JP2012190214A - Image recognition device for vehicle, image recognition method for vehicle, program and medium - Google Patents

Abstract

A vehicle image recognition apparatus, a vehicle image recognition method, a program, and a medium that can improve recognition performance more effectively without increasing processing load.
An image recognition apparatus for a vehicle includes an imaging unit, an extraction unit that extracts a crossable area 25 from a captured image by the imaging unit, and a road in the captured image at a boundary line based on the outline of the crossable area 25. Dividing means for dividing the road surface area including the sidewalk into area parts α to ζ, selection means for selecting a pedestrian pattern corresponding to each of the area parts α to ζ, and pedestrians in each of the area parts α to ζ Recognizing means for recognizing a pedestrian by collation using a pattern.
[Selection] Figure 3

Description

  The present invention relates to an image recognition apparatus for a vehicle using a monocular camera, an image recognition method for a vehicle, a program, and a medium.

  When recognizing and detecting pedestrians with an inexpensive configuration using a monocular camera, postures including pedestrian orientation, stationary state, and walking state vary widely, so prepare pedestrian patterns according to pedestrian orientation, It is effective to collate each in turn. However, there is a problem that the processing load increases in proportion to the number of directions, responsiveness decreases, and quick recognition cannot be performed.

  As a conventional technique for dealing with such a problem, for example, as described in Patent Document 1, it is conceivable to detect a pedestrian movement vector and prepare a pedestrian pattern according to the movement vector.

JP 2009-237897 A

  However, this method is effective for a moving pedestrian, but has a problem that a stationary pedestrian cannot be recognized. In particular, among pedestrians in a stationary state, a pedestrian pattern is significantly different between a left-sided pedestrian and a right-sided pedestrian, and thus there is a problem in that recognition performance is low in recognition using a database having these different pedestrian patterns.

  Therefore, an object of the present invention is to provide a vehicular image recognition apparatus, a vehicular image recognition method, a program, and a medium that can improve the recognition performance more effectively without increasing the processing load.

In order to achieve the above object, a vehicle image recognition apparatus according to the present invention includes:
Imaging means;
Extraction means for extracting a crossable region from the image captured by the imaging means;
A dividing unit that divides a region of a road surface including a road and a sidewalk in the captured image into a region portion at a boundary line based on an outline of the crossable region;
Selecting means for selecting a pedestrian pattern corresponding to each of the region portions;
Recognizing means for recognizing a pedestrian by collation using the pedestrian pattern in each of the region portions.

In order to achieve the above object, a vehicle image recognition method according to the present invention includes:
Imaging step;
An extraction step of extracting a crossable region from an image captured by the imaging means;
A division step of dividing a road surface area including a road and a sidewalk in the captured image into a region portion at a boundary line based on an outline of the crossable area;
A selection step of selecting a pedestrian pattern corresponding to each of the region portions;
A recognition step of recognizing a pedestrian by collation using the pedestrian pattern in each of the region portions.

  A program according to the present invention is a program for executing the vehicle image recognition method, and a medium according to the present invention is a medium storing the program.

  According to the present invention, the recognition performance can be improved more effectively without increasing the processing load.

It is the block diagram which showed the structure of the image recognition apparatus for vehicles 1 of Example 1 which is one Embodiment of this invention. It is the figure which showed the captured image 21 of the advancing direction of the own vehicle obtained with the monocular camera 11. FIG. FIG. 3 is a schematic diagram illustrating a manner of dividing an area in a captured image 21 into an area portion according to the first embodiment. It is a schematic diagram which shows the specific example of the aspect which shows the stationary state of the pedestrian pattern used in Example 1, a walking state, and direction. It is the flowchart which showed the example of a process which extracts the pedestrian crossing 25 (crossing possible area | region). It is a figure for demonstrating the pedestrian crossing 25 on the captured image 21 of FIG. It is a flowchart which shows the whole process of the image recognition apparatus 1 for vehicles of the present Example 1. FIG. It is the block diagram which showed the structure of the image recognition apparatus 31 for vehicles of Example 2 which is one Embodiment of this invention. It is the flowchart which showed the example of a process which calculates the reference distance L by extracting the pedestrian crossing 25 (crossing possible area | region). It is the figure which showed the reference distance L to the pedestrian crossing 25 which has the pixel width a on the captured image 21. FIG. It is a schematic diagram which shows the division | segmentation aspect into the area | region part of the area | region in the captured image 21 in Example 2. FIG. It is a schematic diagram which shows the relationship between the pixel number of a pedestrian pattern used in Example 2, and distance. It is a flowchart which shows the whole process of the image recognition apparatus 31 for vehicles of Example 2. FIG. It is a schematic diagram which shows the recognition form of the pedestrian of the image recognition apparatus 31 for vehicles of Example 2 based on a comparison with a prior art.

