JP6426512B2 - Driving lane recognition system - Google Patents
Driving lane recognition system Download PDFInfo
- Publication number
- JP6426512B2 JP6426512B2 JP2015053084A JP2015053084A JP6426512B2 JP 6426512 B2 JP6426512 B2 JP 6426512B2 JP 2015053084 A JP2015053084 A JP 2015053084A JP 2015053084 A JP2015053084 A JP 2015053084A JP 6426512 B2 JP6426512 B2 JP 6426512B2
- Authority
- JP
- Japan
- Prior art keywords
- line
- candidate
- certainty factor
- lane
- division
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000605 extraction Methods 0.000 claims description 21
- 150000001875 compounds Chemical class 0.000 claims description 13
- 238000005192 partition Methods 0.000 claims description 9
- 230000010354 integration Effects 0.000 claims description 7
- 239000000284 extract Substances 0.000 claims description 5
- 238000000034 method Methods 0.000 description 10
- 239000011280 coal tar Substances 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 244000025254 Cannabis sativa Species 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
Landscapes
- Traffic Control Systems (AREA)
- Image Analysis (AREA)
Description
本発明は、車載カメラにより撮影された画像に基づいて、道路の走行区画線を認識する走行区画線認識装置に関する。 The present invention relates to a traveling parting line recognition apparatus that recognizes a traveling parting line of a road based on an image captured by an onboard camera.
従来、車載カメラで撮影した画像に基づいて、道路の車線を区画する走行区画線を認識する装置が提案されている。このような装置では、走行区画線以外を走行区画線と誤認識することが問題となる。そこで、特許文献１に記載の車線識別装置では、車載カメラで撮影した画像を複数の単位ブロックに区分し、単位ブロックごとに、単位ブロック内での輝度や色差の分散値を算出している。そして、分散値が閾値以上の単位ブロックは、画像の複雑度が高く外乱要因が含まれているとして除外し、分散値が閾値未満の単位ブロックから走行区画線を検出している。 2. Description of the Related Art Conventionally, there has been proposed an apparatus that recognizes a traveling division line that divides a lane of a road based on an image captured by an on-vehicle camera. In such an apparatus, it becomes a problem to misrecognize other than a traveling division line as a traveling division line. Therefore, in the lane identification device described in Patent Document 1, the image captured by the on-vehicle camera is divided into a plurality of unit blocks, and the dispersion value of luminance and color difference in the unit block is calculated for each unit block. Then, unit blocks having a variance value equal to or higher than the threshold value are excluded because they have high complexity of the image and include disturbance factors, and a traveling lane is detected from the unit blocks whose variance value is less than the threshold value.
上記車線識別装置では、単位ブロック内において、走行区画線以外の外乱要因が存在しない場合でも、走行区画線がかすれていたりして路面が荒れている場合には、その単位ブロックの分散値が閾値以上になり除外されてしまうおそれがある。そのため、走行区画線を適切に認識できないおそれがある。 In the above-mentioned lane identification device, even if there is no disturbance factor other than the traveling division line in the unit block, if the traveling division line is blurred or the road surface is rough, the variance value of the unit block is the threshold value. There is a risk of being excluded and being over. Therefore, there is a possibility that it can not recognize a run division line appropriately.
本発明は、上記実情に鑑み、走行区画線の認識精度を向上させることが可能な走行区画線認識装置を提供することを主たる目的とする。 In view of the above situation, the present invention has as its main object to provide a traveling parting line recognition device capable of improving the recognition accuracy of traveling parting lines.
