JP2002083301A - Traffic monitoring device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To solve a problem that a conventional traffic monitoring device judges the external environment of day, night, and dusk by the luminance of a road surface portion, makes an erroneous judgment due to the reflection of headlights of vehicles and street lamps or the like at night, and makes erroneous vehicle detection as a result. SOLUTION: This traffic monitoring device is provided with an image feature extracting means splitting the input image from an image input means inputting a vehicle image into the prescribed size with an image splitting means to obtain image features in split areas, an external environment judging means judging the external environment of a photographed object from the image features, a background difference generating means generating a background difference image from a background image photographed and stored in advance and the input image, a moving body detecting means detecting a moving body from the background difference image, and a moving body tracking means tracking the moving body.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a traffic monitoring apparatus for acquiring information such as the number of vehicles passing and the speed of the vehicles, detecting falling objects on a road surface, and providing a warning to a driver of the vehicles. Things.

[0002]

2. Description of the Related Art Conventional techniques for vehicle detection include, for example, IEICE Technical Report, PRMU97-6, 1997, pp.4.
1-48 There is a vehicle detection method described in "Research on traffic flow monitoring method by image processing". FIG. 10 is a configuration diagram of a traffic monitoring device that performs this detection method. In the figure, 1 is an image input means,
2 is an image feature extracting means, 4 is an external environment determining means, 6 is a moving object detecting means, and 7 is a moving object tracking means. FIG. 11 is a diagram for explaining a vehicle detection method. In the figure, reference numerals 10 and 11 denote moving vehicles, 30 denotes an initial detection area for detecting a vehicle, 31
Is a road surface portion for determining the external environment. FIG. 12 shows a flow of a process performed by the traffic monitoring device using the configuration of FIG.

Next, the operation of the traffic monitoring device shown in FIG. 10 will be described with reference to the flowchart of FIG. Step S
At 61, the input image from the camera is input to the image input means 1. In step S62, the image feature extracting means 2 obtains the luminance distribution of the road surface portion 31 from the input image, and the external environment determining means 4 determines whether it is day / night / dusk based on the luminance distribution.

In step S63, the moving body detecting means 6 detects a light / dark pattern or a headlight pair on the front of the vehicle and stores a small area image (template) at the front of the vehicle. Specifically, based on the result of the external environment determining means 4,
If it is determined that it is daytime, the range in which the vertical and horizontal projection distributions of the brightness equal to or more than a predetermined value in the initial detection area 30 are equal to or more than a predetermined threshold is regarded as a vehicle, and the moving object detection unit 6 detects the vehicle when the detection result can be regarded as a vehicle. . If it is determined to be nighttime, the zero cross point of the horizontal differentiation of the vertical projection distribution of the luminance equal to or higher than the second predetermined value in the initial detection area 30 is set as the tail lamp position, and when this pair of tail lamp positions exists, it is detected as a vehicle. In step S64, the moving body tracking means 7
Detects a vehicle by searching from the current frame of an image similar to the template obtained in the previous frame.

[0005]

As described above, the conventional traffic monitoring device determines the external environment such as daytime, nighttime or dusk based on the brightness of the road surface portion, as described above. As a result, there is a problem in that the daytime is erroneously determined due to reflection of the vehicle, and as a result, an erroneous vehicle detection is performed.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problems, and determines the external environment based on image information of an arbitrary partial area of an image, thereby enabling a vehicle headlight, a streetlight, etc. at night. It is an object of the present invention to be able to correctly determine the external environment even when there is reflection of a vehicle, and as a result, to be able to correctly detect a vehicle.

[0007]

A traffic monitoring device according to the present invention comprises: an image input unit for inputting a captured vehicle image; an image dividing unit for dividing an input image from the image input unit into a predetermined size; Image feature extraction means for obtaining an image feature in the divided area; external environment determination means for determining an external environment of an imaging target based on the image feature; and a background difference image from a previously captured and stored background image and the input image. The image processing apparatus includes a background difference generating unit that generates a moving object, a moving object detecting unit that detects a moving object from the background difference image, and a moving object tracking unit that tracks the moving object.

