted to detect a vehicle.

[Problems to be solved by the invention]

However, the vehicle detection method of the above-mentioned conventional device is such that, for example, when two vehicles are traveling back and forth as shown in FIG. 10 (a), the headlight of the vehicle behind is tail light of the vehicle ahead. When the area is illuminated (hatched area in Fig. 10 (a)), the vertical density addition result is as shown in Fig. 10 (b), and the illumination intensity of the headlight is higher than that of the tail lamp. Because it is expensive
The irradiated part of the headlight is detected as the maximum value of the added density. Therefore, the threshold value is set with this maximum value as a reference. Therefore, when the threshold value processing is performed using this value, the irradiation portion of the headlight of the following vehicle is extracted as shown in FIG. 10 (c). However, the original information of the tail lamp portion cannot be extracted, and the determination may be erroneous.

Further, in the case of rain, if a puddle or the like exists on the road surface as shown in FIG. 11 (a), reflected light from the puddle such as a tail lamp or a night illumination lamp will be captured as an image. Therefore, the vertical density addition result is
As shown in Fig. 11 (b), the 11th
As shown in FIG. 7C, the information on the tail lamp portion cannot be extracted.

As described above, in the conventional method, the presence of reflected light of the same level as or higher than that of headlights and tail lamps causes erroneous detection, and always detects vehicle light sources such as headlights and tail lamps in a stable manner. There was a problem that I could not.

The present invention has been made under such circumstances,
So-called unnecessary reflection areas such as the vehicle body surface of the vehicle and the road surface with puddle that generate reflected light are removed from the image, and only the vehicle light sources such as headlights and tail lamps that are characteristic of the vehicle at night or in the dark are stably extracted. An object of the present invention is to provide a vehicle detection method capable of performing the above.

[Means for solving the problem]

According to a first vehicle detection method of the present invention, a polarizing element set so that polarized light that oscillates in parallel with an incident surface perpendicular to a road surface among incident light from a shooting range is attached to the front surface of a television camera. , The monitoring target area is imaged. Then, by performing predetermined threshold value processing on the obtained density value of the horizontally polarized image, unnecessary irradiation areas other than vehicle light sources such as headlights and tail lamps are removed.

In addition, the second vehicle detection method according to the present invention, in order to more reliably remove the unnecessary irradiation area, polarized light and vertical light that oscillate parallel to an incident surface perpendicular to the road surface among incident light from the shooting range. A polarizing element that can be changed so that the oscillating polarized light can pass through is attached to the front of the TV camera, and the polarization plane of this polarizing element is alternately switched between the horizontal direction and the vertical direction, and the monitored area is synchronized with the switching of this polarization plane. To shoot. By calculating the density ratio of the corresponding pixels of the horizontally polarized image and the vertically polarized image obtained in this way and applying a predetermined threshold value processing to the image having the calculated density ratio, the headlight The unnecessary irradiation area other than the vehicle light source, such as a tail lamp and a tail lamp, is removed more stably.

(Operation)

Before explaining the principle of the vehicle detection method of the present invention, the refraction, reflection, and polarization of light, which is the basis of the unnecessary irradiation area removal processing in the present invention, will be described.

Generally, the vibration plane of natural light is random, but in order to facilitate understanding in the following, vertical polarization (S polarization) that vibrates perpendicularly to the incident surface as shown in FIG.
And horizontally polarized light (P polarized light) that oscillates parallel to the plane of incidence
Only the two lights of

Now, in FIG. 4 (a), I _i , I _r , and I _f are the light intensities of the incident light, the reflected light, and the refracted light, M _i , M _r , and M _f are the incident light area, the reflected light area, Assuming the area of refracted light, the energies W _i , W _r , and W _f of the incident light I _i , the reflected light I _r , and the refracted light I _f can be expressed by the following equations.

Further, the energy reflectance R of the substance A on which light is incident is defined as follows.

Here, when the energy reflectances R _p and R _s of the horizontally polarized light (P-polarized light) and the vertically polarized light (S-polarized light) with respect to an arbitrary incident angle φ are obtained from the Fresnel formula, the following equations are obtained.

Note that φ and γ are the incident angle and the refraction angle shown in FIG. 4 (b).

From this equation (3), when (φ + γ) becomes π / 2, tan ² (φ + γ) in the denominator of R _p becomes infinite, so that the energy reflectance R _p of horizontal polarized light (P polarized light) becomes 0. That is, when (φ + γ) is π / 2, all the horizontally polarized components are transmitted, and the reflected light reflected from the substance A is only the vertically polarized (S polarized) component. Expressing this as an equation, Becomes According to Snell's law, the refractive indices n ₁ and n _{2 in} FIG.
The relationship is Therefore, using the relationship of equation (4), When the relation of this expression (6) is established, all the horizontally polarized components pass through the substance A, and the reflected light from the substance A becomes only the vertically polarized light.

