JPH05314389A - Vehicle kind recognizing device - Google Patents

Abstract

(57) [Summary] [Purpose] To make it possible to easily configure a vehicle type recognition device into various systems that meet market needs. The vehicle type recognition device 10 includes a comparison data storage circuit 11 in which normalized window frame closed curve data ND _{1 to} ND _{N of} various vehicles are stored in advance, and an identification target vehicle image I sent from the television camera 1. Circuit 12 in which is stored
An extraction circuit 13 for extracting window frame closed curve data D from the identification target vehicle image I stored in the memory circuit 12; and a normalization processing circuit 14 for normalizing the window frame closed curve data D extracted by the extraction circuit 13. The normalized window frame closed curve data ND _{1 to} ND _N stored in the comparison data storage circuit 11 are read one by one, and the read normalized window frame closed curve data ND _{1 to} ND _N and the normalization processing circuit 14 are read. A discriminating circuit 15 for discriminating the type of the vehicle photographed by the television camera 1 by comparing the window frame closed curve data D ′ that has been normalized.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a vehicle type recognition device,
In particular, the present invention relates to a vehicle type recognition device that photographs a vehicle that is stopped or running with a television camera to determine the type of the vehicle.

[0002]

2. Description of the Related Art Conventionally, as a vehicle type recognition device of this type,
There is one that determines the type of a vehicle based on a license plate photographed by a television camera for the purpose of classifying vehicles for automatic collection of tolls on toll roads (for example, Japanese Patent Laid-Open No. 60-24700). ). However, since the data showing the correspondence between the license plate and each vehicle is not open to the public at present, such a vehicle type recognizing device cannot precisely determine the types of many vehicles.

An apparatus for extracting a vehicle from a straight edge in a vehicle image has also been proposed (for example, Japanese Unexamined Patent Publication No. Heisei 3-
83474). However, considering that a lot of rounded vehicles have appeared in recent years, it is difficult to extract the vehicles even by using such a device, and therefore the types of the vehicles cannot be determined in detail. In addition, if there is a carrier or loaded luggage, it is easy to erroneously determine the type of vehicle.

Further, as a vehicle type recognition apparatus capable of discriminating the type of vehicle in detail, only an image of a vehicle portion is cut out from an image taken by a television camera, and the image data of the cut out image and pre-registered image data are extracted. A vehicle type recognition device that determines the type of vehicle by comparison has been proposed (for example, JP-A-1-46890). However, in such a vehicle type recognition device, in addition to the problem that the amount of data and the amount of calculation required when collating image data that is two-dimensional data becomes enormous, the image data is more There is also a problem that the image data necessary for accurate matching cannot be easily normalized.

[0005]

Therefore, in the above-described conventional vehicle type recognition apparatus, it is not easy to determine the type of vehicle in detail, so that various systems that meet the market needs as described below can be easily configured. There is a problem that you cannot do it. (1) An automatic monitoring system that can determine the type of vehicle in detail even if the license plate cannot be recognized due to stolen vehicles, bad weather, and detachment due to fraud. (2) Even if the parking ticket is lost in a large machine-transportation type parking lot and information such as entry and departure times is unknown and the owner cannot clearly remember the license plate number of the own vehicle, it is subject to the type of vehicle A system that can find a vehicle. (3) An automatic traffic flow measurement system that can determine the type of vehicle in detail.

An object of the present invention is to provide a vehicle type recognition device which can easily configure various systems that meet market needs.

[0007]

The vehicle type recognition apparatus of the present invention is a vehicle type recognition apparatus for discriminating the type of a vehicle from an image of the vehicle taken by a TV camera. Extracted window frame closed curve data from the respective stored comparison data storage circuits, a memory circuit in which the identification target vehicle image sent from the television camera is stored, and the identification target vehicle image stored in the memory circuit Extracting circuit, a normalization processing circuit for normalizing the window frame closed curve data extracted by the extraction circuit, and reading the normalized window frame closed curve data stored in the comparison data storage circuit one by one. Then, the read normalized window frame closed curve data is compared with the window frame closed curve data normalized by the normalization processing circuit, and the image is taken by the television camera. The and a discrimination circuit for discriminating the type of the vehicle.

