JP6162064B2 - Object detection processing device - Google Patents

Description

  The present invention relates to an apparatus for detecting an object such as a vehicle using a captured image.

  2. Description of the Related Art Conventionally, a technique for detecting an object such as a vehicle based on an image captured by an imaging device is known. For example, in Patent Document 1, a horizontal edge at the end of the vehicle (a horizontal edge corresponding to the lower side of the vehicle rear surface) when approximating the outer shape of the vehicle with a rectangular parallelepiped model from the horizontal edges extracted from the captured image, A technique is described in which a set of a horizontal edge corresponding to the upper side of the back of the vehicle and a horizontal edge corresponding to the front side of the roof of the vehicle is searched, and a vehicle of a rectangular parallelepiped model represented by those horizontal edges is detected. ing.

JP 2009-245042 A

  The technique described in Patent Document 1 uses an imaging device that is fixedly installed on a road, and detects other vehicles such as a preceding vehicle using a captured image from a running vehicle, for example. When trying to do so, an imaging device is mounted on the vehicle. In this case, a variety of objects (signboards, sign boards, structures, etc.) other than the vehicle often appear in the captured image by the imaging device.

  In such a situation, when trying to detect another vehicle using the technique as disclosed in Patent Document 1, the following inconvenience is likely to occur.

  That is, in the technique shown in Patent Document 1, a horizontal edge corresponding to the upper side of the back surface or the front side of the roof is searched on the assumption that the horizontal edge corresponding to the contour of the outer shape of the vehicle can be extracted.

  However, for example, when the front of the vehicle is imaged by an imaging device mounted on the vehicle, a horizontal edge corresponding to the upper side of the back of the other vehicle in front is clearly extracted due to the influence of the background image, etc. There are cases where this is not possible. In addition, images on the upper side of other vehicles ahead also tend to include images such as signs and sign boards.

  For this reason, a horizontal edge different from the upper side of the rear surface of the other vehicle is easily detected as a horizontal edge corresponding to the upper side of the rear surface of the vehicle. As a result, what is different from the actual other vehicle is likely to be detected as another vehicle, or it is difficult to continuously detect the other vehicle.

  The present invention has been made in view of such a background, and an object that can detect a predetermined type of object such as a vehicle appearing in the captured image with high reliability by using a horizontal edge extracted from the captured image. An object is to provide a detection processing apparatus.

  First, the technical matters that form the basis of the present invention will be described. When an image of a predetermined type of object such as a vehicle is captured in the captured image of the imaging apparatus, even when a horizontal edge is extracted from the captured image, a line corresponding to the contour of the object is not extracted In many cases, a plurality of horizontal edges are extracted from the location of the image of the object.

  For example, in an image of the vehicle viewed from the rear side or the front side, in addition to the horizontal edge corresponding to the contour of the upper end of the vehicle, a plurality of horizontal edges corresponding to window frames, license plates, bumpers, lamp frames, etc. are extracted. There are many cases.

  Thus, in the case of a predetermined type of object from which a plurality of horizontal edges can be extracted from an image, whether or not a horizontal edge corresponding to the contour of the upper end of the object is extracted from the image of the object The horizontal edges to be extracted tend to be arranged in the vertical direction of the captured image at a pitch equal to or smaller than a predetermined value (a pitch smaller than an interval corresponding to the height of the object) or at a density equal to or higher than a predetermined value. There is.

  The present invention has been made paying attention to such a phenomenon.

That is, according to one aspect of the object detection processing device of the present invention, in order to achieve the above object, a process of acquiring a captured image around the imaging device from the imaging device and detecting the object using the captured image. An object detection processing device that performs horizontal edge extraction from a captured image, and when a plurality of horizontal edges are extracted by the horizontal edge extraction device, the image detection image is aligned vertically in the captured image. The first condition that the image is composed of a plurality of horizontal edges, and the vertical distance between each pair of two horizontal edges adjacent to each other in the vertical direction of the captured image of the plurality of horizontal edges is less than or equal to a predetermined value. that the horizontal edge group satisfying a first 2a condition that there and a recognizing object recognition means as a horizontal edge group of a predetermined type of the same object shall be the basic configuration of the first embodiment.

In the present invention, the plurality of horizontal edges arranged in the vertical direction of the captured image is more specifically a portion in which each of the plurality of horizontal edges overlaps with any other horizontal edge in the vertical direction. It means a plurality of horizontal edges arranged in the vertical direction of the captured image so as to have (parts where the positions in the lateral direction of the captured image are the same). This is also true inventions described below.

According to the basic configuration of the first aspect , the object recognizing unit predetermines a horizontal edge group that satisfies the first condition and the second a condition among a plurality of horizontal edges extracted by the horizontal edge extracting unit. Recognize as a horizontal edge group of the same type of object.

  That is, when there are a plurality of horizontal edges arranged in the vertical direction in the horizontal edges extracted from the imaging device, a horizontal edge group (horizontal edge group satisfying the first condition) configured by the plurality of horizontal edges is In the horizontal edge group, the predetermined type of the same object is required to satisfy the second b condition that a vertical distance between two horizontal edges of each pair adjacent to each other in the vertical direction is not more than a predetermined value. This is recognized as a group of horizontal edges extracted for the image.

  In this case, the horizontal edge group recognized as the horizontal edge group of the same object does not need to include a horizontal edge corresponding to the outline of the upper end of the predetermined type object in the captured image. Therefore, regardless of whether or not a horizontal edge corresponding to the contour of the upper end of the predetermined type of object is included, a horizontal edge group that satisfies the first condition and the second a condition is recognized as a horizontal edge group of the same object. .