The above content will be described below with reference to an actual captured image in FIG.

That is, when the expression (6) is satisfied, all the horizontally polarized components of the incident light are transmitted through the puddle and are absorbed by the road surface thereunder, and the energy of the reflected light is attenuated as much. Therefore, the light-dark density value of the reflected light from the puddle becomes smaller in the horizontally polarized image (FIG. 5 (a)) than in the vertically polarized image (FIG. 5 (b)), and it is unnecessary as an image. The area gets darker. This phenomenon also holds true with respect to the reflection on the vehicle body surface, although to a different degree.

In other words, in a horizontally polarized image, the density difference between the light source and the vehicle light source, which is a direct light, such as a headlight and a tail lamp, and the light reflected from other unnecessary irradiation areas becomes larger, so that the horizontally polarized image becomes larger. If threshold processing is performed using an image, it is possible to remove an unnecessary irradiation area and obtain an image composed of only vehicle light sources such as headlights and tail lamps.

The present invention has been made based on the above theory. In the first vehicle detection method, the monitoring target area is captured as an image by the image pickup means 1 as shown in the principle thereof in FIG. Imaging means 1
A polarizing element 2 having a polarization plane set in the horizontal direction is disposed on the front surface of the. The image storage means 3 stores the horizontal polarized image output from the image pickup means 1 as digital image data, and sends the image data to the unnecessary irradiation area removing means 4. The unnecessary irradiation area removing means 4 removes the unnecessary irradiation area by performing threshold processing on the sent image data, and obtains image data of only the headlights and tail lamps. Then, the obtained image data is sent to the vehicle detection means 5 to make a vehicle determination.

Furthermore, the second vehicle detection method of the present invention uses a vertically polarized image in order to more stably remove the unnecessary irradiation region.

That is, comparing the horizontally polarized image and the vertically polarized image,
There is almost no difference between the light and dark density values of vehicle light sources such as headlights and tail lamps. The reason for this is that the vehicle light source is direct light and the vibrating surface of the light is random, so there is almost no effect on the polarization component.

On the other hand, as described above, a large difference occurs in the image density value of the unnecessary irradiation area other than the vehicle light source such as the headlight and the tail lamp between the horizontally polarized image and the vertically polarized image. As an example, the reflectance of water is shown in FIG. In the figure, R _P is the reflectance for horizontally polarized light, R _S is the reflectance for vertically polarized light,
As is apparent from this figure, the reflectance of the horizontally polarized light is extremely smaller than that of the vertically polarized light. That is, in the horizontally polarized image, the light and dark density value of the unnecessary irradiation region becomes extremely smaller than that in the vertically polarized image. Therefore, when calculating the density ratio between these two vertically and horizontally polarized images, the density ratio of the vehicle light source parts such as headlights and tail lamps that emit direct light does not change much, but the density ratio of the unnecessary irradiation area is a large value. Become. Therefore, by setting a threshold on the boundary between the vehicle light source and the unnecessary irradiation area in the image of this density ratio, the unnecessary irradiation area can be more reliably removed from the screen.

The second vehicle detection method of the present invention uses the above logic, and the imaging means 11 captures the monitoring target area as an image, as the principle thereof is shown in FIG. At this time, the imaging means
The polarization angle of the polarization element 12 arranged in front of 11 is switched alternately by the polarization angle control means 16 between the horizontal direction and the vertical direction. Therefore, the horizontally polarized image and the vertically polarized image are alternately input to the image pickup means 11, and each image is converted into an image pickup signal giving a density in the image pickup means 11.

The image storage means 13 stores the horizontally polarized image and the vertically polarized image output from the image pickup means 11 as digital image data, respectively, and the image data is stored in the unnecessary irradiation area removing means 14
Send to The unnecessary irradiation area removing means 14 calculates the density ratio of each corresponding pixel using the density data of the horizontally polarized image and the vertically polarized image, and performs a predetermined threshold value process on the image having this density ratio. Unnecessary irradiation areas are removed and image data consisting only of vehicle light sources such as headlights and tail lamps are obtained. Then, this image data is sent to the vehicle detection means 15 to determine the vehicle.

[Embodiment 1] Hereinafter, a first embodiment of the present invention will be described with reference to the drawings.