Further, in the traffic monitoring device according to another invention, the external environment determining means determines only the divided areas having greatly different image characteristics at different times.

In the traffic monitoring device according to another invention, the image feature extracting means extracts a hue and a luminance distribution from an RGB value of the image, and the external environment determining means deteriorates visibility in rain, fog or the like. The state is determined to be based on the luminance and the hue.

In a traffic monitoring apparatus according to another aspect of the invention, the image feature extracting means holds image features at a plurality of different times in advance, and uses an image feature at a time closest to the current time as a reference image feature. The external environment determining means is configured to determine the external environment from the reference image feature and the current image feature.

Further, in the traffic monitoring device according to another invention, the image feature extracting means holds image features at different times in advance, and uses image features at a plurality of times close to the current time as reference image features to determine an external environment. The means is configured to determine an external environment from the plurality of reference image features and the current image features.

Further, in the traffic monitoring device according to another invention, the external environment determining means excludes image characteristics of an area detected as a moving object by the moving object detecting means from a determination target.

[0013]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiment 1 FIG. FIG. 1 is a diagram showing a configuration of the first embodiment of the present invention. In the figure, reference numeral 1 denotes an image input means for inputting an image from a camera (not shown) (pixel values take values from 0 to 255), and 2 denotes an image within a predetermined area of the input image from the image input means 1. Image feature extracting means for obtaining features, 3 is an image dividing means for dividing an input image into a predetermined size and outputting the image to the image feature extracting means, 4 is an external environment for determining an external environment based on image features from the image feature extracting means 2 The determination means 5 is a background difference detection means which previously captures an image with a camera (not shown) and obtains a background difference image from the stored background image having no moving body regions 10 and 11 (described later) and the input image. A moving object detecting means for detecting a moving object area, a moving object tracking means for tracking the moving object and outputting the position, speed, and moving direction of the moving object, and a moving image detecting means for inputting a background image stored in the background difference detecting means to an input image. Based Is the background image updating means for updating.

FIG. 2 is a diagram for explaining the moving object detecting means 6. In the figure, reference numerals 10 and 11 denote moving body regions such as a motorcycle, a four-wheeled vehicle, and a falling load at time t-1.
Reference numerals 2 and 13 denote moving body regions at time t. FIG. 3 is a flowchart illustrating the operation of the first embodiment. Note that the input image from the camera includes an arbitrary area other than the road surface, as is clear from FIG.

Next, the operation of this embodiment will be described with reference to the flowchart of FIG. In step S1, a captured image at time t output from a camera (not shown) that captures an image of a road is converted by the image input unit 1 to a value of each pixel of 0 to 25.
A digital image having a luminance value of 5 is input. In step S2, the image dividing means 3 divides the input image into a predetermined size and outputs the image to the image feature extracting means 2. The image feature extraction means 2 calculates the luminance statistical information V for each divided area according to the equation (1).
Get in.

Vj = (δi) / n · di (1) where Vj is luminance statistical information in the divided area j, n
Is the number of frames to be averaged, δi is the variance of the divided area j at time i, and di is the edge gradient amount of the divided area j at time i. As the edge gradient amount, for example, a difference value between adjacent pixel luminance values in the divided area is set as a sum of the areas.

In step S3, the external environment determining means 4
Calculates the number of areas Vnum satisfying the following equation (2) for all divided areas from the luminance statistical information V.

Vj> Vth (2) where Vj is luminance statistical information in divided area j, Vth
Is a predetermined threshold.