  In addition, by appropriately setting the predetermined value related to the second a condition, horizontal edges extracted from images other than the predetermined type of object are included in the horizontal edge group satisfying the first condition and the second a condition. It can be avoided as much as possible. For example, when the predetermined type of object is a vehicle traveling on a road, a horizontal edge corresponding to an image of a signboard, a sign board, etc. existing above the road is included in the horizontal edge group that satisfies the first condition and the second a condition. Can be avoided as much as possible.

Therefore, according to the basic configuration of the first aspect, it is possible to detect a predetermined type of object such as a vehicle appearing in the captured image with high reliability by using the horizontal edge extracted from the captured image.

According to another aspect of the object detection processing device of the present invention, in order to achieve the above object, a captured image around the imaging device is acquired from the imaging device, and an object is detected using the captured image. An object detection processing apparatus that performs processing to perform, when a horizontal edge extraction unit extracts a horizontal edge from the captured image, and when a plurality of horizontal edges are extracted by the horizontal edge extraction unit, the vertical direction of the captured image A group of horizontal edges satisfying a first condition that is constituted by a plurality of horizontal edges arranged in a row and a second b condition that the density of the horizontal edges within a predetermined range in the vertical direction of the captured image is a predetermined value or more further comprising a recognizing object recognition means as a horizontal edge group type of the same object shall be the basic configuration of the second embodiment.

According to the basic configuration of the second aspect , the object recognizing unit predetermines a horizontal edge group satisfying the first condition and the second b condition among a plurality of horizontal edges extracted by the horizontal edge extracting unit. Recognize as a horizontal edge group of the same type of object.

  That is, when there are a plurality of horizontal edges arranged in the vertical direction in the horizontal edges extracted from the imaging device, a horizontal edge group (horizontal edge group satisfying the first condition) configured by the plurality of horizontal edges is The horizontal edge extracted from the image of the same object of the predetermined type on condition that the second b condition that the density of the horizontal edges in the predetermined range in the vertical direction of the captured image is equal to or higher than a predetermined value is satisfied. Recognized as a group.

  The horizontal edge density can be expressed as the number of horizontal edges included per unit interval in the vertical direction of the captured image.

In this case, as in the basic configuration of the first aspect, the horizontal edge group recognized as the horizontal edge group of the same object includes a horizontal edge corresponding to the outline of the upper end of the predetermined type object in the captured image. It is not what you need. Therefore, a horizontal edge group satisfying the first condition and the second b condition is recognized as a horizontal edge group of the same object regardless of whether or not it includes a horizontal edge corresponding to the upper end contour of the predetermined type of object. .

  Further, by appropriately setting the predetermined value related to the second b condition, horizontal edges extracted from images other than the predetermined type of object are included in the horizontal edge group satisfying the first condition and the second a condition. It can be avoided as much as possible. For example, when the predetermined type of object is a vehicle traveling on a road, a horizontal edge corresponding to an image such as a signboard or a sign board existing above the object is included in the horizontal edge group satisfying the first condition and the second b condition. Can be avoided as much as possible.

Therefore, according to the basic configuration of the second aspect, it is possible to detect a predetermined type of object such as a vehicle appearing in the captured image with high reliability by using the horizontal edge extracted from the captured image.

In the basic configuration of the first aspect or the second aspect , more specifically, the plurality of horizontal edges in the first condition are, for example, three or more horizontal edges (first reference invention) .

According to this, an object of a type having a plurality of horizontal edges inside the image, such as a vehicle, can be detected with higher reliability.

Furthermore, in the basic configuration of the first aspect , the object recognizing means has another horizontal range between the left and right ends of the lowermost horizontal edge of the horizontal edge group and the other upper side of the lowermost horizontal edge. It is more preferable that the horizontal edge group satisfying the third condition that the horizontal position of the center of the horizontal edge exists as a further condition is recognized as the horizontal edge group of the same object of the predetermined type ( first Invention) . The same applies to the basic configuration of the second aspect (second invention).

According to the first invention or the second invention , in the horizontal edge group recognized as the horizontal edge group of the same type of the same object, each horizontal edge existing above the lowest horizontal edge is the lowest edge. It has an overlapping part in the vertical direction with the horizontal edge. However, at least one of one horizontal end and the other horizontal end of each horizontal edge existing above the lowest horizontal edge may protrude laterally beyond the lowest horizontal edge.

  For this reason, it becomes possible to detect a predetermined type of object having various shapes such as a vehicle.

In the basic configuration of the first aspect , the object recognition means estimates a distance between a position in real space corresponding to at least one horizontal edge of the horizontal edge group and the imaging device, and the distance The predetermined value may be variably set according to the estimated value of the distance so that the predetermined value in the second a condition is reduced as the estimated value of the second is increased ( second reference). Invention ).

Similarly, in the basic configuration of the second aspect , the object recognition unit estimates a distance between a position in real space corresponding to at least one horizontal edge of the horizontal edge group and the imaging device, and The predetermined value may be variably set according to the estimated value of the distance so that the larger the estimated value of the distance is, the larger the predetermined value in the second b condition is ( third). Reference invention ).

That is, the image of the object becomes smaller as the distance between the imaging device and the object in real space increases. Therefore, as in the second reference invention or the third reference invention , the above is estimated according to the estimated value of the distance between the position in the real space corresponding to at least one horizontal edge of the horizontal edge group and the imaging device. The predetermined value in the second a condition or the predetermined value in the second b condition is variably set according to the estimated value of the distance.

Accordingly, in the second reference invention , a horizontal edge group satisfying the first condition and the second a condition, or a horizontal edge group satisfying the first condition and the second b condition in the third reference invention , is set as the horizontal edge of a predetermined type of object. The reliability of recognition as an edge group can be further enhanced. As a result, the reliability of object detection by the horizontal edge group can be further enhanced.