請求項１に記載の発明は、上記課題を解決するため、走行区画線認識装置であって、車両に搭載されたカメラにより撮影された画像から、道路の車線を区画する走行区画線の候補である区画線候補を抽出する候補抽出手段と、前記区画線の特徴ごとに、前記候補抽出手段により抽出された前記区画線候補の前記区画線である区画線確信度を算出する第１確信度算出手段と、前記区画線候補が、前記区画線を認識する際に外乱要因となる所定対象物の特徴を有するほど、前記区画線候補の前記区画線である区画線確信度を低く算出する第２確信度算出手段と、前記第１確信度算出手段により算出された前記区画線確信度、及び前記第２確信度算出手段により算出された前記区画線確信度を統合して、統合確信度を算出する統合手段と、前記統合手段により算出された前記統合確信度が第１閾値よりも高い前記区画線候補を前記区画線として認識する区画線認識手段と、を備える。 The invention according to claim 1 is a traveling lane line recognition apparatus for solving the above-mentioned problems, which is a candidate for a traveling lane line which divides a lane of a road from an image photographed by a camera mounted on a vehicle. A candidate extraction unit that extracts a certain partition line candidate, and a first certainty calculation that calculates, for each feature of the partition line, a partition line certainty factor that is the partition line of the partition line candidate extracted by the candidate extraction unit A second calculation unit configured to calculate the division line certainty factor, which is the division line of the division line candidate, lower as the division line candidate has the feature of the predetermined target that is a disturbance factor when recognizing the division line; An integrated certainty factor is calculated by unifying the certainty factor calculating means, the division line certainty factor calculated by the first certainty factor calculating means, and the section line certainty factor calculated by the second certainty factor calculating means Means of integration and said integration And a division line recognizing means for recognizing as the partition lines the integrated confidence that the calculated higher the lane line candidate than the first threshold value by the step.
請求項１に記載の発明によれば、カメラの撮影画像から走行区画線の候補である区画線候補が抽出され、走行区画線の特徴ごとに、区画線候補の確信度が算出される。また、区画線候補が外乱要因となる所定対象物の特徴を有するほど、区画線候補の確信度が低く算出される。すなわち、区画線候補が所定対象物である可能性が高いことが特定されて、区画線候補の確信度が低く算出される。そして、所定対象物の特徴に対して算出された区画線確信度と、走行区画線の特徴ごとに算出された区画線確信度とが統合されて、統合確信度が算出される。さらに、算出された統合確信度が第１閾値よりも高い区画線候補が、区画線として認識される。よって、区画線候補が所定対象物に対応するものである可能性が高い場合には、統合した区画線確信度が低く算出されるため、所定対象物を走行区画線として認識することを抑制できる。したがって、走行区画線の認識精度を向上させることができる。 According to the first aspect of the present invention, a parting line candidate which is a candidate for a traveling parting line is extracted from a photographed image of a camera, and the certainty factor of the parting line candidate is calculated for each feature of the traveling parting line. Further, the certainty factor of the parting line candidate is calculated to be lower as the parting line candidate has the feature of the predetermined object that causes the disturbance. That is, it is identified that the dividing line candidate is likely to be the predetermined target, and the certainty factor of the dividing line candidate is calculated to be low. Then, the dividing line certainty factor calculated for the feature of the predetermined object and the dividing line certainty factor calculated for each feature of the traveling parting line are integrated to calculate the integrated certainty factor. Furthermore, a dividing line candidate whose calculated integrated certainty factor is higher than the first threshold value is recognized as a dividing line. Therefore, when the division line candidate is highly likely to correspond to the predetermined object, the integrated division line certainty factor is calculated to be low, so that recognition of the predetermined object as the traveling division line can be suppressed. . Therefore, the recognition accuracy of the traveling lanes can be improved.
また、請求項２に記載の発明は、走行区画線認識装置であって、車両に搭載されたカメラにより撮影された画像から、道路の車線を区画する走行区画線の候補である区画線候補を抽出する候補抽出手段と、前記区画線の特徴に基づいて、前記候補抽出手段により抽出された前記区画線候補から選択した前記区画線候補を前記区画線として認識する区画線認識手段と、前記区画線を認識する際に外乱要因となる所定対象物の特徴に基づいて、前記候補抽出手段により抽出された前記区画線候補から選択した前記区画線候補を前記所定対象物として認識する対象物認識手段と、を備え、前記区画線認識手段は、前記区画線を認識しなくなった場合に、前記対象物認識手段により前記所定対象物と認識された前記区画線候補を、新たな前記区画線として認識しないことを特徴とする。 The invention described in claim 2 is a traveling lane line recognition apparatus, which is a lane line candidate that is a candidate for a traveling lane line that divides a road lane from an image captured by a camera mounted on a vehicle. A candidate extraction unit for extracting, a division line recognition unit that recognizes, as the division line, the division line candidate selected from the division line candidates extracted by the candidate extraction unit based on the characteristics of the division line; Object recognition means for recognizing the division line candidate selected from the division line candidates extracted by the candidate extraction means as the predetermined object based on the characteristics of the predetermined object which is a disturbance factor when recognizing a line And when the marking line recognition means does not recognize the marking line, the marking line candidate recognized as the predetermined target by the object recognition means is the new marking line. Characterized in that it does not recognize Te.