The external environment determining means 4 determines whether the external environment is day / night / dusk based on the number of areas Vnum, and selects a preset processing parameter for day / night / dusk. For example, when Expression (3) is satisfied, it is determined that the night is at night. This is based on the finding that the camera apparatus performs sensitization because the overall luminance decreases at night, and the noise level increases at the pixel where the input light is insufficient, so that the variance value increases at night.

Vnum> Vth1 (3) where Vth1 is a first predetermined threshold.

Next, it is daytime when the expression (4) is satisfied. Vnum <Vth2... (4) where Vth2 is a second predetermined threshold, and Vth2 <V
th1.

If the above equations (3) and (4) are not satisfied, it is determined that twilight occurs.

In step S4, the background difference detection means 5 obtains a background difference image from the input image and the background image by using equation (5) in order to obtain a moving object region. This is performed for all pixels of the input image.

H (x, y) = | B (x, y) −I (x, y) | (5) where H (x, y) is a point of the background difference image The luminance value at (x, y), B (x, y) is the luminance value at point (x, y) of the background image, I (x,
y) is a luminance value at a point (x, y) of the input image.

This is described, for example, in the document "Image processing and recognition".
(Co-authored by Shokodo, Takeshi Aiin and Tomoharu Nagao, PP. 161-1
62) is one method of moving object detection called a background subtraction method.

In step S5, the moving object detecting means 6 obtains moving object regions 12, 13 from the background difference image. For example, a binarization threshold is obtained for the moving object region based on the result of the external environment determination unit 4 determining the background difference image, and the binarized background difference binary image is detected as a region by a labeling method. Here, as the binarization threshold, for example, a binarization threshold for each of day, night, and dusk is used in advance. The labeling method is disclosed in the document "Introduction to Computer Image Processing" (published by Soken Publishing, Hideyuki Tamura, PP. 75-76), and can be performed by extracting connected pixels from a background difference binary image as one region.

In step S6, the moving body tracking means 7 tracks the object based on the moving body area information at the time t-1 and the moving body area information at the time t. In the tracking method, the representative points of the moving body regions 10 and 11 detected at time t-1 are associated with the representative points of the moving body regions 12 and 13 to perform tracking. Here, for example, the center of gravity of the region or the like is used as the representative point. For the association, for example, a moving body area at time t−1 that is closest to the moving body area representative point at time t and the distance is within a predetermined range is obtained, and the corresponding moving body area is stored as the same moving body.
Further, the speed and the moving direction are obtained from the distance between the corresponding moving object representative points. The speed and the moving direction are obtained by, for example, Expression (6).

Speed = D (Ot, Ot-1) / T Moving direction = arctan (| Ot.x-Ot-1.x | / | Ot.y-Ot-1.y |) ……………… (6) where D (a, b) is the Euclidean distance between two points a and b, T is the time interval between time t and time t-1, Ot is the object position at time t, and Ot.x is the object Is the X coordinate and Ot.y is the Y coordinate of the object.

In step S7, the background image updating means 8 updates the background image based on the background difference image. The updating method is performed by, for example, Expression (7). This is performed for all pixels of the input image.

Bt (x, y) = Bt−1 (x, y) + α · {It−1 (x, y) −Bt−1 (x, y)} (7) where Bt ( (x, y) is the luminance value at time t at point (x, y) in the background image, It (x, y) is the luminance value at time t at point (x, y) in the input image, and α is a predetermined constant It is.

The above operation is executed at regular intervals. For example, it is executed at a video rate (30 ms).

As described above, according to the present embodiment, the external environment is determined based on the total result of the image information of the partial area including the area other than the arbitrary road surface. Erroneous determination of the external environment due to the above.

Embodiment 2 FIG. In the first embodiment, the external environment is determined in all of the determined divided regions. However, in the present embodiment, only the divided regions that are significantly different from the image features at different times in advance are determined by the external environment determining unit 4. It was made to be a judgment target in.

Although the block diagram showing the configuration of the second embodiment is the same as that of the first embodiment, the configuration and operation of the external environment determining means 4 are different.