The second reference invention or the third reference invention may be combined with the first reference invention , or the first invention or the second invention .

Further, in the basic configuration of the first aspect , the object recognizing means decreases the predetermined value in the second 2a condition as the length of at least one horizontal edge in the horizontal edge group is shorter. The predetermined value may be variably set according to the length of the at least one horizontal edge ( third invention ).

Similarly, in the basic configuration of the second aspect , the object recognition means increases the predetermined value in the second b condition as the length of at least one horizontal edge of the horizontal edge group is shorter. The predetermined value may be variably set according to the length of the at least one horizontal edge ( fourth invention ).

  That is, the image of the object becomes smaller as the distance between the imaging device and the object in the real space becomes larger, and becomes smaller as the size of the object in the real space such as the horizontal width and the vertical width becomes smaller. Therefore, the longer the distance between the imaging device and the object in the real space, or the smaller the size of the object in the real space, basically, the length of the horizontal edge included in the horizontal edge group corresponding to the object. (Length on the captured image) is shortened.

Therefore, as in the third or fourth invention , the predetermined value in the 2a condition or the 2b condition according to the length (length in the captured image) of at least one horizontal edge in the horizontal edge group. The predetermined value is variably set.

Thereby, in the third invention , the horizontal edge group that satisfies the first condition and the second a condition, or the horizontal edge group that satisfies the first condition and the second b condition in the fourth invention , the distance of the object from the imaging device, Alternatively, it is possible to further increase the reliability of recognition as a horizontal edge group of a predetermined type of object regardless of the size of the object. As a result, the reliability of object detection by the horizontal edge group can be further enhanced.

In addition, as the length of the horizontal edge in the third or fourth invention , for example, the horizontal edge at the lowest end of the plurality of horizontal edges satisfying the first condition (or satisfying the first condition and the third condition) is satisfied. It is preferable to adopt the length. The third invention or the fourth invention may be combined with the first reference invention , the first invention, or the second invention .

In the basic configuration of the first aspect , when the predetermined type of object is, for example, a vehicle, the object recognizing unit is configured to detect the vertical position of each pair of two horizontal edges adjacent to each other in the vertical direction of the captured image. The predetermined value is variably set according to the vertical position of the two horizontal edges of each pair so that the predetermined value in the condition 2a is increased as the value is higher on the upper side. ( 5th invention).

When the predetermined type of object is a vehicle, for example, the vertical width of the vehicle window may be relatively large. Even in such a case, according to the fifth invention, the horizontal edge group to be recognized as the horizontal edge group of the predetermined type of object to be so as to minimize include the horizontal edges in the image of the object Can do.

The fifth invention may be combined with the first reference invention , the first invention, or the second invention .

The figure which shows the system configuration | structure of the object detection processing apparatus in embodiment of this invention. The flowchart which shows the process of the object recognition part shown in FIG. The flowchart which shows the detail of the process of STEP3 of FIG. 2 in 1st Embodiment. The figure which shows the example of the horizontal edge extracted by the process of STEP2 of FIG. The figure for demonstrating the process of STEP3 of FIG. 2 in 1st Embodiment. The flowchart which shows the detail of the process of STEP3 of FIG. 2 in 2nd Embodiment. The figure for demonstrating the process of STEP3 of FIG. 2 in 2nd Embodiment.

[First Embodiment]
A first embodiment of the present invention will be described below with reference to FIGS.

  The object detection processing device 1 of this embodiment is mounted on a vehicle 100 that travels on a road. An image captured by the imaging device 2 is input to the object detection processing device 1 from the imaging device 2. Then, the object detection processing device 1 executes processing for detecting a predetermined type of object such as another vehicle existing in the imaging region of the imaging device 2 using the captured image acquired from the imaging device 2.

  The imaging device 2 is configured by a monocular camera or the like, and is mounted on the vehicle 100 so as to image the periphery of the vehicle 100, for example, the front area. The captured image of the imaging device 2 may be a color image or a monotone image. Further, the imaging device 2 may be a stereo camera configured by two or more cameras.

  The object detection processing device 1 is constituted by an electronic circuit unit including a CPU, a RAM, a ROM, an interface circuit, and the like.

  This object detection processing device 1 performs horizontal edge extraction for executing processing for extracting a horizontal edge from a captured image of the imaging device 2 as a function realized by a mounted program or a function realized by a hardware configuration. Unit 5 and an object recognition unit 6 for recognizing a predetermined type of object in the captured image based on the extracted horizontal edge.

  In the present embodiment, the object to be recognized by the object recognition unit 6 is another vehicle that exists in the imaging region of the imaging device 2 (the front region of the vehicle 100). The object recognizing unit 6 searches for a horizontal edge group satisfying a predetermined condition from a plurality of horizontal edges extracted by the horizontal edge extracting unit 5, and searches each horizontal edge group for an image of the same other vehicle. A process of recognizing that corresponding to (extracted from an image of one other vehicle) is executed.

  The horizontal edge extraction unit 5 and the object recognition unit 6 correspond to the horizontal edge extraction unit and the object recognition unit in the present invention, respectively.

  Next, the process of the object detection processing apparatus 1 will be described in detail. In the following description, the vehicle 100 may be referred to as the host vehicle 100.

  The object detection processing device 1 detects other vehicles existing in the imaging region of the imaging device 2 (the front region of the host vehicle 100) according to the procedure shown in the flowchart of FIG. Note that the flowchart of FIG. 2 does not show the processing for each control processing cycle of the object detection processing device 1, but schematically shows a series of procedures until the other vehicle is finally detected based on the captured image. Show.