  Hereinafter, embodiments for carrying out the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing a configuration of a vehicle image recognition apparatus 1 according to an embodiment of the present invention. The vehicular image recognition apparatus 1 is mounted on a vehicle and includes a monocular camera 11, a crossable area extraction unit 12, a division unit 13, a selection unit 14, and a recognition unit 15. Part or all of the functions of each unit can be realized by a processor, a microcomputer, or the like including an ECU (Electronic Control Unit) that executes image recognition processing.

  The monocular camera 11 is an imaging unit that captures an image of the periphery of the host vehicle (for example, the traveling direction of the host vehicle) at a predetermined angle of view and outputs the captured image. The traveling direction of the host vehicle may be the forward direction or the reverse direction. The monocular camera 11 includes an image sensor such as a CCD or a CMOS.

  The crossable area extraction unit 12 is an extraction unit that extracts a pedestrian crossing that is a crossable area from an image captured by the monocular camera 11.

  A pedestrian crossing is generally composed only of a combination of a thick white line extending in the front-rear direction of the road and a thin white line extending in the left-right direction or a thick white line. The crossable area may be a character, a symbol, or a combination thereof marked on the road surface by paint or the like, and a condition that the pedestrian signal is blue or there is no vehicle traveling on the road Refers to the area where pedestrians can safely cross. The size of the pedestrian crossing is based on dimensions predetermined by laws and regulations.

  FIG. 2 is a view showing a captured image 21 of the traveling direction of the host vehicle obtained by the monocular camera 11. 22 indicates a pedestrian, 23 indicates a white line on the side, and 24 indicates a white chain line of a separation band. The captured image 21 includes a pedestrian crossing 25 marked on the course ahead of the host vehicle as a crossable area. The code | symbol 26 represents a vanishing point (infinity point). In the case of FIG. 2, the crossable area extracting unit 12 extracts candidate lines p <b> 1 to p <b> 4 constituting the pedestrian crossing 25 from the captured image 21, for example. The captured image 21 is an image in which the number of pixels in the horizontal direction is W and the number of pixels in the vertical direction is H.

  As shown in FIG. 3, the dividing unit 13 includes nine road surface areas including the road and the sidewalk in the captured image 21 at the boundary line based on the outline of the pedestrian crossing 25 extracted by the crossable area extracting unit 12. Thus, the dividing means divides into six types of region portions α to ζ.

  The selection unit 14 is a selection unit that selects a database including a pedestrian pattern corresponding to each of the above-described region portions, and the recognition unit 15 uses a database including a pedestrian pattern corresponding to each of the region portions. Recognition means for recognizing a pedestrian by matching.

  In the first embodiment, the boundary lines described above are left and right extending in the left-right direction along each of the front edge and the rear edge of the outline of the pedestrian crossing 25 as shown in FIG. 3 which is an overhead view from the information of FIG. It includes direction lines L1 and L2, and includes front and rear direction lines L3 and L4 extending in the front and rear direction along the left end edge and the right end edge of the contour. Note that the front and rear direction lines L3 and L4, which are the latter boundary lines, coincide with the white lines on the side of the road that are easier to detect.

  Furthermore, in the first embodiment, the pedestrian pattern includes the pedestrian's stationary state or the walking state and any direction of the front, back, left and right. Further, in the region portions γ and δ located adjacent to the left and right of the crossable region, that is, the region portion ε shown in FIG. 3, each pedestrian pattern is as shown second from the left and left in FIG. It corresponds to a stationary state.

  That is, as shown in FIG. 3, in the crossable region, that is, the region portion γ located adjacent to the left of the region portion ε, the pedestrian pattern is directed to the right, and is positioned adjacent to the right of the crossable region, that is, the region portion ε. In the area portion δ to be operated, the pedestrian pattern is facing left.

  Further, in the region portion β located in front of the left or right of the region portion ε that is a crossable region, the pedestrian pattern corresponds to the walking state shown second from the right in FIG. In a region portion α located adjacent to the left rear or right rear of the region portion ε, which is a possible region, the pedestrian pattern corresponds to the walking state shown on the rightmost side in FIG.