Next, the operation of this embodiment will be described. The operations of the image input unit 1, the image dividing unit 3, the background difference detecting unit 5, the moving object detecting unit 6, the moving object tracking unit 7, and the background image updating unit 8 are the same as those in the first embodiment. The operation of the image feature extraction unit 2 and the external environment determination unit 4 at the time of device adjustment for determining the determination target divided region will be described with reference to the flowchart of FIG. 4 for explaining the operation of the present embodiment.

First, at the time of device adjustment, the moving object regions 10, 1 at a predetermined time by a camera (not shown) are set.
A background image without 1 is captured and input. Step S11
Then, the image feature extracting means 2 divides the image into a predetermined size in the same manner as in the first embodiment, and obtains an image feature of each divided region. Further, in step S12, a plurality of predetermined image feature storage times and the respective image features at that time are stored in the external environment determining means 4. The time to save is, for example, day / night /
Times 12:00, 00:00, 17:00 corresponding to twilight
Set to 0. In step S13, the image features at each time are compared, and a region where the image features are significantly different is determined.
Specifically, a j-th divided area D satisfying the expression (8) is obtained.

Max (| Vj−Vj ′ |) (8) where Vj corresponds to luminance statistical information at the current time in the divided area j, and Vj ′ corresponds to a determination result in the divided area j. This is luminance statistical information at the image feature saving time.

Next, at the time of operation of the apparatus, the image input means 1, the image feature extracting means 2, the image dividing means 3, the background difference detecting means 5, the moving object detecting means 6, the moving object tracking means 7, and the background image updating means 8 are the same as those of the embodiment. In the same manner as in the first embodiment, the external environment determining means 4 determines the external environment as in the first embodiment, using the divided area D as a determination target area.

As described above, according to the present embodiment, only the divided regions that are greatly different from the image features at different times are determined as the objects to be determined by the external environment determining means 4 in advance. This eliminates the possibility of erroneously determining the external environment due to sunlight fluctuation.

Third Embodiment In the first embodiment, the external environment determining means 4 determines the external environment from the luminance statistical information. In the present embodiment, the average luminance and the color distribution in the divided area are further determined. In consideration of the above.

The third embodiment is the same as the first embodiment in the block diagram showing the configuration thereof, but the image feature extraction means 2
And the operation of the external environment determining means 4 is different. FIG. 5 is a flowchart illustrating the operation of the present embodiment.

Next, the operation of this embodiment will be described. The operations of the image input unit 1, the image dividing unit 3, the background difference detecting unit 5, the moving object tracking unit 7, and the background image updating unit 8 are the same as those in the first embodiment. First, the operation of the image feature extraction unit 2 and the external environment determination unit 4 will be described with reference to the flowchart of FIG. In step S21, the image feature extraction means 2 obtains the luminance I and the saturation S according to the equations (9) and (10), and then calculates the average luminance Ij 'and the average saturation Sj' in the divided area.
Ask for.

I = (1/3) × (R + G + B) (9) S = 1- (3 / (R + G + B)) × a Where R is the luminance value of the red component, G is the luminance value of the green component,
B is the luminance value of the blue component, a is the minimum value of R, G, B,
S takes a value from 0 to 1.0 in saturation, and I takes 0 to luminance in brightness.
It takes values up to 255.

In step S22, the external environment determining means 4 obtains an external environment from the average luminance Ij 'and the average saturation Sj'. Specifically, the number of divided regions Vnu that satisfies Expression (11)
Find m '.

Sj ′ <Sth and Ij ′> Ith (11) where Sj ′ is the average saturation in the divided region j, Sth is the predetermined saturation threshold, and Ij ′ is the divided saturation j in the divided region j. Average brightness,
Ith is a predetermined luminance threshold value.

Next, when Vnum 'satisfies the expression (12), the visibility is regarded as poor. This is based on the finding that when the visibility is deteriorated due to rain or fog, the brightness is uniformly increased by the reflected light from the water droplets in the air, and the saturation is reduced.