  In STEP 1, the object detection processing device 1 acquires a captured image of the imaging device 2. Then, in STEP 2, the object detection processing device 1 executes the process of extracting the horizontal edge from the captured image by the horizontal edge extraction unit 5.

  The horizontal edge extraction unit 5 extracts a horizontal edge using a differential filter or the like based on a change in pixel value (for example, luminance value) in the vertical direction of the captured image. When the captured image is a color image, a horizontal edge is extracted based on hue or saturation instead of the luminance value, or by using the luminance value and hue or saturation together. It may be. Various known methods can be adopted as a method for extracting the horizontal edge.

  By the processing of the horizontal edge extracting unit 5, a plurality of horizontal edges Line are extracted as illustrated in FIG. The example illustrated in FIG. 4 schematically illustrates an example of the horizontal edge Line extracted in a situation where three other vehicles 201, 202, and 203 existing in front of the host vehicle 100 are captured in the captured image of the imaging device 2. Is shown.

  In this case, there are a plurality of tire ground contact points (lines on the road surface) of each other vehicle 201, 202, 203, vehicle body bottom line, window frame line, bumper line, tail lamp line, license plate line, etc. Are extracted from the images of the other vehicles 201, 202, 203.

  As described above, when other vehicles are captured in the captured image of the imaging apparatus 2, generally, a plurality of horizontal edges are extracted from the images of the other vehicles. In addition, the number of horizontal edges extracted from each other vehicle image is usually three or more. Note that the upper end line (roof line) in the contour of the contour of the other vehicle on the captured image is not necessarily extracted as a horizontal edge. For example, as shown in FIG. 4, the horizontal edge corresponding to the uppermost line of the image of the other vehicle 201, 202, 203 may not be extracted.

  Next, the object detection processing device 1 executes the processing of the object recognition unit 6 from STEP3. In STEP 3, the object recognition unit 6 searches for a horizontal edge group of vehicle candidates from the horizontal edges extracted by the horizontal edge extraction unit 5.

  The horizontal edge group of the vehicle candidates is a set of a plurality of horizontal edges that are estimated to have a relatively high possibility of being extracted from an image of another vehicle in the captured image.

  The processing of STEP3 is executed as shown in the flowchart of FIG. Specifically, the object recognition unit 6 selects one horizontal edge among the horizontal edges extracted by the horizontal edge extraction unit 5 as a reference horizontal edge Line (1) in STEP 3-1.

  In this case, the reference horizontal edge Line (1) is selected from the extracted horizontal edges closest to the lower end of the captured image (the one having the lowest height position in the captured image). However, in STEP 3-1 that is executed again after STEP 3-6, which will be described later, a horizontal edge that has already been selected or a horizontal edge that is included in the horizontal edge group of the vehicle candidate is selected. Excluded from the target.

  Further, in STEP 3-1, the object recognition unit 6 initializes the value of the count variable n for counting the number of horizontal edges included in the horizontal edge group of the vehicle candidate to “1”.

  Next, in STEP 3-2, the object recognition unit 6 has a horizontal edge Line (n + 1) that satisfies the following conditions A and B above the horizontal edge Line (n) (n: the value of the current count variable n). Determine if it exists. Line (n + 1) is the horizontal edge closest to Line (n) among the horizontal edges existing above the horizontal edge Line (n).

Condition A:
The horizontal position Cx (n + 1) (the horizontal position of the captured image) of the center of the horizontal edge Line (n + 1) is the left end position sx and the right end position ex of the reference horizontal edge Line (1). It is a position within the range between.
Condition B:
The vertical distance D (n) (distance in the vertical direction of the captured image) between the horizontal edge Line (n + 1) and the horizontal edge Line (n) is equal to or less than a predetermined value Dth.

  When the determination result in STEP 3-2 is affirmative (when there is a horizontal edge Line (n + 1) that satisfies the conditions A and B above Line (n)), the object recognition unit 6 In STEP 3-3, after increasing the value of the count variable n by “1”, the processing in STEP 3-2 is repeated.

  If there are a plurality of horizontal edges arranged in the vertical direction so as to satisfy the conditions A and B by repeating the above, the horizontal edges Line (2), Line (3), ... Are sequentially detected from the lower side to the upper side of the captured image.

  Here, the predetermined value Dth in the condition B is a value set in advance based on experiments or the like so that a plurality of horizontal edges extracted in the image of the other vehicle satisfy the condition B. The predetermined value Dth is basically the height of the image of the other vehicle (in the vertical direction) when another vehicle within a predetermined distance from the host vehicle 100 (or from the imaging device 2) is reflected in the captured image. The width is not exceeded.

  Therefore, corresponding to the image of the other vehicle (for example, the other vehicle 202 shown in FIG. 4) shown in the captured image of the imaging device 2, as shown in FIG. 5, the seven horizontal edges Line (1) to Line (1) to Line (7) When (Line (1) is a horizontal edge corresponding to the line of the ground contact point of the tire of the other vehicle 202), the lowermost horizontal edge Line (1) is STEP3-1. When the reference horizontal edge Line (1) is selected, Line (2) to Line (7) are sequentially detected by the processing of STEP3-2.

  When the determination result in STEP 3-2 is negative (when there is no horizontal edge Line (n + 1) that satisfies the conditions A and B above Line (n)), the object recognition unit 6 In STEP 3-4, it is determined whether or not the current value of the count variable n is 3 or more. The current value of the count variable n indicates the total number of horizontal edges detected by the processing in STEP 3-2 and the reference horizontal edge Line (1).

  When the determination result in STEP 3-4 is positive (when n ≧ 3), n horizontal edges Line (1) to Line (n) are extracted from the image of the other vehicle. Probability is high. In this case, the object recognition unit 6 performs the processing of STEP3-5.