  In the region portion ε itself, the pedestrian pattern is in a walking state and corresponds to leftward or rightward. In the region portion ζ before or after the region portion ε, the pedestrian pattern corresponds to the omnidirectional direction. As is clear from the illustration of FIG. 3, the region portion ζ is a road, the region portions α and β are sidewalks, and the region portions γ and δ are sidewalks located on both sides of the pedestrian crossing 25.

  FIG. 5 is a flowchart showing an example of processing for extracting the pedestrian crossing 25. The flowchart will be described with reference to FIGS.

  In step S <b> 10, the crossable area extracting unit 12 extracts an edge in the vertical direction (vertical direction) where the luminance of the pixel suddenly changes from the captured image 21 based on the pixel luminance distribution of the captured image 21.

  In step S20, the crossable area extracting unit 12 classifies the edges extracted in step S10 according to a luminance change pattern in the horizontal direction (left and right direction when the pedestrian crossing 25 in FIG. 2 is viewed from the host vehicle). The traversable region extraction unit 12 determines that the left region is “rising edge” when the left region is darker than the right region based on the pixel luminance distribution of the left region and the right region of each edge, and the left region is more than the right region. If it is bright, it is determined as a “falling edge”.

  In step S <b> 30, the crossable area extracting unit 12 searches the captured image 21 in the horizontal direction, and sets the lines in which the edges classified in step S <b> 20 appear alternately and repeatedly as candidate lines constituting the pedestrian crossing 25. Group. In step S40, the traversable region extraction unit 12 searches the captured image 21 in the vertical direction, and determines each candidate line according to the adjacent relationship between the grouped candidate lines and the connection relationship of the edges constituting the candidate line. Are extracted, and the contours of the pedestrian crossing 25 and the pedestrian crossing 25 are extracted.

  For example, the traversable region extraction unit 12 has a predetermined number or more pairs (for example, three or more pairs) with a pair of rising edges and falling edges adjacent to each other, and about those predetermined number or more pairs. When the horizontal (left and right) distances between the two edges constituting each pair (for example, distances b1, b2, and b3 in the case of FIG. 6) are within a predetermined range, each pair is designated as a crosswalk. Are extracted as candidate lines that form a group 25, and the extracted candidate lines are grouped as a pedestrian crossing 25.

  The recognition unit 15 captures a pedestrian in the vicinity of the pedestrian crossing 25 existing in the traveling direction of the host vehicle (for example, on a road on which the pedestrian crossing 25 is marked or on the side of the road), and is a captured image obtained by the monocular camera 11. 21 is a recognition means for recognizing by pattern recognition such as template matching using the pedestrian pattern as shown in FIG.

  FIG. 7 is a flowchart illustrating the entire processing of the vehicular image recognition apparatus 1 according to the first embodiment. Hereinafter, the entire process of the first embodiment will be described with reference to FIG.

  In step S50, the region portion ε that is the pedestrian crossing 25 is extracted as described in the description of step S40 in the flowchart of FIG. 5, and in step S60, the crossable region extracting unit 12 extracts the pixel luminance distribution of the captured image 21. The white line 23 that is an edge in the vertical direction (vertical direction) where the luminance of the pixel changes suddenly and extends in the front-rear direction and reaches the vanishing point 26 is extracted from the captured image 21.

  In step S <b> 70, the dividing unit 13 divides the region in the captured image 21 into six types of region portions α to ζ, and uses these region portions α to ζ as search areas for pedestrians. Select a database containing pedestrian patterns corresponding to.

  In step S80, the recognition unit 15 recognizes each region portion α to ζ using a database including a corresponding pedestrian pattern.

  The vehicle image recognition method of the present invention is executed based on the flowchart shown in FIG. The program of the present invention may be installed or stored in advance in the vehicle image recognition apparatus 1 of the first embodiment, and this program may be stored in a predetermined medium.

  According to the vehicular image recognition apparatus 1 of the first embodiment, a search is performed using the corresponding pedestrian pattern for each of the divided area portions α to ζ, and a pedestrian is recognized. Can reduce the pedestrian pattern included, reduce the processing load, increase the processing speed, and increase the responsiveness and accuracy of recognition.