Vnum ′> VSth (12) where VSth is a predetermined threshold value.

Next, the operation of the moving object detecting means 6 will be described. When the external environment determining unit 4 determines that the visibility is poor, the moving object detecting unit 6 performs moving object detection from the background image output from the background difference detecting unit 5 and sets a threshold value to 0 during binarization processing.
The moving object detection result is set to 0. Subsequent processing is the same as in the first embodiment.

As described above, according to the present embodiment, the color distribution within the divided area is also determined, so that
It can be determined that the visibility is deteriorated due to fog or the like, and when the vehicle detection result is erroneous in an unstable image, the vehicle detection result is set to 0, thereby preventing an erroneous vehicle detection result.

Fourth Embodiment In the first embodiment, the external environment is determined based on the current image feature. However, in the present embodiment, the image feature having different image times and having the closest time to the current time is provided. With the feature as a reference image feature, an external environment is determined from the reference image feature and the current image feature.

Although the block diagram showing the configuration of the fourth embodiment is the same as that of the first embodiment, the configurations and operations of the image feature extraction means 2 and the external environment judgment means 4 are different. FIG. 6 is a diagram for explaining the operation of the external environment determining means. 2 in the figure
0 is a previously obtained image feature for each daytime, 21 is a previously obtained image feature for each night time, 22 is a previously obtained image feature for each hour of dusk, and 23 is a previously calculated heavy rainfall image feature. 24, the image feature for each time of the fog obtained in advance, 25 is the time zone in which the image feature for the day is stored,
26 is a time zone where night image features are stored, 27 is a time zone where dusk image features are stored, 28 is a time zone where heavy rain image features are stored, and 29 is a time zone where fog image features are stored. Bands, which are image feature tables (20 to 24)
Is stored in the external environment determination means 4. FIG. 7 is a flowchart illustrating the operation of the present embodiment.

Next, the operation of this embodiment will be described. The operations of the image input unit 1, the image dividing unit 3, the background difference detecting unit 5, the moving object detecting unit 6, the moving object tracking unit 7, and the background image updating unit 8 are the same as those in the first embodiment. The operation of the image feature extraction unit 2 and the external environment determination unit 4 will be described with reference to the flowchart of FIG.

In step S31, the image feature extracting means 2 obtains image features of a partial area in the input image. In step S32, the external environment determining means 4 compares the image features obtained in advance, for example, the image features obtained by the image feature extracting means 2 with the hourly hourly hour in day / night / dusk / heavy rain / fog. . Specifically, for example, the image feature tables (20 to 24) for day / night / dusk / heavy rain / fog are obtained in advance. However, during day / night / dusk storage time (25
Except for (27), the image feature is set to 0. Current time is 1
In the case of 2:14, 12:00 noon, night, twilight,
The difference between the image feature table value (20 to 24) under heavy rain / fog conditions and the image feature of the input image partial area obtained in step S31 is obtained, and the image feature with the smallest difference value is obtained. In step S33, the weather condition corresponding to the image feature table having the minimum difference value is set as the current external environment.

As described above, according to the present embodiment, the image features at different times are obtained in advance, and the stored image feature closest to the current time is set as the reference image feature, and this reference image feature and the current Since the external environment is determined from the image characteristics, the external environment can be determined with higher accuracy even when the image characteristics fluctuate due to sunshine fluctuation due to sunset such as at twilight.

Fifth Embodiment In the fourth embodiment, the external environment determination means 4 makes a determination based on a pre-stored image feature closest to the current time. Judgment was made from image features at a plurality of close times.

The block diagram showing the structure of the fifth embodiment is the same as that of the first embodiment, but the operations of the image feature extraction means 2 and the external environment judgment means 4 are different. FIG. 8 is a flowchart illustrating the operation of the present embodiment.