  In STEP 3-5, the object recognizing unit 6 includes n (the horizontal edges Line (2) to Line (n) detected by repeating the processing in STEP 3-2 and the reference horizontal edge Line (1) ( Three or more) horizontal edges Line (1) to Line (n) are determined as a horizontal edge group of vehicle candidates. In this case, the object recognizing unit 6 determines the reference horizontal edge Line (1) as the horizontal edge at the lower end of the vehicle candidate. That is, the horizontal edge Line (1) corresponds to the line at the lowest position of the image of the vehicle candidate that is likely to be another vehicle.

  Next, the object recognition unit 6 executes the determination process in STEP 3-6. In STEP 3-6, the object recognition unit 6 determines whether there is another horizontal edge to be selected that can be selected as the reference horizontal edge. This determination is performed by excluding the horizontal edges already selected in STEP 3-1 and the horizontal edges included in the horizontal edge group determined in STEP 3-4.

  If the determination result in STEP 3-6 is affirmative, the object recognition unit 6 repeats the processing from STEP 3-1 to search for a horizontal edge group of a new vehicle candidate.

  When the determination result in STEP 3-4 is negative (when n <2), there is no horizontal edge that satisfies the conditions A and B above the reference horizontal edge Line (1), or This is a case where there is only one horizontal edge Line (2) that satisfies the conditions A and B. In this case, it is highly possible that the horizontal edge Line (1), or Line (1), Line (2) is not extracted from the image of the other vehicle.

  Therefore, if the determination result in STEP 3-4 is negative, the object recognition unit 6 immediately executes the determination process in STEP 3-6 without executing the process in STEP 3-5.

  By executing the process of searching for the horizontal edge group of the vehicle candidate as described above, the horizontal edge group of the vehicle candidate determined in STEP 3-5 includes a plurality of (three in this embodiment) arranged in the vertical direction of the captured image. The above-mentioned conditions including the horizontal edges Line (1) to Line (n) (first condition in the present invention) and the vertical direction of the captured image of the plurality of horizontal edges Line (1) to Line (n) The vertical distance D (i) between the two horizontal edges Line (i), Line (i + 1) (i = 1, 2,...) Of each pair adjacent to each other is equal to or less than a predetermined value Dth. (Condition 2a in the present invention), and within the horizontal range between the left and right ends of the lowermost horizontal edge Line (1) of the horizontal edge group, the other upper side of the lowermost horizontal edge Line (1) The condition that the horizontal position (position in the horizontal direction of the captured image) of the center of the horizontal edges Line (2) to Line (n) exists (in the present invention) The third condition).

  Returning to the description of FIG. 2, when the horizontal edge group of the vehicle candidate is searched as described above by the processing of STEP 3, the object recognition unit 6 next tracks the vehicle candidate corresponding to each horizontal edge group. The process is executed in STEP4.

  In this tracking process, in each captured image obtained for each predetermined control processing cycle from the imaging device 2 within a period from the time when the horizontal edge group of each vehicle candidate is searched by the process of STEP 3 to a predetermined time later, Whether or not a vehicle candidate image corresponding to the group is included is confirmed by a technique such as image matching.

  Then, after executing this tracking process, the object recognition unit 6 determines whether or not the image of the vehicle candidate corresponding to the horizontal edge group is an actual other vehicle in STEP 5. In this case, when it is confirmed by the tracking process in STEP 4 that the image of the vehicle candidate corresponding to the horizontal edge group is continuously included in the captured image, the object recognition unit 6 corresponds to the horizontal edge group. It is determined that the image of the vehicle candidate to be performed is an actual image of another vehicle.

  When the tracking process in STEP 4 cannot confirm that the image of the vehicle candidate corresponding to the horizontal edge group is continuously included in the captured image, the object recognition unit 6 corresponds to the horizontal edge group. It is determined that the image of the vehicle candidate is not an actual image of another vehicle.

  As described above, when another vehicle exists in the imaging region of the imaging device 2, the presence of the other vehicle is detected using the horizontal edge of the captured image in the captured image.

  According to the present embodiment, the condition that the captured image includes three or more horizontal edges Line (1) to Line (n) arranged in the vertical direction, and the plurality of horizontal edges Line (1) to Line (n). The vertical distance D (i) between two horizontal edges Line (i), Line (i + 1) (i = 1, 2,...) Adjacent to each other in the vertical direction of the captured image is Within the horizontal range between the left and right ends of the lowermost horizontal edge Line (1) of the horizontal edge group and other conditions above the lowermost horizontal edge Line (1) within the horizontal range between the left and right ends of the horizontal edge group. The presence of an image of another vehicle is recognized in the captured image as having as image features a horizontal edge group that satisfies the condition that the horizontal position of the center of the horizontal edges Line (2) to Line (n) exists.

  In this case, since it is not necessary to specify the horizontal edge at the upper end of the contour of the image of the other vehicle, it is possible to prevent the horizontal edge that is not that of the other vehicle from being recognized as a component of the horizontal edge group of the vehicle candidate.

  The horizontal edge group determined as the vehicle candidate horizontal edge group needs to include three or more horizontal edges including the reference horizontal edge Line (1).

  Further, each horizontal edge of the horizontal edge group determined as the horizontal edge group of the vehicle candidate can take various lengths.

  Therefore, according to the present embodiment, other vehicles existing in the imaging region of the imaging device 2 can be detected from the captured image with high reliability.