  Particularly in the first embodiment, the crossable area is the pedestrian crossing 25, and it is assumed in advance that there are many pedestrians waiting for crossing in the area portions γ and δ shown in FIG. In view of the fact that there are many cases where the direction is fixed to either the left or right, and that the pedestrian is in the walking state and the direction is either left or right in the region portion ε, the pedestrian included in the corresponding database Since the number of patterns can be reduced as much as possible, the processing load can be reduced and the recognition accuracy can be improved more effectively.

  Also, in the region portion α located in front of the traveling direction of the host vehicle from the pedestrian crossing 25, the pedestrian pattern included in the corresponding database is considered in view of the fact that there are many cases where the pedestrian is facing the front and in a walking state. The number can be reduced, the processing load can be reduced, and the recognition accuracy can be increased. Similarly, in the area portion β located after the pedestrian crossing 25, in view of the fact that there are many cases where the pedestrian is facing the back and is in a walking state, the processing load can be reduced and the recognition accuracy can be increased.

  According to the recognition by the pedestrian pattern matching with the division of the region to be searched in the first embodiment, the entire captured image 21 is the omnidirectional pedestrian similar to the region portion ζ in the first embodiment. Compared with the case of searching and recognizing a pedestrian in a database including patterns, the processing time can be set to about 40%.

  In the first embodiment described above, the size of the pedestrian crossing 25 and the distance from the host vehicle are not obtained. Of course, after obtaining the size and distance, a boundary line is set and the region is set to the region portions α to ζ. It can also be divided.

  In the schematic diagram shown in FIG. 3, the region portions α to ζ have portions that overlap each other, but it is considered that the overlap margin absorbs the pedestrian from each region portion. Is set as appropriate. In addition, the pedestrian pattern may be in the form of displaying belongings, and other methods may be used for expressing the stationary state, the walking state, and the orientation.

  Next, after obtaining the reference distance L from the host vehicle to the pedestrian crossing 25, the area in the captured image 21 can be divided only in the front-rear direction. The second embodiment will be described below.

  Hereinafter, the form for implementing this invention about this Example 2 is demonstrated, referring drawings. FIG. 8 is a block diagram showing the configuration of the vehicle image recognition apparatus 31 according to the embodiment of the present invention. The vehicle image recognition device 31 is mounted on a vehicle, and recognizes a monocular camera 11, a crossable area extraction unit 12, a division unit 13, a selection unit 14, a size detection unit 16, and a distance calculation unit 17. Part 15. Some or all of the functions of each unit can be realized by a processor, a microcomputer, or the like that executes image recognition processing, as in the first embodiment.

  The monocular camera 11 is an imaging unit that captures an image of the periphery of the host vehicle (for example, the traveling direction of the host vehicle) at a predetermined angle of view and outputs the captured image. The traveling direction of the host vehicle may be the forward direction or the reverse direction. The monocular camera 11 includes an image sensor such as a CCD or a CMOS.

  The crossable area extracting unit 12 is a unit that extracts a predetermined part of a pedestrian crossing that is a crossable area from an image captured by the monocular camera 11. The predetermined part extracted by the crossable area extracting unit 12 is predetermined in design, and may be the whole or a part of the pedestrian crossing.

  As shown in FIG. 2 already shown, the captured image 21 includes a pedestrian crossing 25 marked on the course ahead of the host vehicle as a crossable area. The code | symbol 26 represents a vanishing point (infinity point). In the case of FIG. 2, the crossable area extracting unit 12 extracts candidate lines p <b> 1 to p <b> 4 constituting the pedestrian crossing 25 from the captured image 21, for example. The captured image 21 is an image in which the number of pixels in the horizontal direction is W and the number of pixels in the vertical direction is H.

  The size detection unit 16 is a means for detecting the size (preferably the lateral width) on the image captured by the monocular camera 11 for a predetermined part of the pedestrian crossing 25 extracted by the crossable area extraction unit 12. For example, the size detection unit 16 can detect the size of the predetermined part on the captured image by acquiring the number of pixels constituting the predetermined part of the pedestrian crossing 25 from the captured image.

  The distance calculation unit 17 includes a predetermined size in the standard for a predetermined portion of the pedestrian crossing 25 extracted by the crossable area extraction unit 12, a size actually detected by the size detection unit 16, and a monocular camera. 11 is a means for calculating a reference distance L from the own vehicle to the pedestrian crossing 25 based on the angle resolution of 11.