請求項２に記載の発明によれば、カメラの撮影画像から走行区画線の候補である区画線候補が抽出される。そして、走行区画線の特徴に基づいて、抽出された区画線候補から選択された区画線候補が走行区画線として認識される。また、外乱要因となる所定対象物の特徴に基づいて、抽出された区画線候補から選択された区画線候補が所定対象物として認識される。そして、走行区画線の特徴に基づいて走行区画線が認識されなくなった場合に、所定対象物と認識された区画線候補が、新たな走行区画線として認識されない。したがって、走行区画線の認識精度を向上させることができる。 According to the second aspect of the present invention, division line candidates which are candidates for traveling division lines are extracted from the photographed image of the camera. Then, on the basis of the characteristics of the traveling division lines, the division line candidates selected from the extracted division line candidates are recognized as the traveling division lines. Further, based on the feature of the predetermined object which is a disturbance factor, the dividing line candidate selected from the extracted dividing line candidates is recognized as the predetermined object. Then, when the traveling parting line is not recognized based on the characteristics of the traveling parting line, the parting line candidate recognized as the predetermined object is not recognized as a new traveling parting line. Therefore, the recognition accuracy of the traveling lanes can be improved.
以下、走行区画線認識装置を具現化した各実施形態について、図面を参照しつつ説明する。なお、以下の各実施形態相互において、互いに同一もしくは均等である部分には、図中、同一符号を付しており、同一符号の部分についてはその説明を援用する。 Hereinafter, each embodiment that embodies the traveling lane marking recognition apparatus will be described with reference to the drawings. In addition, in the following each embodiment, the same code | symbol is attached | subjected to the mutually same or equal part in the figure, and the description is used about the part of the same code | symbol.
（第１実施形態）
まず、図１及び２を参照して、本実施形態に係る走行区画線認識装置について説明する。本実施形態に係る走行区画線認識装置は、ＥＣＵ２０により構成され、車載カメラ１０により撮影された前方画像から、走行車線の左右を区画する走行区画線を認識する車載装置である。
First Embodiment
First, with reference to FIGS. 1 and 2, a travel lane line recognition apparatus according to the present embodiment will be described. The travel lane line recognition device according to the present embodiment is an on-vehicle device that is constituted by the ECU 20 and recognizes travel lane lines dividing the left and right of the traveling lane from the front image captured by the on-vehicle camera 10.
車載カメラ１０は、ＣＣＤイメージセンサ、ＣＭＯＳイメージセンサ等の少なくとも１つから構成されている。図１に示すように、車載カメラ１０は、例えば車両５０のフロントガラスの上端付近に搭載されており、車両前方へ向けて所定角度範囲で広がる領域を撮影する。すなわち、車載カメラ１０は、車両５０の前方の道路を含む周辺環境を撮影する。 The on-vehicle camera 10 is configured of at least one of a CCD image sensor, a CMOS image sensor, and the like. As shown in FIG. 1, the on-vehicle camera 10 is mounted, for example, in the vicinity of the upper end of the windshield of the vehicle 50, and captures an area extending in a predetermined angle range toward the front of the vehicle. That is, the on-vehicle camera 10 captures the surrounding environment including the road ahead of the vehicle 50.
走行支援装置４０は、ＥＣＵ２０により認識された走行区画線に基づいて、車線の逸脱を警告する逸脱警報装置や、運転支援を行う運転支援装置である。走行支援装置４０は、逸脱警報装置である場合、ディスプレイ、スピーカ、バイブレータ等のヒューマンマシンインターフェースとして構成され、車両５０が車線を逸脱する際に、運転者に警報を出力する。また、走行支援装置４０は、運転支援装置である場合、操舵アクチュエータや制動アクチュエータとして構成され、車両５０の操舵制御やブレーキ制御を行う。 The driving support device 40 is a deviation warning device that warns of a lane deviation based on the driving division lines recognized by the ECU 20, and a driving support device that performs driving support. When the travel support device 40 is a departure warning device, the travel support device 40 is configured as a display, a speaker, a human machine interface such as a vibrator, and outputs a warning to the driver when the vehicle 50 deviates from the lane. In addition, when the driving support device 40 is a driving support device, the driving support device 40 is configured as a steering actuator or a braking actuator, and performs steering control and brake control of the vehicle 50.