[Second Embodiment]
Next, a second embodiment of the present invention will be described with reference to FIGS. Note that this embodiment is different from the first embodiment only in the processing of STEP 3 (processing for searching for a horizontal edge group of vehicle candidates) by the object recognition unit 6. For this reason, in the description of the present embodiment, the description of the same matters as in the first embodiment is omitted.

  In the present embodiment, the processing of STEP 3 is executed as shown in the flowchart of FIG. Specifically, the object recognizing unit 6 selects one horizontal edge of the horizontal edges extracted by the horizontal edge extracting unit 5 as the reference horizontal edge Line (1) in STEP 3-11 and count variable The value of n is initialized to “1”. The processing in STEP 3-11 is the same as the processing in STEP 3-1 of the first embodiment.

  Next, in STEP 3-12, the object recognition unit 6 sets the condition described in the first embodiment above the horizontal edge Line (n) within a predetermined range in the vertical direction with the reference horizontal edge Line (1) as the lower end. It is determined whether or not there is a horizontal edge Line (n + 1) satisfying A.

  As shown in FIG. 7, the predetermined range in the vertical direction is a range in which the interval from the lower end (reference horizontal edge Line (1)) to the upper end is the predetermined interval ΔH.

  If the determination result in STEP 3-12 is affirmative (when there is a horizontal edge Line (n + 1) that satisfies the condition A above Line (n)), the object recognition unit 6 determines that STEP3- In step 13, after incrementing the value of the count variable n by “1”, the processing of STEP 3-12 is repeated.

  If there are a plurality of horizontal edges arranged in the vertical direction so as to satisfy the condition A in the predetermined range in the vertical direction, the horizontal edges Line (2), Line (3),. It will be detected in order from the lower side to the upper side of the captured image.

  Here, the interval ΔH of the predetermined range in the vertical direction is the vertical interval of ΔH when the reference horizontal edge Line (1) is the lower end line of the image of the other vehicle (the line of the ground contact point of the tire of the other vehicle). This value is set in advance based on experiments or the like so that a plurality of horizontal edges satisfying the condition A are detected within the predetermined direction range.

  For this reason, as illustrated in FIG. 7, when seven horizontal edges Line (1) to Line (7) are extracted in the image of the other vehicle 202 shown in the captured image of the imaging device 2 (first) In the case similar to FIG. 5 in the embodiment), when the lowermost horizontal edge Line (1) is selected as the reference horizontal edge Line (1) in STEP 3-11, Line (2) to Line (7) is detected in order by the processing of STEP 3-12.

  When the determination result in STEP 3-12 is negative (when there is no horizontal edge Line (n + 1) satisfying the condition A on the upper side of Line (n) within the predetermined range in the vertical direction of the interval ΔH) Then, the object recognition unit 6 determines whether or not the value of the count variable n is 3 or more in STEP 3-14.

  When the determination result in STEP 3-14 is affirmative (when n ≧ 3), the object recognition unit 6 further determines the density of horizontal edges detected within a predetermined range in the vertical direction in STEP 3-15. It is determined whether or not a condition (hereinafter referred to as condition C) that (average density) is equal to or greater than a predetermined value αth is satisfied.

  Here, the density of the horizontal edges is expressed as the number of horizontal edges per predetermined unit interval in the vertical direction of the captured image. Therefore, if the unit interval is Δh (constant value), a value (= n · Δh / ΔH) obtained by multiplying the number n of horizontal edges detected within the predetermined range in the vertical direction by Δh / ΔH (= n · Δh / ΔH) is the predetermined range in the vertical direction. It becomes the density (average density) of the horizontal edges inside.

  For example, in the example shown in FIG. 7, seven horizontal edges including the reference horizontal edge Line (1) are detected within a predetermined range in the vertical direction of the interval ΔH. In this case, the density of horizontal edges in the predetermined range in the vertical direction is 7 · Δh / ΔH. Note that the value of the unit interval Δh can be arbitrarily determined. For example, Δh = 1 or Δh = ΔH can be set.

  The predetermined value αth in STEP 3-15 is set in advance based on experiments or the like so that the density (average density) of a plurality of horizontal edges extracted in the image of the other vehicle is equal to or greater than αth. For example, if the predetermined value αth is set to a value smaller than 7 · Δh / ΔH, the density (= 7 · Δh / ΔH) of the horizontal edges Line (1) to Line (7) in the example shown in FIG. , Becomes a predetermined value αth or more.

  When the predetermined value αth is set to a value of 3 · Δh / ΔH or more, if the density of horizontal edges detected within the predetermined range in the vertical direction of the interval of ΔH is equal to or greater than the predetermined value αth, STEP 3-14 The judgment result is always positive. Therefore, in this case, the determination process in STEP 3-14 may be omitted.

  When the determination results in STEPs 3-14 and 3-15 are affirmative (when n ≧ 3 and the condition C is satisfied (the density of horizontal edges ≧ αth)), n horizontal edges Line There is a high possibility that (1) to Line (n) are extracted from the image of the other vehicle. In this case, the object recognition unit 6 performs the same processing as STEP 3-5 in the first embodiment in STEP 3-16. Accordingly, a horizontal edge group composed of n horizontal edges Line (1) to Line (n) is determined as a horizontal edge group of vehicle candidates.

  Next, the object recognizing unit 6 executes the same determination processing in STEP 3-17 as that in STEP 3-6 in the first embodiment. When the determination result is affirmative (when there is another horizontal edge that can be selected as the reference horizontal edge), the processing from STEP 3-11 is repeated.

  Further, when the determination result of any one of STEPs 3-14 and 3-15 is negative (when n <3 or when the condition C is not satisfied), a horizontal edge within a predetermined range in the vertical direction is detected. Since there are too few, there is a high possibility that the horizontal edge is not extracted in the image of the other vehicle.