FIG. 7 shows an example of a vehicle detection device configured by applying the first vehicle detection method of the present invention. In the figure, 21 is a TV camera such as an ITV camera as an image pickup means, and 22 is a TV camera 2
Linear polarizer placed in front of 1, 23 is a low-pass filter, 24 is an AD converter, 25 is an image memory that can store image data of one or more screens of the input image, and 26 is image data transferred from the image memory 25. Is an output memory for storing, and 27 is a microprocessor for performing vehicle determination processing.

 Next, the operation of the above embodiment will be described.

The polarization plane of the polarizer 22 is directly set in the horizontal direction, and the television camera 21 captures an image of the monitoring target area. The captured image signal is sent to the AD converter 24 through the low pass filter 23. Then, the AD converter 24 samples the image signal of this horizontally polarized image, converts it into multi-tone digital density data a, and sends it to the image memory 25.

The low pass filter 23 is inserted for the purpose of removing high frequency noise components. Generally, noise generated from an amplifier inside an image pickup device such as a television camera exists almost uniformly from a high frequency component to a low frequency component. Therefore, by removing the high frequency component with a low pass filter 23, a random noise component In addition to reducing the number, it also prevents the generation of aliasing noise during sampling.

When the acquisition of the digital density data for one screen of the horizontally polarized image is completed as described above, the digital density data stored in the image memory 25 is sent to the output memory 26 through the data bus b. When the microprocessor 27 receives the data transfer end signal c, it reads the image data from the output memory 26 through the data bus d, and performs threshold processing with a threshold value set on the basis of the image data to remove the unnecessary irradiation area. Image data consisting of only light sources such as headlights and tail lamps is obtained. Then, the vehicle determination described below is executed based on this image data, and when the vehicle determination processing ends, the microprocessor
27 outputs the processing end signal e to the AD converter 24 and repeats the above-mentioned processing again.

FIG. 9 is a flowchart of the vehicle determination process of the microprocessor 27 described above. After removing the unnecessary irradiation area, the microprocessor 27 specifies the detection area in which the vehicle determination is to be performed (step [1]), and the vertical projection addition of the density value of each pixel existing in the specified area. Is performed (step [2]).

Then, in this addition result, if there is a positive polarity of extremely high brightness, it is determined that "there is a vehicle" (steps [3] and [4]), and if it does not exist, it is determined that "there is no vehicle" (step [ 3] [5]), the vehicle is detected.

By arranging the linear polarizer in the horizontal direction on the front surface of the television camera as described above, unnecessary irradiation areas other than the vehicle light source such as headlights and tail lamps can be removed to perform stable vehicle determination.

Second Embodiment Next, a second embodiment of the present invention will be described.

FIG. 8 shows an example of a vehicle detection device configured by applying the second vehicle detection method of the present invention. In the figure, 31 is a TV camera such as an ITV camera as an imaging means, and 32 is a TV camera 3.
Linear polarizer arranged in front of 1, 33 is a low-pass filter, 34 is an AD converter, 35 and 36 are image memories capable of storing image data of one or more screens of the input image, 37 and 38 are logarithmic calculators, 39 is a difference calculator, 40 is an output memory for storing calculation results, 41 is a microprocessor for performing vehicle determination processing, and 42 is control for performing variable control of the polarization angle of the linear polarizer 32 and sampling control for capturing image data. Part 43 is a sampling timer.

 Next, the operation of the above embodiment will be described.

An image of the monitoring target area is input to the television camera 31. The control unit 42 outputs the polarization angle control signal a, rotates the linear polarizer 32 to the vertical polarization position, and outputs the image capture signal b to the AD converter 34 in synchronization with this.

When the linear polarizer 32 is set to the vertical polarization position, the television camera 31 outputs a vertically polarized image image signal, and the image signal is sent to the AD converter 34 through the low-pass filter 33. Then, the AD converter 34 samples the image signal of this vertically polarized image, converts it into multi-tone digital density data c, and sends it to the image memories 35 and 36.

The low-pass filter 33 is inserted for the purpose of removing high frequency noise components and the like as in the first embodiment.

Next, the control unit 42 outputs the image memory selection vibration d,
One of the two image memories 35 is selected, and digital density data for one screen of the vertically polarized image is stored in the image memory 35.

When the capture of the digital density data for one screen of the vertically polarized image in the image memory 35 is completed, the control unit 42 outputs the timer start signal k and activates the sampling timer 43. Then, when the sampling timer 43 times out after the elapse of a certain period of time, the control unit 42 outputs the time-up signal e.
Output to

After receiving the time-up signal e, the control unit 42 outputs the polarization angle control signal a again, rotates the linear polarizer 32 to change its polarization plane from the vertical polarization position to the horizontal polarization position, and further synchronizes with this. Then, the image capture signal b is output to the AD converter 34.