ＥＣＵ２０は、ＣＰＵ、ＲＡＭ、ＲＯＭ、及びＩ／Ｏ等を備えたコンピュータである。ＣＰＵが、ＲＯＭに記憶されている各種プログラムを実行することにより、候補抽出部２１、判定部２２、区画線確信度算出部２３、路側物確信度算出部２４、統合部２８、区画線認識部２９、及び学習部３０の各機能を実現する。 The ECU 20 is a computer provided with a CPU, a RAM, a ROM, an I / O, and the like. When the CPU executes various programs stored in the ROM, the candidate extraction unit 21, the determination unit 22, the dividing line certainty calculating unit 23, the roadside object certainty calculating unit 24, the integrating unit 28, and the dividing line recognition unit 29 and each function of the learning unit 30 are realized.
候補抽出部２１（候補抽出手段）は、車載カメラ１０により撮影された画像から、走行区画線の候補である区画線候補を抽出する。具体的には、候補抽出部２１は、画像を水平方向に走査し、輝度値が大きく上昇するアップエッジ点、及び輝度値が大きく下降するダウンエッジ点を抽出する。候補抽出部２１は、抽出したアップエッジ点及びダウンエッジ点にハフ変換を適用して、アップエッジ点からなるエッジ線、及びダウンエッジ点からなるエッジ線を抽出する。そして、候補抽出部２１は、アップエッジ点からなるエッジ線と、ダウンエッジ点からなるエッジ線を輪郭とする線を、区画線候補として抽出する。 The candidate extraction unit 21 (candidate extraction unit) extracts division line candidates that are candidates for traveling division lines from the image captured by the on-vehicle camera 10. Specifically, the candidate extraction unit 21 scans the image in the horizontal direction, and extracts an up edge point where the luminance value greatly increases and a down edge point where the luminance value largely decreases. The candidate extraction unit 21 applies the Hough transform to the extracted up edge point and down edge point to extract an edge line composed of the up edge point and an edge line composed of the down edge point. Then, the candidate extraction unit 21 extracts an edge line formed by the up edge point and a line whose outline is an edge line formed by the down edge point as division line candidates.
判定部２２（判定手段）は、周知のいずれかの方法により、候補抽出部２１により抽出された区画線候補が複合線か否か判定する。例えば、判定部２２は、車両５０に対して同じ側で抽出された、複数の区画線候補同士の間隔が、予め規定されている複合線のパターンに一致する場合に、その複数の区画線候補を複合線に含まれるものと判定する。 The determination unit 22 (determination unit) determines whether the division line candidate extracted by the candidate extraction unit 21 is a compound line by any known method. For example, when the interval between the plurality of dividing line candidates extracted on the same side with respect to the vehicle 50 matches the pattern of the compound line defined in advance, the determining unit 22 determines the plurality of dividing line candidates Is determined to be included in the compound line.
区画線確信度算出部２３（第１確信度算出手段）は、走行区画線の特徴ごとに、候補抽出部２１により抽出された区画線候補の区画線確信度を算出する。区画線確信度は、区画線候補が走行区画線である確からしさを表す。詳しくは、区画線確信度算出部２３は、走行区画線の特徴ごとに、区画線候補が走行区画線の特徴を備える度合が高いほど、区画線確信度を高く算出する。走行区画線の特徴は、路面に対する走行区画線のコントラスト、走行区画線の太さ、走行区画線の連続する距離等である。 The dividing line certainty calculating unit 23 (first certainty calculating unit) calculates dividing line certainty of the dividing line candidate extracted by the candidate extracting unit 21 for each feature of the traveling dividing line. The lane line certainty factor indicates the likelihood that the lane line candidate is a traveling lane line. In more detail, the dividing line certainty calculating unit 23 calculates the dividing line certainty higher as the degree of the dividing line candidate including the characteristic of the running parting line is higher, for each of the features of the running parting line. The characteristics of the traveling division line are the contrast of the traveling division line with respect to the road surface, the thickness of the traveling division line, the continuous distance of the traveling division line, and the like.
