JP2000090268A - Vehicle area detecting method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a vehicle area detecting method improved in the detection accuracy of an edge area belonging to a vehicle. SOLUTION: The set area of link edge areas is detected from the front image provided by image pickup (step S1). Next, the edge strength and length of the link edge area, the total value of horizontal and vertical variant values and edge strength are calculated (steps S2-S6). Then, the respective link edge areas are respectively rearranged in order from highest PWR, highest LEN and smallest VA (step S7), a high-order R (R is a predetermined number of areas scheduled to detect) of link edge areas are judged as the edge areas belonging to the vehicle and a vehicle candidate area is detected (step S8). Thus, the vehicle candidate area can be highly accurately detected.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for detecting a vehicle region, and more particularly, to a method for detecting a vehicle region with high accuracy.

[0002]

2. Description of the Related Art As a device for judging safety in car navigation, a device for recognizing a preceding vehicle (hereinafter referred to as a preceding vehicle) by imaging the front has been put to practical use. In this device, as a method of distinguishing a preceding vehicle from a row of trees,
The fact that an edge image obtained by differentiating a grayscale value of a captured image contains many edge features of a horizontal component and a vertical component is used. That is, the fact that the continuity and strength of the edges of the horizontal component and the vertical component belonging to the image of the distant view are weaker than the edges of the horizontal component and the vertical component belonging to the vehicle candidate area. The vehicle candidate region is, for example, a rectangular image region detected in a forward video, and is a region where it is detected that there is a high possibility that a vehicle is present. Hereinafter, a conventional vehicle area detection method will be described with reference to the drawings and by way of example.

FIG. 3 is a schematic diagram showing an image (hereinafter, referred to as a forward image) obtained by photographing the front of a traveling vehicle with a CCD camera, for example. In the conventional vehicle region detection method, the grayscale value of the forward image is differentiated, and the edge features of the horizontal component and the vertical component are detected from the edge image. Then, a region where many edge features are distributed is determined as an edge region belonging to the preceding vehicle, and a region where many edge features are not distributed such as a row of trees is determined not to be an edge region belonging to the preceding vehicle.
0 has been determined.

[0004]

However, the conventional method of detecting a vehicle region has a problem that the detection accuracy of a vehicle candidate region is low because the detection accuracy of an edge region belonging to a vehicle is low. In view of the circumstances as described above, an object of the present invention is to provide a vehicle area detection method with high detection accuracy of an edge area belonging to a vehicle.

[0005]

In order to achieve the above object, a method for detecting a vehicle region according to the present invention forms an edge image based on a captured front image, and then forms a connected edge based on the edge image. A first step of performing a predetermined filtering process on a captured front image in a vehicle region detection method for forming a region and detecting a vehicle candidate region of a front vehicle from among a set region of connected edge regions; A binarizing process and a labeling process on at least one of a vertical component and a horizontal component of each edge image obtained in the step (a) to form a connected edge region and form a set region of the connected edge regions; For the edge area,
At least one of the horizontal variance value or the vertical variance value, at least one of the horizontal length or the vertical length, and the third step of calculating the edge strength, respectively, at least one of the horizontal variance value or the vertical variance value, A fourth step of extracting a connected edge area constituting the vehicle candidate area from a set area of the connected edge areas based on at least one of the horizontal length or the vertical length and the edge strength. It is characterized by:

As the filter processing, noise removal processing, contrast improvement processing, and differentiation processing are usually performed.

Preferably, in the third step, at least one of a horizontal variance value or a vertical variance value of each edge image constituting the connected edge region, at least one of a horizontal length or a vertical length, and The edge strength is calculated, and the respective average values are calculated.

As an example of the processing performed in the fourth step,
At least one of the horizontal variance value or the vertical variance value, at least one of the horizontal length or the vertical length, and, for the edge strength, respectively, extract a set number of higher-order connected edge regions in the set region. , A connected edge region constituting the vehicle candidate region. As another example of the process performed in the fourth step, a predetermined weight constant is assigned to at least one of the horizontal variance value or the vertical variance value, at least one of the horizontal length or the vertical length, and the edge strength. An evaluation value obtained by the multiplication is calculated, and the connected edge regions are extracted from the set region by the set number in the descending order of the evaluation values, and are set as the connected edge regions constituting the vehicle candidate region.