Here, each of the normalized window frame closed curve data indicates the shape of the window frame of the side window of each of the various vehicles, and the window frame closed curve data of the vehicle photographed by the television camera. It may indicate the shape of the window frame of the side window.

[0009]

The vehicle type recognizing device of the present invention makes it possible to easily make a detailed determination of the vehicle type by determining the vehicle type by paying attention to the shape of the window frame of the vehicle for the following reason. .. (1) Assuming that the shape of the window frame of the side window of the vehicle is used, the shape of the upper part of the cabin of the vehicle can be regarded as the main elements of the side window and the pillars. It is considered that the shape of the constructed closed curve indirectly reflects the shape of the rooftop of the vehicle. (2) When the closed curve formed by the window frame is regarded as a mathematical shape, several types of shapes are clearly found even in the vehicles currently on the market. (3) The data of the closed curve formed by the window frame is a sequence of coordinate points (one-dimensional data), and although the vehicle shape is reflected, the actual shape of the vehicle (three-dimensional data) and the vehicle image (two-dimensional data). The amount of data is extremely small compared to (data), and the direction histogram, Fourier expansion, and generalized Hough
Data can be easily compared by signal processing techniques such as conversion.

That is, in the vehicle type recognition apparatus of the present invention, the discriminator circuit reads the normalized window frame closed curve data stored in the comparison data storage circuit one by one, and the read normalized window frame closed curve data and the normalized window frame closed curve data. By comparing the window frame closed curve data that has been subjected to the normalization processing by the signal processing circuit and determining the type of the vehicle photographed by the television camera, detailed determination of the vehicle type can be easily performed.

[0011]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of a vehicle type recognition device of the present invention.

The vehicle type recognizing device 10 includes a television camera 1
The type of the vehicle is discriminated from the image of the vehicle photographed in 1. The normalized window frame closed curve data ND _{1 of} various vehicles
To ND _N are respectively stored in advance, the comparison data storage circuit 11, the memory circuit 12 in which the identification target vehicle image I sent from the television camera 1 is stored, and the memory circuit 1
Extraction circuit 13 for extracting the window frame closed curve data D from the identification target vehicle image I stored in 2, and a normalization processing circuit 1 for normalizing the window frame closed curve data D extracted by the extraction circuit 13.
4 and the normalized window frame closed curve data ND _{1 to} ND _N stored in the comparison data storage circuit 11 one by one, and the read normalized window frame closed curve data ND _{1 to} ND _N and the normalization processing circuit. It includes a discriminating circuit 15 for discriminating the type of the vehicle photographed by the television camera 1 by comparing the window frame closed curve data D ′ that has been normalized in 14.

Next, regarding the operation of the vehicle type recognition device 10,
An example will be described in which the television camera 1 is installed on the side of the vehicle, the vehicle is observed, and the type of the vehicle is determined from the shape of the window frame on the side surface of the vehicle.

The vehicle to be recognized is photographed by the television camera 1 so that the side window is completely included and the side window is located substantially in the center of the image. It should be noted that such photographing can be easily realized even when the vehicle is moving by using a conventionally used light-shielding sensor or ultrasonic sensor. In addition, since the height position of the side window of the vehicle can be roughly known by using the vehicle height measurement using ultrasonic waves or the like, it is possible to easily photograph the side window completely. FIG. 2 shows an example of the identification target vehicle image I captured by the television camera 1. The identification target vehicle image I is displayed from the television camera 1 to the memory circuit 1.
2 and is stored in the memory circuit 12.