  Therefore, if the determination result in any of STEPs 3-14 and 3-15 is negative, the object recognition unit 6 immediately executes the determination process in STEP 3-17 without executing the process in STEP 3-16. Execute.

  By executing the process of searching for the horizontal edge group of the vehicle candidate as described above, the horizontal edge group of the vehicle candidate determined in STEP 3-16 includes a plurality of (three in this embodiment) arranged in the vertical direction of the captured image. The above-mentioned conditions including the horizontal edges Line (1) to Line (n) (first condition in the present invention) and the condition that the density of the horizontal edges within a predetermined range in the vertical direction is not less than a predetermined value αth (present invention). And the other horizontal edge Line above the lowermost horizontal edge Line (1) within the horizontal range between the left and right ends of the lowermost horizontal edge Line (1) of the horizontal edge group. (2) to the condition (third condition in the present invention) that the lateral position (position in the lateral direction of the captured image) of the center of Line (n) exists.

  The present embodiment is the same as the first embodiment except for the matters described above.

  According to this embodiment, the condition that the captured image includes three or more horizontal edges Line (1) to Line (n) arranged in the vertical direction, and the density of the horizontal edges within the vertical range is a predetermined value αth. The above conditions and the other horizontal edge Line above the lowermost horizontal edge Line (1) within the horizontal range between the left and right ends of the lowermost horizontal edge Line (1) of the horizontal edge group. The presence of the image of the other vehicle is recognized in the captured image as having as the image feature a horizontal edge group that satisfies the condition that the horizontal position of the center of (2) to Line (n) exists.

  In this case, as in the first embodiment, there is no need to specify the horizontal edge at the upper end of the contour of the image of the other vehicle, so the horizontal edge that is not that of the other vehicle is recognized as a component of the horizontal edge group of the vehicle candidate. Can be prevented.

  The horizontal edge group determined as the vehicle candidate horizontal edge group needs to include three or more horizontal edges including the reference horizontal edge Line (1).

  Further, each horizontal edge of the horizontal edge group determined as the horizontal edge group of the vehicle candidate can take various lengths.

  Therefore, according to the present embodiment, other vehicles existing in the imaging region of the imaging device 2 can be detected from the captured image with high reliability.

[Modification]
Next, some modifications related to the above-described embodiment will be described. In the first embodiment, the predetermined value Dth related to the condition B is a constant value.

  However, for example, as the pair of horizontal edges Line (i), Line (i + 1) (i = 1, 2,...) Adjacent in the vertical direction is further away from the reference horizontal edge Line (1), The predetermined value Dth to be compared with the vertical distance D (i) between the horizontal edges Line (i) and Line (i + 1) may be increased. For example, the lower horizontal edge Line (i) (or the upper horizontal edge Line (i + 1)) of the horizontal edges Line (i) and Line (i + 1) and the reference horizontal edge Line (1) When the distance in the up and down direction is greater than or equal to a predetermined value, Dth may be set larger than when the distance is smaller than the predetermined value.

  In this case, even when the vertical width of the window on the back of the other vehicle is relatively large, a group of horizontal edges including the horizontal edges at the upper and lower ends of the window can be searched. .

  Supplementally, according to the second embodiment, even when the vertical width of the window on the back of another vehicle is relatively large, a search is made for a group of horizontal edges including the horizontal edges at the upper and lower ends of the window. can do.

  In addition, the predetermined value Dth related to the condition B in the first embodiment or the predetermined value αth related to the condition C in the second embodiment is one of the horizontal edges (for example, the reference horizontal edge Line ( It may be set variably according to the distance in the real space between 1)) and the imaging device 2.

  That is, as the distance between the other vehicle in the real space and the imaging device 2 of the host vehicle 100 increases, the image of the other vehicle in the captured image becomes smaller. Therefore, the pitch of the horizontal edge extracted in the image of the other vehicle (Spacing between horizontal edges adjacent in the vertical direction) is reduced.

  In consideration of this, when a plurality of horizontal edges arranged in the vertical direction are extracted, the position in the real space corresponding to one of the horizontal edges (for example, the reference horizontal edge Line (1)) and the imaging device 2 (or the host vehicle 100) is estimated according to the estimated value of the distance so that the larger the estimated value of the distance is, the smaller the predetermined value Dth related to the condition B in the first embodiment is. Dth is variably set.

  Similarly, αth is variably set according to the estimated value of the distance so that the larger the estimated value of the distance, the larger the predetermined value αth related to the condition C in the second embodiment.

  By doing in this way, the reliability that the horizontal edge group determined as a horizontal edge group of a vehicle candidate is searched as a thing corresponding to the image of an actual other vehicle can be improved further.

  As the distance estimation method, a known method based on a captured image (monocular captured image or stereo image) can be employed. Alternatively, the distance may be estimated using a ranging device such as a radar.

  The predetermined value Dth related to the condition B in the first embodiment or the predetermined value αth related to the condition C in the second embodiment is any horizontal edge that can constitute a horizontal edge group, for example, a reference horizontal edge Line. The length may be variably set according to the length of (1) (the length on the captured image).

  That is, as the distance between the other vehicle in the real space and the imaging device 2 of the host vehicle 100 increases, or as the size (horizontal width and vertical width) of the other vehicle in the real space decreases, the other vehicle in the captured image. The image becomes smaller. Eventually, the pitch of horizontal edges (interval between horizontal edges adjacent in the vertical direction) extracted in the image of the other vehicle becomes small.

  Then, the longer the distance between the other vehicle in the real space and the imaging device 2 of the host vehicle 100 is, or the smaller the size (horizontal width and vertical width) of the other vehicle in the real space corresponds to the other vehicle. Thus, the length of the horizontal edges constituting the horizontal edge group extracted in this manner is shortened.