Next, the operation of this embodiment will be described. The operations of the image input unit 1, the image dividing unit 3, the background difference detecting unit 5, the moving object detecting unit 6, the moving object tracking unit 7, and the background image updating unit 8 are the same as those in the first embodiment. The operation of the image feature extraction unit 2 and the external environment determination unit 4 will be described with reference to the flowchart of FIG.

In step S41, the image feature extracting means 2 obtains an image feature of the input image partial area. Step S42
Then, the external environment determining means 4 obtains image features obtained in advance and stored in the external environment determining means 4, such as day / night / dusk /
The image feature obtained at 0 hour / hour in light rain, heavy rain, and fog is compared with the image feature obtained by the image feature extraction means 2. Specifically, if the current time is 12:14, the time is 12: 0
Noon at 0 minutes and at 11:00 and 13:00
The difference between the image feature in the night / dusk / light rain / heavy rain / fog situation and the image feature obtained in step S41 is obtained, and the image feature having the smallest difference value is obtained. In step S43, the weather condition corresponding to the image feature having the smallest difference value is set as the current external environment.

As described above, according to the present embodiment, a plurality of image features having image features at different times and close to the current time are set as reference image features, and the reference image features and the current image features are used. Because we decided to judge the external environment,
The external environment can be determined with higher accuracy even when there is seasonal variation in sunshine.

Sixth Embodiment In the first embodiment, the external environment is determined in all of the defined divided regions. In the present embodiment, however, a divided region overlapping with a previously detected moving object region is determined. To be excluded from

The block diagram showing the configuration of the sixth embodiment is the same as that of the first embodiment, but the operations of the image feature extraction means 2 and the external environment judgment means 4 are different. FIG. 9 is a flowchart illustrating the operation of the present embodiment.

Next, the operation of this embodiment will be described. The operations of the image input unit 1, the image dividing unit 3, the background difference detecting unit 5, the moving object detecting unit 6, the moving object tracking unit 7, and the background image updating unit 8 are the same as those in the first embodiment. The operation of the image feature extraction unit 2 and the external environment determination unit 4 will be described with reference to the flowchart of FIG.

In step S51, based on the result of the moving object detecting means 6 at the previous execution, the image feature extracting means 2 excludes the partial area where the moving object is present from the image feature calculation target, and performs the image feature calculating as in the first embodiment. Ask for. For example, the image feature of the excluded partial area is set to a negative value (-1). Step S5
In 2, the external environment determination unit 4 excludes the partial area where the moving object exists from the image feature calculation target, and determines the external environment as in the first embodiment. For example, when the image feature of the partial area is a negative value (-1), it is not determined to be a determination target.

As described above, according to the present embodiment, the divided area overlapping with the previously detected moving object area is excluded from the judgment object, so that if the image feature of the moving object part is erroneously judged as the judgment object, Therefore, it is possible to reduce the erroneous determination of the external environment, and it is possible to reduce the processing amount of the external environment determination unit 4.

[0065]

As described above, according to the present invention, an external environment is determined based on the comparison result by individually comparing the image information of a partial area including an area other than an arbitrary road surface with a predetermined reference. This eliminates the possibility of erroneously determining the external environment due to the headlights and street lights of the vehicle.

According to another aspect of the present invention, only the divided regions that are significantly different from the image features at different times are set as the determination targets of the external environment determining means, so that a slight change in the sun may cause an error. There is no need to determine the external environment.

According to another aspect of the present invention, the determination is also made based on the color distribution in the divided area. Therefore, it is possible to determine that the visibility is deteriorated due to rain, fog, etc. When an error occurs, the vehicle detection result is set to 0, thereby preventing an incorrect vehicle detection result.

According to another aspect of the present invention, image features at different times are obtained in advance, and a stored image feature closest to the current time is set as a reference image feature. Since the external environment is determined, it is possible to determine the external environment with higher accuracy even when image characteristics change due to sunshine fluctuation due to sunset such as at twilight.