路側物確信度算出部２４（第２確信度算出手段）は、走行区画線を認識する際に外乱要因となる所定対象物の特徴に基づいて、候補抽出部２１により抽出された区画線候補の区画線確信度を算出する。本実施形態では、所定対象物を路側物とする。路側物としては、ガードレール、路側壁、路肩、ポール等が挙げられる。 The roadside object certainty factor calculation unit 24 (second certainty factor calculation means) determines the division line candidate extracted by the candidate extraction unit 21 based on the feature of the predetermined object which is a disturbance factor when recognizing the traveling division line. Calculate the lane line certainty factor. In the present embodiment, the predetermined object is a roadside object. Roadside items include guard rails, road sidewalls, road shoulders, poles, and the like.
図３に示すように、車線に沿って延伸した白色のガードレールは、周囲よりも輝度が高いため、本来の走行区画線と同様に、区画線候補として抽出される。また、図４に示すように、路肩の縁石の表面は、画面上で光って見えるため、本来の走行区画線と同様に、区画線候補として抽出される。図３及び４に示すように、路側物が区画線候補として抽出されているときに、本来の走行区画線がかすれて、区画線候補として抽出されなかった場合、区画線候補として抽出された路側物を、走行区画線として認識するおそれがある。路側物が走行区画線として認識されると、走行支援を行う際に支障が出るおそれがある。 As shown in FIG. 3, since the white guardrail extended along the lane has a higher luminance than the surrounding area, it is extracted as a candidate for a dividing line in the same manner as an original running dividing line. Further, as shown in FIG. 4, the surface of the curb on the road shoulder looks bright on the screen, and therefore, it is extracted as a division line candidate in the same manner as the original travel division line. As shown in FIGS. 3 and 4, when the roadside object is extracted as a dividing line candidate, the original traveling dividing line is faded and not extracted as a dividing line candidate, the roadside extracted as a dividing line candidate An object may be recognized as a running section line. When a roadside object is recognized as a travel division line, there is a possibility that trouble may occur when performing travel support.
そこで、路側物確信度算出部２４は、路側物に対応する区画線候補が走行区画線として認識されないように、区画線候補が路側物の特徴を有するほど、区画線候補の区画線確信度を低く算出する。路側物の特徴としては、画像の色情報のばらつき又は輝度情報のばらつき、画面の幅方向の中央から区画線候補までの距離に対する区画線候補の幅の割合、区画線候補の幅が走行区画線の幅からかけ離れている、車両５０の右側及び左側において抽出された区画線候補が、走行車線の中央に対して非対称である等が挙げられる。以下、各路側物の特徴に対する区画線確信度の算出について説明する。 Therefore, the roadside object certainty factor calculation unit 24 determines the parting line certainty factor of the parting line candidate as the parting line candidate has the feature of the roadside object so that the parting line candidate corresponding to the roadside object is not recognized as the traveling parting line. Calculate low. The characteristics of the roadside object include the variation of color information of the image or the variation of luminance information, the ratio of the width of the dividing line candidate to the distance from the center of the screen in the width direction to the dividing line candidate The lane line candidates extracted on the right and left sides of the vehicle 50, which are far from the width of the vehicle, are asymmetric with respect to the center of the traveling lane, and the like. Hereinafter, calculation of division line certainty factor to the feature of each roadside thing is explained.
路側物の周囲には、植込みや草等があるために、画像上において、路側物に対応する区画線候補の周辺では、色情報のばらつきや輝度情報のばらつきが大きくなる。よって、路側物確信度算出部２４は、画像において、区画線候補の周辺の色情報のばらつき又は輝度情報のばらつきが大きいほど、区画線確信度を低く算出する。色情報のばらつき又は輝度情報のばらつきは、ばらつき算出部２５により算出される。 Since there are plantings, grass, and the like around the roadside object, variations in color information and variations in luminance information become large in the vicinity of the parting line candidate corresponding to the roadside object on the image. Therefore, the roadside object certainty factor calculation unit 24 calculates the parting line certainty factor lower as the variation in the color information or the variation in the brightness information around the parting line candidate increases in the image. The variation in color information or the variation in luminance information is calculated by the variation calculating unit 25.