When the direct polarizer 32 is set to the horizontal polarization position, the image signal of the horizontal polarization image is output from the television camera 31 and sent to the AD converter 34 through the low pass filter 33. The AD converter 34 samples the image signal of the horizontally polarized image and converts it into multi-tone digital density data c,
Send to the image memory 35, 36.

The control unit 42 outputs the image memory selection signal and outputs the image memory.
36 is selected and digital density data for one screen of the horizontally polarized image is stored in the image memory 36.

When the capture of the digital density data for one screen of the vertically polarized image and the horizontally polarized image is completed as described above,
The digital density data stored in the image memories 35 and 36 are logarithmically calculated by the logarithmic calculators 37 and 38, and then sent to the difference calculator 39 through the data buses g and h.

The difference calculator 39 performs difference calculation for each pixel corresponding to the vertically polarized image and the horizontally polarized image sent from the logarithmic calculators 37 and 38, and outputs the logarithmic difference calculation value 1 to the output memory 40. Store. In this way, the density data of the two images are logarithmically converted, and then the difference operation is performed to perform the same operation as the division, that is, the same operation as the density ratio of each pixel corresponding to the vertically polarized image and the horizontally polarized image. I went to.

When the logarithmic difference calculation is completed for one screen of the vertically polarized image and the horizontally polarized image, the difference calculator 39 outputs a calculation end signal i to the microprocessor 41.

Upon receipt of the calculation end signal i, the microprocessor 41 reads the logarithmic difference calculation data from the output memory 40 through the data bus j, and performs a predetermined threshold value processing based on the logarithmic difference calculation data. As an example of this threshold value processing, for example, "0" may be set if the density ratio is smaller than the set threshold value, and "1" may be set if it is large. As a result, image data in which the unnecessary irradiation area is removed can be obtained. The vehicle determination process shown in FIG. 9 is executed on this image data as in the case of the first embodiment. When the vehicle determination process ends, the microprocessor 41 outputs a process end signal m to the control unit 42, restores the state of the control unit 42 to the initial state, and repeats the above-described processes.

Generally, when the imaging condition changes, the image density value also changes. However, if the density ratio between the horizontally polarized image and the vertically polarized image is obtained as described above, the threshold value can be set without considering the change in the density value. Therefore, the unnecessary irradiation area can be removed more stably.

〔The invention's effect〕

As is clear from the above description, when the vehicle detection method of the present invention is used, the image information of the unnecessary irradiation area other than the vehicle light source such as headlights and tail lamps, which is a major drawback of the conventional method, can be reliably obtained from the image. It is possible to remove, and it is possible to stably extract only vehicle light sources such as headlights and tail lamps, which are features of the vehicle at night or in a dark place, and it is possible to improve vehicle detection accuracy.

[Brief description of the drawings]

1 is a principle diagram of a first vehicle detection method of the present invention, FIG. 2 is a principle diagram of a second vehicle detection method of the present invention, FIG. 3 is an explanatory diagram of horizontal polarization and vertical polarization, and FIG. Is an explanatory view of refraction / reflection of light, FIG. 5 is a view showing an example of a photographed image, FIG. 6 is a measurement view of reflectance of water, FIG. 7 is a block diagram of the first embodiment of the present invention, FIG. 8 is a block diagram of a second embodiment of the present invention, FIG. 9 is a flowchart of vehicle determination in the above embodiment, and FIGS. 10 and 11 are explanatory diagrams of vehicle erroneous detection due to unnecessary irradiation.

Claims (2)

(57) [Claims] 1. A polarizing element, which is set so that polarized light that oscillates in parallel with an incident surface perpendicular to a road surface among incident light from a photographing range passes, is arranged in front of a photographing means such as a television camera. The horizontal polarization image obtained by the image pickup means is subjected to predetermined threshold value processing to extract an image composed of only vehicle light sources such as headlights and tail lamps, and the vehicle is judged based on the extracted image. Vehicle detection method. 2. A polarizing element is disposed in front of an image pickup means such as a television camera, and polarized light vibrating in parallel with or perpendicular to an incident surface perpendicular to a road surface among incident light from a photographing range. Of the horizontal polarization image and the vertical polarization image obtained by the image pickup means are calculated, and predetermined threshold processing is performed on the image having this density ratio. A vehicle detection method characterized by extracting an image consisting only of a vehicle light source such as a headlight or a tail lamp by performing the application, and determining a vehicle based on the extracted image.