According to another aspect of the present invention, a plurality of image features having image features at different times and close to the current time are set as reference image features, and the external environment is determined from the reference image features and the current image features. Therefore, the external environment can be determined more accurately even when there is a seasonal variation in sunshine.

According to another aspect of the present invention, a divided region overlapping with a moving body region detected in advance is excluded from the determination target, so that the image feature of the moving body part is not erroneously determined as a determination target and an erroneous external feature is determined. The determination of the environment can be reduced, and the processing amount of the external environment determining means 4 can be reduced.

[Brief description of the drawings]

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

FIG. 2 is an explanatory diagram of a moving object detecting means 6 according to the first embodiment.

FIG. 3 is a flowchart illustrating the operation of the first embodiment.

FIG. 4 is a flowchart of FIG. 4 for explaining the operation of the second embodiment.

FIG. 5 is a flowchart illustrating the operation of the third embodiment.

FIG. 6 is a diagram illustrating an operation of an external environment determination unit according to the fourth embodiment.

FIG. 7 is a flowchart illustrating the operation of the fourth embodiment.

FIG. 8 is a flowchart illustrating the operation of the fifth embodiment.

FIG. 9 is a flowchart illustrating the operation of the sixth embodiment.

FIG. 10 is a configuration diagram of a conventional traffic monitoring device.

FIG. 11 is a diagram illustrating a vehicle detection method of a conventional traffic monitoring device.

FIG. 12 is a flowchart illustrating the operation of a conventional traffic monitoring device.

[Explanation of symbols]

1: image input means, 2: image feature extraction means, 3: image division means, 4: external environment determination means, 5: background difference detection means, 6: moving object detection means, 7: moving object tracking means, 8: background image update Means 10, 11: moving body region at time t-1, 12, 13: moving body region at time t.

 ────────────────────────────────────────────────── ─── Continued on the front page F term (reference) 5B057 AA16 BA02 CA01 CA08 CA12 CA16 DA06 DB02 DB06 DB09 DC25 DC32 5H180 AA01 CC04 DD03 5L096 AA02 AA06 BA04 CA02 FA32 GA06 GA08 GA19 HA03

Claims (6)

[Claims] 1. An image input unit for inputting a captured vehicle image, an image dividing unit for dividing an input image from the image input unit into a predetermined size, and an image feature extracting unit for obtaining an image feature in the divided area. An external environment determination unit that determines an external environment of an imaging target based on the image characteristics; a background difference image generation unit that generates a background difference image from a previously captured and stored background image and the input image; A traffic monitoring device, comprising: a moving object detection unit that detects a moving object; and a moving object tracking unit that tracks the moving object. 2. The traffic monitoring device according to claim 1, wherein the external environment determination unit is configured to determine only the divided regions having significantly different features of the background image at different times. 3. The image feature extracting means according to claim 1, wherein said image feature extracting means comprises an RGB
The traffic according to claim 1, wherein a distribution of hue and luminance is extracted from the value, and the external environment determining means is configured to determine a visibility deterioration state by rain and fog based on the luminance and the hue. Monitoring device. 4. The image feature extraction means holds in advance the background image features at a plurality of different times, and sets the background image feature at the time closest to the current time as the reference image feature, 2. The traffic monitoring apparatus according to claim 1, wherein an external environment is determined from the features and the current image features. 5. The image feature extracting means holds background image features at different times in advance, sets background image features at a plurality of times close to the current time as reference image features, and the external environment determining means includes the plurality of reference image features. 5. The method according to claim 4, wherein the external environment is determined from the current image characteristics and the current image characteristics.
The traffic monitoring device as described. 6. The apparatus according to claim 1, wherein said external environment determining means excludes image characteristics of a region detected as a moving object by said moving object detecting means from a determination target. The traffic monitoring device as described.