As still another example of the processing performed in the fourth step, a fifth step of extracting a connected edge area by a first set ratio in the descending order of edge strength in a set area;
It is determined whether or not the number of connected edge regions extracted in the fifth step is larger than the first set number. If the number is larger, at least one of the horizontal variance value and the vertical variance value is determined in the descending order from the highest. A sixth step of further extracting the connected edge regions by the set ratio, and determining whether the number of the connected edge regions extracted in the sixth step is larger than the second set number. And a seventh step of further extracting a connected edge region by a third set ratio in ascending order of at least one of the length in the vertical direction and the length in the vertical direction.

[0010]

Embodiments of the present invention will be described below in detail with reference to the accompanying drawings. Embodiment 1 Embodiment 1 is an embodiment of the present invention. FIG.
FIG. 1 is a schematic diagram showing an image (hereinafter, referred to as a forward image) obtained by capturing an image ahead of a traveling vehicle with a CCD camera, for example, in the present embodiment. In the present embodiment, image processing is performed on the forward video in the procedure described below.

FIG. 2 is a flowchart showing the processing procedure of this embodiment. First, predetermined filter processing such as noise removal, contrast improvement, and differentiation is performed on the forward video obtained by imaging. Next, binarization processing and labeling processing are performed on the detected edge image (differential image) of the vertical component or horizontal component, and a set area of the formed connected edge areas is detected (step S1).

Next, the following evaluation values are calculated for each pixel group of each label after the labeling process (steps S2 to S4). First, in order to evaluate the continuity of a vertical edge and a horizontal edge, an area, a length, a horizontal variance value, and a vertical variance value are calculated as follows. (A) The area is calculated as the number of pixels of the same label pixel by the following equation. S [n] = (total number of pixels belonging to label [n] (n-th label)), that is, (b) The length is calculated by the following formula using the maximum value of the X coordinate and the Y coordinate for the same label pixel. It is calculated as the difference between the minimum values. Length of vertical component edge image: LEN [n] = Xn # max-Xn # mi
n Length of horizontal component edge image: LEN [n] = Yn # max-Yn # mi
n where, Xn # max, Yn # max: maximum X coordinate value and Y coordinate value of pixel belonging to label [n] Xn # min, Yn # min: minimum X coordinate value of pixel belonging to label [n] And Y coordinate values. (C) The horizontal variance value and the vertical variance value are calculated as the variance values for the same label pixel from the vertical component edge image and the horizontal component edge image, respectively, by the following equations. Horizontal variance: VA [n] = (Σ (Yn (i)-Yn # av)) / S [n] Vertical variance: VA [n] = (Σ (Xn (i)-Xn # av)) / S [n] Here, Xn (i), Yn (i): X coordinate value and Y of the pixel belonging to label [n]
Coordinate values Xn # av, Yn # av: The average value of the X coordinate values and the average value of the Y coordinate values of the pixels belonging to the label [n].

Further, the average value of the edge strength in the edge image is calculated for the pixels in the same label for each label of each connected edge area by the following equation (step S5). pn (x, y) = edge strength of pixel belonging to label [n] Edge strength: PWR [n] = Σpn (x, y) / Sn In the vertical component edge image, the horizontal line image is the horizontal component edge image In the figure, vertical line images are respectively detected. Label numbers do not overlap between the vertical component edge image and the horizontal component edge image.

After calculating the above evaluation values, for each connected edge area for each label, the total value of the edge strength is calculated from the evaluation values of the labels belonging to the connected edge area: PWR = ΣPWR [n] Total length : LEN = ΣLEN [n] Total value of horizontal dispersion value and vertical dispersion value: VA = ΣVA [n] is calculated (step S6).

The connected edge areas are rearranged in the order of PWR, LEN, and VA (step S7), and the top R (R is the predetermined number of detection areas) connection is performed. The edge area is determined to be the edge area belonging to the vehicle, and the vehicle candidate area 11 is detected (step S8). Thereby, the vehicle candidate area can be detected with high accuracy. Further, for example, the region 12 (FIG. 1) is efficiently excluded from the vehicle candidate regions.

Second Embodiment In the second embodiment, compared to the first embodiment, step S7 is performed.
Instead of using PWR, LEN, and VA,
The weighting constants A, B, and C, which are set in advance, are respectively multiplied, and the total evaluation point value (POINT) is calculated. POINT = A * PWR + B * LEN + C * VA Next, sort the connected edge areas in descending order of POINT,
A predetermined number of higher-order connected edge areas are determined to be edge areas belonging to the vehicle, and a vehicle candidate area is detected.