ばらつき算出部２５（ばらつき算出手段）は、画像において、区画線候補に近接する所定領域Ｒｒ，Ｒｌの色情報のばらつき又は輝度情報のばらつきを算出する。ばらつきは、例えば分散値とすればよい。所定領域Ｒｒ，Ｒｌは、図４及び５に示すように、区画線候補に近接した右側の領域と、左側の領域である。 The variation calculation unit 25 (variation calculation means) calculates variation in color information or variation in luminance information of predetermined regions Rr and Rl close to the dividing line candidate in the image. The variation may be, for example, a variance value. The predetermined regions Rr and Rl are, as shown in FIGS. 4 and 5, a region on the right side adjacent to the division line candidate and a region on the left side.
ただし、複合線には、白色の線及び黄色の線から構成されるものがある。そのため、区画線候補が、複合線に含まれる線に対応するものである場合、ばらつき算出部２５により算出されたばらつきが比較的大きくなることがある。そこで、路側物確信度算出部２４は、判定部２２により、区画線候補が複合線に含まれるものであると判定された場合は、色情報のばらつき又は輝度情報のばらつきを、路側物の特徴としない。すなわち、判定部２２により、区画線候補が複合線に含まれるものであると判定された場合、ばらつき算出部２５は、色情報のばらつき又は輝度情報のばらつきを算出しない。これにより、区画線候補が複合線に対応するものである場合に、区画線候補の区画線確信度を低下させることがない。 However, some composite lines are composed of white lines and yellow lines. Therefore, when the dividing line candidate corresponds to a line included in the compound line, the variation calculated by the variation calculating unit 25 may be relatively large. Therefore, when the roadside object certainty factor calculation unit 24 determines that the dividing line candidate is included in the compound line by the determination unit 22, the variation of the color information or the variation of the brightness information is determined by the feature of the roadside object. And not. That is, when the determining unit 22 determines that the dividing line candidate is included in the compound line, the variation calculating unit 25 does not calculate the variation in color information or the variation in luminance information. As a result, when the parting line candidate corresponds to the compound line, the parting line certainty factor of the parting line candidate is not reduced.
また、ガードレール等の路側物は、路面よりも高い位置にあるため、道路の幅方向において、実際の路側物の位置よりも、車両から遠い位置にある走行区画線に対応する区画線候補として抽出される。例えば、図３に示すように、車両５０の左側にガードレールがある場合、ガードレールは、道路の幅方向において、実際のガードレールの位置よりも左側の位置、すなわち、ガードレールを路面に投影した位置に存在する区画線候補として抽出される。したがって、抽出された区画線候補が路側物に対応するものである場合、実際の路側物は、路面上に抽出された区画線候補よりも、車載カメラ１０に近い位置に存在する。そのため、区画線候補が路側物に対応するものである場合、区画線候補が走行区画線に対応するものである場合よりも、区画線候補の幅Ｗａが太く見える。 In addition, roadside objects such as guardrails are at a higher position than the road surface, so they are extracted as division line candidates corresponding to traveling division lines located farther from the vehicle than the actual roadside objects in the width direction of the road. Be done. For example, as shown in FIG. 3, when there is a guardrail on the left side of the vehicle 50, the guardrail exists at a position on the left side of the actual guardrail in the width direction of the road, ie, a position where the guardrail is projected onto the road surface. It is extracted as a candidate for dividing line. Therefore, when the extracted lane line candidate corresponds to a roadside object, the actual roadside object is closer to the on-vehicle camera 10 than the lane line candidate extracted on the road surface. Therefore, when the parting line candidate corresponds to a roadside object, the width Wa of the parting line candidate looks wider than in the case where the parting line candidate corresponds to the traveling parting line.
よって、路側物確信度算出部２４は、画像の幅方向の中央から区画線候補までの距離に対して、候補幅算出部２６により算出された区画線候補の幅の割合が大きいほど、区画線確信度を低く算出する。候補幅算出部２６（候補幅算出手段）は、区画線候補の幅、すなわちアップエッジ点からなるエッジ線と、ダウンエッジ点からなるエッジ線との間隔を算出する。 Therefore, the roadside object certainty factor calculation unit 24 determines the division line as the ratio of the width of the division line candidate calculated by the candidate width calculation unit 26 to the distance from the center in the width direction of the image to the division line candidate increases. Calculate the degree of certainty low. The candidate width calculation unit 26 (candidate width calculation means) calculates the width of the dividing line candidate, that is, the interval between the edge line consisting of the up edge point and the edge line consisting of the down edge point.
