JP3379416B2 - Object detection device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an object detecting device, and more particularly to an object detecting device that can be suitably used for detecting the presence or absence of a vehicle in each parking section of a parking lot.

[0002]

2. Description of the Related Art It is necessary to detect the presence / absence of a vehicle in each parking section in order to confirm an empty space in a parking lot.
Various attempts have been made to realize this by image processing. For example, in Japanese Unexamined Patent Publication No. 4-216182, a feature vector of a predetermined area is calculated from a reference image and a learning image to design an object identification system for the area, thereby detecting the presence or absence of an object in the area in the processed image. A method of doing so is shown (above, the first conventional example).

Further, Japanese Patent Application Laid-Open No. 7-41118 discloses that
An apparatus for detecting the presence / absence of an object by comparing the average density and density dispersion of a predetermined area with a reference average density and density dispersion is shown (above, second conventional example).

[0004]

However, in the above-mentioned first conventional example, since it is necessary to obtain an image in which no object exists as a reference image, when a parking lot or the like is targeted, a timing when there is no vehicle is present. In some cases, it is difficult to obtain the image at the time, and the reference image and the processed image to be determined are obtained at different times. However, there is a problem that the determination accuracy is lowered due to the positional deviation. Also,
There is also a problem that the calculation is relatively complicated.

Further, in the second conventional example, since the density and the density variance are directly compared with the reference value, there is a problem in that they are easily affected by the sunlight and the reliability of the determination is lacking.

An object of the present invention is to solve such a problem, and an object thereof is to provide an object detecting device which can detect the presence or absence of an object simply and with high reliability without requiring a reference image or the like.

[0007]

In order to achieve the above object, the present invention is a device for detecting the presence or absence of an object in a predetermined place, and an image pickup means (2) for obtaining an image of the predetermined place, Evaluation values calculated based on the size of the small area and the rank assigned in descending order of the size, by integrating pixels having similar properties in the image area obtained by the image capturing unit to form a small area. With a predetermined threshold to determine the presence or absence of an object (3, steps 103 to 106)
And have. The "property" is density, color, texture or the like. The presence / absence of an object is determined, for example, when there is a small number of large small areas in the image, it is determined that there is no object, and when there are a large number of small areas in the image, it is determined that there is an object. To do. The size of the small area is represented by the area, the number of pixels, the perimeter, and the like.
Here, the evaluation value f can be calculated by the following mathematical expression (1).

[Equation 4] Here, k is the rank of the small area, Sk is the area of the k-th small area, Sa is the area of the entire image area, and n is the largest of the ranks k of the small areas having a certain size or more.

In the present invention, when there is no object at a predetermined place, the obtained image is an image having only a substantially uniform background, so that a large number of large image regions having similar properties occur in this image. On the other hand, if there is an object at the above-mentioned predetermined location, a large number of small image areas having similar properties are generated in the image corresponding to each part of the object. Therefore, by knowing the size and the distribution of image areas having similar properties in the image, it is possible to reliably determine the presence or absence of an object at a predetermined location.

According to the present invention, the presence or absence of an object can be detected easily and with high precision because a reference image as in the prior art is not required, and since the density and the density dispersion are not directly compared, the influence of the sunshine It is possible to detect with high reliability without receiving the error.

The present invention can also be implemented as a method,
In this case, the object detection method obtains an image of a predetermined place, calculates the size and the number of image regions having similar properties in the image, and determines the presence or absence of the object based on the calculated size.

[0011]

BEST MODE FOR CARRYING OUT THE INVENTION (First Embodiment) In the present embodiment, a case will be described below in which the present invention is applied to the detection of the presence or absence of a vehicle in a parking lot (that is, a parking space).

In FIG. 1, a plurality of parking sections 11 are formed in a parking lot 1 with a white line 12 as a boundary, and a TV camera 2 as an image pickup means is installed at a high place with its direction fixed. Predetermined range L including all parking areas 11
Are getting the image of.

As shown in FIG. 2, an image obtained by the television camera 2 is converted into a digital signal by the A / D converter 31 of the image processing apparatus 3 and then stored in the frame memory 32 and stored in the input interface 33. Through the CPU 34
The presence / absence of a vehicle in each parking zone 11 is determined by the processing that is taken in by and described later. The determination result is displayed on the display device 4 via the output interface 35.