The extraction circuit 13 reads out the identification target vehicle image I from the memory circuit 12, classifies the vehicle body portion of the identification target vehicle image I and the inside of the window frame, and then extracts and extracts the closed figure by a known labeling process. A closed figure showing the window frame of the side window is estimated and selected from the size and position of the closed figure, and the window frame closed curve data D is extracted from the identification target vehicle image I (see FIGS. 3 to 5). In the extraction circuit 13, a color TV camera is used as the TV camera 1 because the paint color on the vehicle body surface is uniform in general vehicles except for bright-colored special vehicles and competition vehicles displaying sponsor names. The body part and the inside of the window frame are classified by color information. Thereby, even when a part or all of the window glass of the side window is vacant, the vehicle body part of the identification target vehicle image I and the inside of the window frame can be classified. In addition, when a stable light source is set indoors and the vehicle is photographed so that the vehicle portion occupies most of the observation area from diagonally above in the vicinity,
The vehicle body portion of the identification target vehicle image I and the inside of the window frame can be particularly easily classified.

The normalization processing circuit 14 normalizes the window frame closed curve data D extracted by the extraction circuit 13,
This is because the window frame closed curve data D differs depending on the direction of observation of the vehicle by the TV camera 1, so that it is the same as the normalized window frame closed curve data ND _{1 to} ND _{N of} various vehicles stored in the comparison data storage circuit 11 in advance. This is for converting the window frame closed curve data D into the observation direction condition. The normalization processing circuit 14 uses a known perspective projection based on a camera parameter that is generally used in the field of machine pigeons, computer graphics, etc., which indicates the positional relationship between the television camera 1 and the vehicle in the observation conditions. The window frame closed curve data D extracted by the extraction circuit 13 is normalized by a mapping operation called exchange.

In the discriminating circuit 15, the normalized window frame closed curve data ND _{1 to} ND _N read out one by one from the comparison data storage circuit 11 are normalized by the direction histogram which is one of the known closed figure contour analysis methods. From the window frame closed curve data D ′ that has been normalized by the processing circuit 14, the television camera 1
The type of the vehicle photographed in is determined.

Since each of the normalized window frame closed curve data ND _{1 to} ND _N is a discrete coordinate data string, the center of gravity of the normalized window frame closed curve data ND _{1 to} ND _N is obtained, and the center of gravity is constant. Draw a straight line for each angle, find the distance from the center of gravity to the contour, and convert it to a so-called γ (θ) curve, then discretely sample the normal direction of the minutely divided line segment, and measure its length. A histogram in the normal direction (that is, a direction histogram) is obtained by accumulating all the directions in the same normal direction and scaling so that the total sum for all directions becomes "1". The same applies to the normalized window frame closed curve data D. Therefore, the window frame closed curve data D of most of the vehicles can be regarded as a convex shape. Therefore, with respect to the direction histogram, (1) a closed curve like the window frame closed curve data D and a direction histogram are essentially paired. Corresponds to one. (2) Even if the sample center position changes at the time of conversion from a closed curve such as the window frame closed curve data D to a γ (θ) curve,
It has been theoretically elucidated that the directional histogram is invariant. Therefore, the correlation value between the direction histogram obtained for the normalized window frame closed curve data ND _{1 to} ND _N and the direction histogram obtained for the window frame closed curve data D is calculated,
Normalized window frame closed curve data ND ₁ ~ having the largest correlation value
By selecting the type of vehicle corresponding to ND _N , the type of vehicle captured by the TV camera 1 can be determined.