  FIG. 9 is a flowchart showing a processing example until the reference distance L to the pedestrian crossing 25 is calculated. This flowchart will be described with reference to FIGS. 2, 6, and 10.

  In step S <b> 10, the crossable area extracting unit 12 extracts an edge in the vertical direction (vertical direction) where the luminance of the pixel suddenly changes from the captured image 21 based on the pixel luminance distribution of the captured image 21.

  In step S20, the crossable area extracting unit 12 classifies the edges extracted in step S10 according to a luminance change pattern in the horizontal direction (left and right direction when the crossable area is viewed from the host vehicle). The traversable region extraction unit 12 determines that the left region is “rising edge” when the left region is darker than the right region based on the pixel luminance distribution of the left region and the right region of each edge, and the left region is more than the right region. If it is bright, it is determined as a “falling edge”.

  In step S <b> 30, the crossable area extracting unit 12 searches the captured image 21 in the horizontal direction, and sets the lines in which the edges classified in step S <b> 20 appear alternately and repeatedly as candidate lines constituting the pedestrian crossing 25. Group. Then, the traversable region extracting unit 12 searches the captured image 21 in the vertical direction, and groups each candidate line according to the adjacent relationship between the grouped candidate lines and the connection relationship of the edges constituting the candidate line. The pedestrian crossing 25 is extracted.

  For example, the traversable region extraction unit 12 has a predetermined number or more pairs (for example, three or more pairs) with a pair of rising edges and falling edges adjacent to each other, and about those predetermined number or more pairs. When the horizontal (left and right) distances between the two edges constituting each pair (for example, distances b1, b2, and b3 in the case of FIG. 6) are within a predetermined range, each pair is designated as a crosswalk. Are extracted as candidate lines constituting 25, and the plurality of extracted candidate lines are grouped as crossable areas.

  In step S40-1, the size detection unit 16 detects the number of pixels (width) in the left-right direction of the pedestrian crossing 25, and the distance calculation unit 17 uses the detected number of pixels in the left-right direction. The reference distance L from the monocular camera 11 to the pedestrian crossing 25 is calculated.

As shown in FIG. 6, among the pairs extracted as the crossable area by the crossable area extraction unit 12, the size detection unit 16 is, for example, the third from the left from the rising edge p <b> 1 a of the leftmost pair. Three pairs of pixel widths a corresponding to the distance to the falling edge p3b of the pair are detected. The width length between both ends of the three pairs on the standard of the pedestrian crossing 25 (that is, a value required in the standard of the part corresponding to the portion of the pixel width a) is A, and the angular resolution in the horizontal direction of the monocular camera 11 is Assuming that P is the reference distance L, as is clear from FIG.
L = A / tan (a × P) (1)
It is expressed by the relational expression. Here, the angle resolution P is defined as “P = when the horizontal angle of view of the monocular camera 11 is α [degrees] and the number of pixels of the captured image 21 obtained by the monocular camera 11 is W [pix.]. It is a fixed value that can be represented by “α / W [degree / pix.]”. That is, the distance calculation unit 17 can calculate the reference distance L to the pedestrian crossing 25 based on the formula (1). The distance calculating unit 17 calculates the reference distance L ′ to the front end of the pedestrian crossing 25 by the same method as calculating the reference distance L for the length of the pedestrian crossing 25 in the front-rear direction, and calculates the reference distance L Calculated by subtracting the reference distance L from '.

  Of course, as long as it is a pair of edges extracted as a traversable region, the pixel width between any edges may be detected. If the detected pixel width is a, the part corresponding to the portion of the pixel width a The standard value corresponds to the fixed value A.

  Therefore, the vehicular image recognition apparatus 1 includes the monocular camera 11, the crossable area extraction unit 12, the size detection unit 16, and the distance calculation unit 17 having the above-described configuration. Considering the above size, the reference distance L to the crossable area is calculated, so the reference distance L to the pedestrian crossing 25 is determined even if the crossable area exists far from the vehicle. It can be calculated with high accuracy from an image captured by a monocular camera.

  By the way, a certain tolerance is recognized in the standard value of the size of the predetermined part of the pedestrian crossing 25. For example, the standard value of one white line area of the pedestrian crossing 25 (that is, a high brightness area sandwiched between a rising edge and a falling edge of a pair) is 45 cm to 50 cm. Currently, there is a difference in the values adopted.