  Therefore, when a plurality of horizontal edges arranged in the vertical direction are extracted, the shorter the length of any one of the horizontal edges (for example, the reference horizontal edge Line (1)), the condition B in the first embodiment. Dth is variably set according to the length of the horizontal edge so as to reduce the predetermined value Dth.

  Similarly, the shorter the length of any one of the plurality of horizontal edges arranged in the vertical direction (for example, the reference horizontal edge Line (1)), the smaller the predetermined value αth related to the condition C in the second embodiment. Αth is variably set according to the length of the horizontal edge so as to increase.

  By doing in this way, the horizontal edge group determined as the horizontal edge group of the vehicle candidate irrespective of the distance from the own vehicle 100 to the other vehicle or the size of the other vehicle is an image of the actual other vehicle. It is possible to further increase the reliability of the search as corresponding to the above.

  Further, in each of the embodiments described above, the case where another vehicle is detected as an object of a predetermined type in the present invention has been described as an example. However, the present invention can be applied even when an object other than another vehicle is detected.

  In addition, the imaging device 2 is not limited to the imaging device 2 mounted in advance in the vehicle 100, and may be an imaging device that can be carried by the driver.

  DESCRIPTION OF SYMBOLS 1 ... Object detection processing apparatus, 2 ... Imaging device, 5 ... Horizontal edge extraction part (horizontal edge extraction means), 6 ... Object recognition part (object recognition means).

Claims (5)

An object detection processing device that acquires a captured image around the imaging device from the imaging device, and performs processing for detecting an object using the captured image,
Horizontal edge extraction means for extracting a horizontal edge from the captured image;
When a plurality of horizontal edges are extracted by the horizontal edge extracting means, the first condition is constituted by a plurality of horizontal edges arranged in the vertical direction of the captured image, and the captured image of the plurality of horizontal edges A second a condition that the vertical distance between two horizontal edges of each pair adjacent to each other in the vertical direction is less than or equal to a predetermined value, and between the left and right ends of the lowest horizontal edge of the plurality of horizontal edges A horizontal edge group that satisfies the third condition that the horizontal position of the center of the other horizontal edge above the lowest horizontal edge exists within the horizontal range of An object detection processing apparatus comprising: an object recognition means for recognizing. An object detection processing device that acquires a captured image around the imaging device from the imaging device, and performs processing for detecting an object using the captured image,
Horizontal edge extraction means for extracting a horizontal edge from the captured image;
When a plurality of horizontal edges are extracted by the horizontal edge extracting means, the first condition is constituted by a plurality of horizontal edges arranged in the vertical direction of the captured image, and within a predetermined range in the vertical direction of the captured image. The second edge condition that the density of the horizontal edges is greater than or equal to a predetermined value, and the upper side of the lowermost horizontal edge within a lateral range between the left and right ends of the lowermost horizontal edge of the plurality of horizontal edges And object recognition means for recognizing a horizontal edge group satisfying the third condition that a lateral position of the center of another horizontal edge exists as a horizontal edge group of a predetermined type of the same object. Processing equipment. An object detection processing device that acquires a captured image around the imaging device from the imaging device, and performs processing for detecting an object using the captured image,
Horizontal edge extraction means for extracting a horizontal edge from the captured image;
When a plurality of horizontal edges are extracted by the horizontal edge extracting means, the first condition is constituted by a plurality of horizontal edges arranged in the vertical direction of the captured image, and the captured image of the plurality of horizontal edges An object that recognizes a horizontal edge group satisfying the condition 2a that the vertical distance between two horizontal edges of each pair adjacent to each other in the vertical direction is a predetermined type of horizontal edge group of the same object A recognition means ,
The object recognizing unit responds to a length of the at least one horizontal edge so that the predetermined value in the condition 2a is reduced as the length of at least one horizontal edge in the horizontal edge group is shorter. The object detection processing device is configured to variably set the predetermined value . An object detection processing device that acquires a captured image around the imaging device from the imaging device, and performs processing for detecting an object using the captured image,
Horizontal edge extraction means for extracting a horizontal edge from the captured image;
When a plurality of horizontal edges are extracted by the horizontal edge extracting means, the first condition is constituted by a plurality of horizontal edges arranged in the vertical direction of the captured image, and within a predetermined range in the vertical direction of the captured image. Object recognition means for recognizing a horizontal edge group satisfying the second b condition that the density of the horizontal edges is equal to or higher than a predetermined value as a horizontal edge group of the same type of the same object ;
According to the length of the at least one horizontal edge, the object recognizing means increases the predetermined value in the second b condition as the length of at least one horizontal edge in the horizontal edge group is shorter. The object detection processing device is configured to variably set the predetermined value . An object detection processing device that acquires a captured image around the imaging device from the imaging device, and performs processing for detecting an object using the captured image,
Horizontal edge extraction means for extracting a horizontal edge from the captured image;
When a plurality of horizontal edges are extracted by the horizontal edge extracting means, the first condition is constituted by a plurality of horizontal edges arranged in the vertical direction of the captured image, and the captured image of the plurality of horizontal edges An object that recognizes a horizontal edge group satisfying the condition 2a that the vertical distance between two horizontal edges of each pair adjacent to each other in the vertical direction is a predetermined type of horizontal edge group of the same object A recognition means ,
The predetermined type of object is a vehicle;
The object recognizing means increases the predetermined value in the second a condition as the vertical position of the two horizontal edges of each pair adjacent to each other in the vertical direction of the captured image is higher. An object detection processing apparatus configured to variably set the predetermined value in accordance with a vertical position of two horizontal edges of a pair .