また、区画線候補が路側物に対応するものである場合、区画線候補の幅Ｗａは、区画線の幅からかけ離れていることが多い。よって、路側物確信度算出部２４は、算出された区画線候補の幅と、後述する学習部３０により学習された走行区画線の幅の学習値との偏差が大きいほど、区画線確信度を低く算出する。なお、学習値の代わりに、予め規定されている標準的な走行区画線の幅を用いてもよい。 In addition, when the dividing line candidate corresponds to a roadside object, the width Wa of the dividing line candidate is often far from the width of the dividing line. Therefore, the roadside object certainty factor calculation unit 24 determines the certainty factor as the deviation between the calculated width of the dividing line candidate and the learning value of the width of the traveling dividing line learned by the learning unit 30 described later increases. Calculate low. It should be noted that, instead of the learning value, a standard travel division line width defined in advance may be used.
また、車両５０の右側及び左側で抽出された区画線候補の少なくとも一方が、路側物に対応するものである場合、左右の区画線候補は、走行中の車線の中央に対して非対称となることが多い。よって、路側物確信度算出部２４は、車両５０の右側及び左側において抽出された区画線候補が、走行中の車線の中央に対して非対称である場合に、区画線確信度を第１閾値よりも低い値である第３閾値よりも低く算出する。この場合、片側の区画線候補は走行区画線に対応するものである可能性がある。よって、路側物確信度算出部２４は、後述する統合部２８により区画線確信度を統合した際に、他の区画線確信度が十分に高い場合には統合確信度が第１閾値を超えるように、非対称に関する区画線確信度を下げすぎないようにする。例えば、第３閾値を５０％とし、非対称に関する区画線確信度を５０％よりも低い数十％とする。 In addition, if at least one of the dividing line candidates extracted on the right side and left side of the vehicle 50 corresponds to a roadside object, the left and right dividing line candidates are asymmetric with respect to the center of the running lane. There are many. Therefore, when the lane line candidates extracted on the right and left sides of the vehicle 50 are asymmetrical with respect to the center of the traveling lane, the roadside object certainty factor calculation unit 24 determines the lane line certainty from the first threshold value. It is calculated lower than the third threshold which is also a low value. In this case, the lane line candidate on one side may correspond to the traveling lane line. Therefore, when the roadside item certainty factor calculating unit 24 integrates the dividing line certainty factor by the integration unit 28 described later, the integrated certainty factor exceeds the first threshold when the other parting line certainty factor is sufficiently high. So as not to underestimate the line-line certainty about the asymmetry. For example, the third threshold is set to 50%, and the dividing line certainty factor regarding the asymmetry is set to several tens percent lower than 50%.
ただし、車線の分岐地点や合流地点のように、車線幅が標準的な幅から大きく変化している地点では、左右の走行区画線が車線の中央に対して非対称となる。よって、予め有している車線幅と所定幅との偏差が、所定偏差よりも大きい場合に、車両５０の右側及び左側において抽出された区画線候補について、非対称であることを路側物の特徴としない。車線幅は、走行区画線の認識結果の履歴から推定する。なお、車両５０が、車線幅の情報を含む地図情報及びＧＰＳ受信機を備えている場合は、現在位置の車線幅を地図情報から取得してもよい。 However, at a point where the width of the lane is largely changed from the standard width, such as a branch point of a lane or a junction point, the left and right travel division lines become asymmetric with respect to the center of the lane. Therefore, when the deviation between the lane width and the predetermined width, which is included in advance, is larger than the predetermined deviation, the feature of the roadside is that the division line candidates extracted on the right and left sides of the vehicle 50 are asymmetric. do not do. The vehicle width is estimated from the history of the recognition result of the traveling lanes. When the vehicle 50 includes map information including lane width information and a GPS receiver, the lane width at the current position may be acquired from the map information.
また、保持部２７（保持手段）は、車両５０が所定時間又は所定距離走行する間、路側物確信度算出部２４により算出された区画線確信度を保持する。詳しくは、保持部２７は、路側物確信度算出部２４により算出された区画線確信度を、その後に撮影される複数フレームの画像において、車線の幅方向の同じ位置に抽出された区画線候補の区画線確信度として保持する。保持部２７は、路側物の特徴ごとに算出された区画線確信度をそれぞれ保持してもよいし、路側物の特徴ごとに算出された区画線確信度を統合して保持してもよい。 Further, while the vehicle 50 travels for a predetermined time or a predetermined distance, the holding unit 27 (holding unit) holds the marking line certainty factor calculated by the roadside object certainty factor calculation unit 24. Specifically, the holding unit 27 is a lane line candidate extracted at the same position in the width direction of the lane in the plurality of frames of images taken after that, with the lane line certainty factor calculated by the roadside object certainty factor calculating unit 24. Hold as a certainty for The holding unit 27 may hold division line certainty factors calculated for each of the features of the roadside object, or may integrate and hold the division line certainty factors calculated for each of the features of the roadside object.