FIG. 3 shows a procedure of image processing in the CPU 34. Prior to the image processing, in step 101, an image division area M corresponding to each parking division 11 is set. This image division area M is normally the white line 1 forming the parking division 11.
2 is a quadrangle, but when there is a blind spot area of the TV camera 2 when there is a vehicle in the adjacent parking section 11, as shown in FIG. 1, it should be slightly shifted from the actual parking section 11. good. Explaining the image division area M further, as shown in FIG. 4, vertices P1 to P1 of a polygon (the figure shows the case of a quadrangle) surrounding the corresponding parking division 11 are shown.
P4 is taken in the image, and the following image processing is performed on each pixel 3 in each image division area M, with the straight line connecting the vertices P1 to P4 and the area surrounded by these straight lines as the image division area M. . Therefore, the image division area M
Does not necessarily have to be along the white line 12 as long as it corresponds to each parking section 11, and each parking section 11 does not necessarily have to be partitioned by the white line 12.

That is, in step 102 of FIG. 3, area division is performed. This is for sequentially adjoining, for example, adjacent pixels 3 having close density levels within the image division area M and dividing the areas into small areas. For example, the method proposed by Yokoya et al. No.8, pp730-737 (1978)) can be used. In this step 102, as shown in FIG. 5, when the area division is performed on each of the adjacent image division areas M1 and M2 in which no vehicle exists, each of the image division areas M1 and M2 as shown in FIG. It is divided into a relatively small number of small regions 13 including a region occupying a considerably large area with respect to the entire area. This is because most of the image division areas M1 and M2 are divided into large areas having similar density levels corresponding to the road surface, and there are areas corresponding to dirt such as white lines and oil drawn on the road surface. This is because it is divided into areas having similar density levels.

On the other hand, in step 102, when the image division areas M3 and M4 adjacent to each other in which the vehicle exists as shown in FIG.
As shown in, each of the image division areas M3 and M4 is divided into a large number of small areas 13 each having a relatively small area with respect to the entire area. This is because the regions corresponding to the components such as the vehicle hood, window glass, roof, bumper, and front grille are partitioned as regions having similar concentration levels. Note that the image division areas M1 and M2 and the image division areas M3 and M4 have different overall areas because the photographing angles from the television camera 2 are different.

In step 103 of FIG. 3, the area of each partitioned small area 13 is calculated, and in the following step 104,
The area ratio (%) of each small area 13 to the total area of the image partition areas M1 to M4 in which they are present is calculated. FIG. 9 shows the small areas 13 in the image partitioned areas M1 and M2 where there is no vehicle, which are numbered and arranged in descending order of area ratio. As is clear from the figure, the image division area M1 is divided into nine small areas from No. 1 having an overwhelmingly large area ratio to Nos. 2 to 9 having a sharp area ratio, and the image division area M2. Are divided into three small regions from No. 1 having an overwhelmingly large area ratio to No. 2 and No. 3 where the area ratio sharply decreases.

On the other hand, in FIG. 10, the small areas 13 in the image divided areas M3 and M4 in which the vehicle is present are arranged by numbering in order from the largest area ratio.
3 has 21 small areas from No. 1 having a relatively large area ratio to No. 2 to 21 having a small area ratio, and the image division area M4 has a gradually increasing area ratio from No. 1 having a relatively large area ratio. It is divided into 16 small areas from 2 to 16 that are smaller.

In step 105 of FIG. 3, an evaluation value f for determining the presence or absence of a vehicle is calculated by the following equation (1).

[0020]

[Equation 1]

Here, k is the rank of the small area, n is the largest of the ranks k satisfying the following equation (2), and Sk is the k-th rank.
The area of the numbered small area, Sa is the area of the entire image division area. The area Sk for which the evaluation value is to be calculated is a certain size or more that satisfies the following expression (2).

[0022]

[Equation 2]

The reason is that the area of the image portion corresponding to the license plate or the headlight, which is relatively small in appearance among the constituent parts of the vehicle, is usually 1% or more of the entire area of the image division area, and therefore 1 This is because it is considered that a small area having an area of less than% does not correspond to a component part of the vehicle but corresponds to dirt on the road surface or the like.

FIG. 11 shows the accurate area ratio of each small area 13 in each image division area M1 to M4 and the evaluation value f of each image division area M1 to M4 obtained from the above equation (1) based on this. Show. As is apparent from the figure, the evaluation value f of the image division areas M1 and M2 without the vehicle is small as it is close to “1”, and the evaluation value f of the image division areas M3 and M4 with the vehicle is small.
Is a value close to "4", which is large.

Therefore, in step 106 of FIG. 3, the threshold value is set to, for example, "2" and it is determined that "the vehicle is present" when the evaluation value f is larger than this value, and "there is no vehicle" when it is smaller. Then, in step 107, the determination result of vehicle presence / absence is displayed on the display device 4 for each parking section corresponding to the image section areas M1 to M4.