As described above, the vehicle type recognition device 10 has the side surface.
Focusing on the shape of the window frame to determine the type of vehicle
Therefore, compared to the conventional vehicle type recognition device described above,
It has the following advantages. (1) Vehicle type identification that distinguishes the type of a vehicle from the contour of the vehicle
In the identification device, misidentification due to carrier or loaded luggage
There was a problem, but the vehicle type recognition device 10 has the shape of the window frame of the side window.
The carrier is used to distinguish the type of vehicle
And the loaded luggage does not affect the discrimination performance. (2) Images cut out from images taken with a TV camera
Compare the image data of and the image data registered in advance
In the vehicle type recognition device that determines the type of vehicle, the amount of data and
And the enormous amount of calculation became a problem, but the closed music composed by the window frame
Since the line data is one-dimensional data,
The amount of calculation can be significantly reduced. (3) Determine the type of vehicle by paying attention to the shape of the side window frame
By doing so, the window frame closed curve data required for discrimination
The normalization of data D can be easily performed as described above.
It (4) Regular data stored in the comparison data storage circuit 11 in advance
Window frame closed curve data ND ₁~ ND_NIs a vehicle manufacturer
Pamphlet issued by the company or design registration that can be viewed for a fee
Since it can be easily obtained from the vehicle shape diagram in the official gazette,
Can be more easily dealt with.

In the above description, the extraction circuit 13 is
Although the body part and the inside of the window frame are classified according to the color information, since the body part is generally made of metal and the window frame is generally connected to the glass through a rubber frame, an infrared television camera is used as the television camera 1. May be used to classify the vehicle body portion and the inside of the window frame by thermal information.

The normalization processing circuit 14 normalizes the window frame closed curve data D extracted by the extraction circuit 13, but if the window frame shooting by the television camera 1 is always the same in the three-dimensional arrangement. , The camera parameters are the same, so normalization is not particularly necessary. Further, when a long focus television camera is used as the television camera 1 and the vehicle is photographed from the same photographing direction, the normalization need only be simple scaling.

Although the type of the vehicle is determined from the shape of the window frame of the side window, the type of the vehicle can be similarly determined from the shape of any window frame of the vehicle. However, by determining the type of vehicle from the shape of the window frame of the side window, it is possible to determine the type of vehicle in more detail.

[0024]

Since the present invention is configured as described above, it has the following effects.

In the discriminating circuit, the normalized window frame closed curve data stored in the comparison data storage circuit are read out one by one, and the read out normalized window frame closed curve data and the window which is normalized by the normalization processing circuit. By comparing the frame closed curve data with the type of vehicle captured by the TV camera, it is possible to easily perform detailed determination of the type of vehicle, making it easy to implement various systems that meet market needs. Can be configured.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a vehicle type recognition device of the present invention.

FIG. 2 is a diagram showing an example of an identification target vehicle image captured by the television camera shown in FIG.

FIG. 3 is a diagram for explaining the operation of the extraction circuit shown in FIG.

FIG. 4 is a diagram for explaining the operation of the extraction circuit shown in FIG.

5 is a diagram for explaining the operation of the extraction circuit shown in FIG.

[Explanation of symbols]

1 a television camera 10 vehicle type recognizing apparatus 11 compares data storage circuit 12 memory circuit 13 extracting circuit 14 normalization processing circuit 15 discriminating circuit ND ₁ to ND _N normalized window frame closed curve data I identified target vehicle image D windows closed curve data D ' Normalized window frame closed curve data

Claims (2)

[Claims] 1. A vehicle type recognition device for determining the type of a vehicle from an image of the vehicle taken by a television camera, comprising: a comparison data storage circuit in which normalized window frame closed curve data of various vehicles are stored in advance; A memory circuit that stores the identification target vehicle image sent from the television camera, an extraction circuit that extracts window frame closed curve data from the identification target vehicle image stored in the memory circuit, and an extraction circuit that is extracted by the extraction circuit A normalization processing circuit for normalizing the window frame closed curve data and the normalized window frame closed curve data stored in the comparison data storage circuit one by one, and the read out normalized window frame closed curve data. A discriminating circuit for discriminating the type of the vehicle photographed by the television camera by comparing the window frame closed curve data normalized by the normalizing circuit with the window frame closed curve data. Vehicle type recognition apparatus characterized by comprising. 2. Each of the normalized window frame closed curve data represents a shape of a window frame of a side window of each of the various vehicles, and the window frame closed curve data is a side surface of the vehicle photographed by the television camera. The vehicle type recognition device according to claim 1, wherein the vehicle type recognition device shows a shape of a window frame of the window.