  In order to absorb such a difference, when the predetermined part of the pedestrian crossing 25 is at a predetermined position in the vertical direction on the captured image, the size (for example, the length in the horizontal direction) of the predetermined part on the captured image is determined. The value (calibration value) obtained by calibrating the standard value of the size of the predetermined part is calculated. By using such a calibration value for the distance calculation after the next time, the calculation accuracy can be improved. For example, the calibration value F1 for the standard value F of the size of the white line area is calculated by using the pixel width f constituting the white line area of the pedestrian crossing 25 when it is at a predetermined position on the image. Used to calculate the distance.

  As shown in FIG. 11, the dividing unit 13 has five areas of the road surface including the road and the sidewalk in the captured image 21 on the boundary line based on the outline of the pedestrian crossing 25 extracted by the crossable area extracting unit 12. Dividing means for dividing into parts η to λ.

  The selection unit 14 is a selection unit that selects a database including a pedestrian pattern corresponding to each of the above-described region portions, and the recognition unit 15 uses a database including a pedestrian pattern corresponding to each of the region portions. Recognition means for recognizing a pedestrian by matching.

  In the second embodiment, as shown in FIG. 11, the boundary line includes a plurality of left and right direction lines L5 to L8 extending in the left and right direction along or parallel to the front end edge and the rear end edge of the contour. As shown in FIG. 12, the lengths in the front-rear direction of the region portions η to λ are set so that the number of pixels in the captured image 21 corresponding to this length is equal to each other (here, 20 pixels). As shown in FIG. 12, the number of pixels in the captured image 21 corresponding to the pedestrian pattern is set according to the distance from the vehicle to each of the region portions η to λ. The larger the distance, the smaller the number of pixels. Is done.

  In the second embodiment, a distance calculation unit 17 is included. The distance calculation unit 17 calculates a reference distance L to the crossable area extracted by the crossable area extraction unit 12 serving as an extraction unit, and the reference distance L It also functions as a calculation means for calculating the above-described distance from the region portion η to λ based on the above.

  In the specific example shown in FIG. 12, the reference distance L of the pedestrian crossing 25 is 25 m, the length of the pedestrian crossing 25 itself in the front-rear direction is 10 m, and the center distance in the front-rear direction is, for example, 30 m. The rear end edge of the region portion η, which is the rear end edge of the region portion θ, is set to a distance of 20 m from the own vehicle, the rear end edge of the region portion ι is set to a distance of 25 m from the own vehicle, and the region portion κ The rear end edge is set at a distance of 35 m from the own vehicle, and the rear end edge of the region portion λ is set at a distance of 60 m from the own vehicle.

  FIG. 13 is a flowchart illustrating the entire processing of the vehicular image recognition apparatus 31 according to the second embodiment. The overall processing of the second embodiment will be described below with reference to FIG.

  In step S90, as shown in the description of step S40-1 in the flowchart of FIG. 9, the area portion ι including the pedestrian crossing 25 is extracted, and the reference distance L from the host vehicle is calculated.

  In step S100, the dividing unit 13 divides the region in the captured image 21 into five region parts η to λ arranged in parallel in the front-rear direction, and selects these region parts η to λ as a pedestrian search region. 14 selects a database including a pedestrian pattern in which the number of pixels is set to be small as the distance corresponding to each increases.

  In step S110, the recognition unit 15 recognizes a pedestrian using the database including the corresponding pedestrian pattern selected in the previous step for each region portion η to λ.

  The vehicle image recognition method of the present invention is executed based on the flowchart shown in FIG. The program of the present invention may be installed or stored in advance in the vehicle image recognition apparatus 31 of the second embodiment, and this program may be stored in a predetermined medium.

  According to the vehicular image recognition apparatus 31 of the second embodiment, as shown above the double arrows in the vertical double line in FIG. 14, the corresponding walking is performed for each divided region portion η to λ. As the pedestrian is recognized by searching using a person's pattern, the search area is reduced and the processing load is reduced as compared with the conventional technique in which the entire area is set as the search area as shown below the arrow in FIG. Thus, the processing speed can be increased, and the responsiveness and accuracy of recognition can be increased.

  In particular, in the second embodiment, the number of pixels constituting the pedestrian pattern is gradually decreased from the front toward the heel corresponding to the distance, so that the search area can be more effectively reduced. In addition, by setting the number of pixels of the pedestrian pattern according to the distance in advance, it is possible to distinguish between an object that may be recognized as a pedestrian and a real pedestrian more precisely in the prior art. Can prevent misrecognition.