According to an experiment conducted by the inventor, the presence or absence of a vehicle in each parking section in a parking lot was determined under various weather conditions by using the above-described evaluation value calculation formula and a threshold value. As a result, the correct answer rate was 99% or more. It was possible to make a decision, and the result of this decision did not depend on the time of day, the sunshine, or the difference in the parking lot. Note that the above threshold values are merely examples, and are experimentally determined according to various situations. Further, a plurality of types of threshold values may be prepared and used properly according to the type of vehicle for which the presence / absence of parking is determined.

(Other Embodiments) The TV camera for picking up an image of the parking lot does not necessarily have to be fixed, and may be a camera that moves so as to scan the entire parking lot. In this case, it is necessary to control the attitude of the television camera so that a predetermined field of view can be obtained at each scanning position.

The expression for calculating the evaluation value is not limited to the above expression (1), and the value of the evaluation value becomes small when there are a small number of small areas that occupy the majority in the image division area, A calculation formula may be used that increases the evaluation value when the image division area is divided into a large number of small areas having a relatively small area. Therefore, the following equation (3) can be used. According to this, it is not necessary to order the small regions in the descending order of area as in the first embodiment, and therefore the image processing speed is improved.

[0029]

[Equation 3]

In the above equation (3), w is a weight, for example Sk
/Sa<0.2, w = 1; Sk / Sa ≧ 0.2, w =
Set to 0.1.

In the first embodiment, the division into small areas and the calculation of the area thereof are performed for each image division area. However, the entire image obtained from the television camera is divided into the small areas and the area thereof is calculated. After that, the evaluation value may be calculated for each image division area.

In the first embodiment, the integrated division into the small areas is performed for the pixels having similar density levels, but the integrated division may be performed based on other properties such as color and texture.

In the first embodiment, the area is used to determine the size of the small area, but the number of pixels, the perimeter, etc. can be used.

The present invention can be applied not only to the presence / absence of parking in a parking lot but also to the presence / absence of luggage in a warehouse.

The program used in the hardware system to implement the present invention can be provided in a state recorded on a medium. As the medium storing the program, for example, a flexible disk, a CD-ROM, a memory card or the like can be used. The program recorded in the medium is installed in a memory incorporated in the hardware system, for example, a hard disk device to execute the program to realize the present invention.

[0036]

As described above, according to the present invention, the presence or absence of an object can be detected easily and with high reliability without requiring a reference image or the like.

[Brief description of drawings]

FIG. 1 is a schematic perspective view showing a parking lot and a television camera that images the parking lot.

FIG. 2 is a block configuration diagram of a parking space detection device.

FIG. 3 is a flowchart showing an image processing procedure in a CPU.

FIG. 4 is a diagram showing an image division area.

FIG. 5 is a diagram showing an image of a parking section without a vehicle.

FIG. 6 is a diagram showing an image division area divided into areas.

FIG. 7 is a diagram showing an image of a parking section in which a vehicle is located.

FIG. 8 is a diagram showing an image division area divided into areas.

FIG. 9 is a graph showing the respective small areas arranged in order and the area ratio thereof in the image division area where there is no vehicle.

FIG. 10 is a graph showing the respective small areas arranged in order and the area ratio thereof in the image division area where the vehicle is present.

FIG. 11 is a diagram showing an area ratio of each small area in an image divided area where there is no vehicle and an image divided area where there is a vehicle, and an evaluation value calculated from these.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Parking lot, 11 ... Parking area, 13 ... Small area (image area), 2 ... Television camera, 3 ... Image processing apparatus, 34 ... C
PU, M, M1, M2, M3, M4 ... Image division area.

─────────────────────────────────────────────────── ─── Continuation of front page (51) Int.Cl. ⁷ Identification code FI G08G 1/14 G08G 1/14 A (72) Inventor Yamamoto Shin Aichi-gun Nagakute-cho, Aichi-gun, Nagakute-cho, Yokoshiro, Yoko 41 Toyota Central Research Institute (56) Reference JP 5-324954 (JP, A) JP 9-102040 (JP, A) (58) Fields investigated (Int.Cl. ⁷ , DB name) G06T 1 / 00 330 B65G 1/137 E04H 6/00 G06T 7/00 300 G08G 1/04 G08G 1/14

Claims (2)

(57) [Claims] 1. An apparatus for detecting the presence or absence of an object at a predetermined place, comprising: an image pickup means for obtaining an image of the predetermined place; and a pixel having similar characteristics in an image area obtained by the image pickup means. Integrated into a small area, the evaluation value calculated based on the size of the small area and the rank assigned in the descending order of the size is compared with a predetermined threshold value, and a determination unit for determining the presence or absence of an object. An object detection device comprising. 2. The evaluation value f is calculated from the following mathematical expression (1).
The object detection device according to claim 1, which calculates. [Equation 3] Here, k is the order of the small area, Sk is the surface of the k-th small area
Product, Sa is the area of the entire image area, and n is a certain size or more.
It is the largest of the ranks k of the small areas.