Embodiment 3 In this embodiment, when rearranging the connected edge regions,
Focusing on the importance of each evaluation value in the order of PWR, VA, LEN, the following processing is performed instead of step S7 as compared with the first embodiment. First, the connected edge regions are rearranged in descending order of PWR, and the top 20% is extracted. If the number of extracted connected edge regions is larger than R (R is the predetermined number of regions to be detected), the extracted connected edge regions are further rearranged in ascending order of VA, and higher (in the case of a variance value, the value is smaller). 50% is extracted. If the number of extracted connected edge areas is still larger than R, the rearranged area is further sorted in descending order of LEN, and R connected edge areas are extracted from the top, and are set as edge areas belonging to the vehicle. Determine the area.

[0018]

According to the present invention, a connected edge region is formed from a captured front image, and for each connected edge region, at least one of a horizontal variance value or a vertical variance value, a horizontal length or a vertical variance value. At least one of the lengths and the edge strength are calculated, and a connected edge region constituting the vehicle candidate region is extracted based on the calculated values. As a result, the vehicle candidate region can be detected with high accuracy and efficiency.

[Brief description of the drawings]

FIG. 1 is a schematic diagram showing a video image of the front of a traveling vehicle with, for example, a CCD camera in a first embodiment.

FIG. 2 is a flowchart illustrating a processing procedure according to the first embodiment.

FIG. 3 is a schematic diagram showing an image obtained by capturing an image of the front of a traveling vehicle by a conventional method using, for example, a CCD camera.

[Explanation of symbols]

 10 vehicle candidate area 11 vehicle candidate area 12 area

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 5B057 AA16 BA29 CE02 CE11 DA08 DC14 DC16 5L096 BA04 EA35 EA43 FA06 FA32 FA33 FA64 GA02 GA34

Claims (6)

[Claims] 1. An edge image is formed based on a captured front image, a connected edge region is formed based on the edge image, and a vehicle candidate region of a forward vehicle is selected from a set of connected edge regions. In a method for detecting a vehicle region, a first step of performing a predetermined filtering process on a captured front image, and a binarization process of at least one of a vertical component and a horizontal component of each edge image obtained in the first step And a labeling process to form a connected edge region to form a set region of the connected edge regions, and for each connected edge region, at least one of a horizontal variance value or a vertical variance value, a horizontal length or a vertical length. A third step of calculating at least one of the length in the direction and the edge strength, and at least one of the horizontal variance value and the vertical variance value; Extracting at least one of the length in the direction or the length in the vertical direction, and a connected edge region forming the vehicle candidate region from a set region of the connected edge regions based on the edge strength. A vehicle area detection method characterized by the above-mentioned. 2. Filter processing includes noise removal processing,
The vehicle area detection method according to claim 1, wherein a contrast improvement process and a differentiation process are performed. 3. In a third step, at least one of a horizontal variance value or a vertical variance value, at least one of a horizontal length or a vertical length, and an edge intensity of each edge image forming the connected edge region. 3. The method according to claim 1, further comprising:
5. The method for detecting a vehicle region according to claim 1. In a fourth step, at least one of a horizontal variance value or a vertical variance value, at least one of a horizontal length or a vertical length, and an edge strength
The vehicle according to any one of claims 1 to 3, wherein a set number of higher-order connected edge regions in the set region are extracted as connected edge regions constituting a vehicle candidate region. Region detection method. In a fourth step, at least one of a horizontal variance value or a vertical variance value, at least one of a horizontal length or a vertical length, and an edge strength are multiplied by a predetermined weighting constant, respectively. 4. An evaluation value calculated as follows, and a set number of connected edge regions are extracted from the set region in the order of higher evaluation values, and are set as connected edge regions constituting a vehicle candidate region. 2. The method for detecting a vehicle region according to claim 1. 6. In a fourth step, a fifth step of extracting a connected edge area by a first set ratio in descending order of edge strength in the set area, and the number of connected edge areas extracted in the fifth step is the Determining whether the number is larger than the set number of 1, and if so, further extracting a connected edge area by a second set ratio in ascending order of at least one of the horizontal variance value and the vertical variance value; It is determined whether or not the number of connected edge regions extracted in the sixth step is larger than a second set number. If the number is larger, at least one of the horizontal length and the vertical length is determined. The vehicle area detecting method according to any one of claims 1 to 3, further comprising: a seventh step of further extracting a connected edge area by a third set ratio in ascending order.