  In the schematic diagram shown in FIG. 11, the region portions η to λ have portions that overlap each other, but the overlap margin absorbs the protrusion of each pedestrian from each region portion as in the first embodiment. It is set as appropriate in consideration of this. Also in the second embodiment, the pedestrian pattern may be in the form of displaying belongings.

  The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the above-described embodiments, and various modifications and substitutions can be made to the above-described embodiments without departing from the scope of the present invention. Can be added.

  For example, the division in the direction perpendicular to the front-rear direction in the second embodiment can also be applied in the first embodiment. For example, the region portions α, β, and ζ shown in FIG. Can do.

DESCRIPTION OF SYMBOLS 1, 31 Image recognition apparatus for vehicles 11 Monocular camera (imaging means)
12 Crossable area extraction unit (extraction means)
13 Dividing part (dividing means)
14 Selection part (selection means)
15 Recognition part (recognition means)
16 Size detector 17 Distance calculator (calculation means)
21 Imaged image 22 Pedestrian 23 White line 24 White chain line 25 Crosswalk (crossable area)
26 Vanishing point (infinity point)
p1-p4 candidate line L reference distance L 'reference distance L1-L4 boundary line α-ζ region portion L5-L8 boundary line η-λ region portion

Claims (16)

  An imaging means, an extracting means for extracting a crossable area from a captured image by the imaging means, and a road surface area including a road and a sidewalk in the captured image as a boundary line based on a contour of the crossable area as an area portion A dividing means for dividing; a selecting means for selecting a pedestrian pattern corresponding to each of the area parts; and a recognizing means for recognizing a pedestrian by collation using the pedestrian pattern in each of the area parts. A vehicular image recognition apparatus comprising:   The vehicular image recognition device according to claim 1, wherein the boundary line includes a left-right direction line extending in a left-right direction along each of a front end edge and a rear end edge of the contour.   The vehicular image recognition device according to claim 2, wherein the boundary line includes a front-rear direction line extending in a front-rear direction along each of a left end edge and a right end edge of the contour.   The vehicular image recognition apparatus according to claim 3, wherein the pedestrian pattern includes a stationary state or a walking state of the pedestrian and any direction of the front, back, left and right.   The vehicular image recognition device according to claim 4, wherein the pedestrian pattern corresponds to the stationary state in the area portion adjacent to the left and right of the crossable area.   In the region portion located adjacent to the left of the crossable region, the pedestrian pattern is directed to the right, and in the region portion located adjacent to the right of the crossable region, the pedestrian pattern is directed to the left. The vehicular image recognition apparatus according to claim 5, wherein the vehicular image recognition apparatus is provided.   In the area portion located at the left front or right front of the crossable area, the pedestrian pattern corresponds to the walking state and is front-facing, and is adjacent to the left rear or right rear of the crossable area. The vehicular image recognition apparatus according to claim 6, wherein the pedestrian pattern corresponds to the walking state and faces the back in the region portion to be performed.   The vehicular image recognition apparatus according to claim 1, wherein the boundary line includes a plurality of left and right direction lines extending in the left and right direction along or parallel to the front end edge and the rear end edge of the contour.   The vehicular image according to claim 8, wherein the length of each of the region portions in the front-rear direction is set so that the number of pixels in the captured image corresponding to the length is equal to each other. Recognition device.   The vehicular image recognition apparatus according to claim 8, wherein the number of pixels in the captured image corresponding to the pedestrian pattern is set according to a distance from the vehicle to each of the region portions.   The vehicle image recognition apparatus according to claim 10, further comprising a calculation unit that calculates a reference distance to the crossable area extracted by the extraction unit, and calculates the distance based on the reference distance.   The vehicular image recognition apparatus according to any one of claims 1 to 11, wherein the crossable area is a pedestrian crossing.   The vehicular image recognition apparatus according to claim 12, wherein the pedestrian pattern includes belongings.   An imaging step, an extraction step for extracting a crossable area from the image captured by the imaging means, and a road surface area including a road and a sidewalk in the captured image at a boundary line based on an outline of the crossable area A dividing step for dividing, a selecting step for selecting a pedestrian pattern corresponding to each of the region portions, and a recognition step for recognizing a pedestrian by collation using the pedestrian pattern in each of the region portions. An image recognition method for a vehicle, comprising:   The program which performs the image recognition method for vehicles of Claim 14.   A medium storing the program according to claim 15.

Images