統合部２８（統合手段）は、区画線確信度算出部２３により算出された全ての区画線確信度、及び路側物確信度算出部２４により算出された全ての区画線確信度を統合して、統合確信度を算出する。区画線確信度を統合する手法としては、例えば、ベイズ推定がある。さらに、統合部２８は、保持部２７により区画線確信度が保持されている場合には、保持部２７により保持されている区画線確信度も統合して、統合確信度を算出する。これにより、一旦路側物らしい区画線候補が抽出されると、車線の幅方向の同じ位置で抽出された区画線候補は、しばらくの間、継続して統合確信度が比較的低く算出される。 The integration unit 28 (integration means) integrates all the division line certainty degrees calculated by the division line certainty calculation unit 23 and all the division line certainty degrees calculated by the roadside object certainty calculation unit 24, Calculate the integrated certainty factor. For example, Bayesian estimation is an example of a method of integrating parcel line certainty factors. Furthermore, when the dividing line certainty factor is held by the holding unit 27, the integrating unit 28 integrates the dividing line certainty factors held by the holding unit 27 to calculate the integrated certainty factor. Thereby, once the lane line candidates that are likely to be roadside objects are extracted, the lane line candidates extracted at the same position in the width direction of the lane continue to be calculated with relatively low integrated certainty for a while.
区画線認識部２９（区画線認識手段）は、車両５０の右側及び左側のそれぞれにおいて、統合部２８により算出された統合確信度が第１閾値よりも高い区画線候補であって、統合確信度が最も高い区画線候補を、走行区画線として認識する。 The dividing line recognition unit 29 (division line recognition unit) is a division line candidate whose integrated certainty factor calculated by the integrating unit 28 is higher than the first threshold in each of the right side and the left side of the vehicle 50 The division line candidate having the highest value is recognized as a traveling division line.
学習部３０（学習手段）は、区画線認識部２９により認識された走行区画線の幅を学習する。例えば、学習部３０は、記憶されている学習値と、新たに認識された走行区画線の幅とを、加重平均して新しい学習値を算出し、算出した学習値を記憶装置に格納する。 The learning unit 30 (learning unit) learns the width of the traveling lane line recognized by the lane line recognition unit 29. For example, the learning unit 30 calculates a new learning value by performing weighted averaging on the stored learning value and the width of the newly recognized traveling division line, and stores the calculated learning value in the storage device.
次に、本実施形態に係る走行区画線を認識する処理手順について、図６のフローチャートを参照して説明する。本処理手順は、ＥＣＵ２０が、車載カメラ１０により１フレームの画像が撮影された都度、繰り返し実行する。 Next, the processing procedure for recognizing the traveling lane lines according to the present embodiment will be described with reference to the flowchart of FIG. The processing procedure is repeatedly executed by the ECU 20 each time an image of one frame is taken by the on-vehicle camera 10.
まず、車載カメラ１０により撮影された前方画像を取得する（Ｓ１０）。続いて、Ｓ１０で取得した画像から、区画線候補を抽出する（Ｓ１１）。 First, a forward image captured by the on-vehicle camera 10 is acquired (S10). Subsequently, dividing line candidates are extracted from the image acquired in S10 (S11).
続いて、Ｓ１１で抽出した区画線候補の確信度を算出する（Ｓ１２）。詳しくは、走行区画線の特徴ごとに、走行区画線の特徴を有している度合が高いほど、区画線確信度を高く算出する。また、路側物の特徴ごとに、路側物の特徴を有している度合が高いほど、区画線確信度を低く算出する。 Subsequently, the certainty factor of the dividing line candidate extracted in S11 is calculated (S12). Specifically, for each of the features of the travel division line, the higher the degree of having the feature of the travel division line is, the higher the division line certainty factor is calculated. Further, the lane line certainty factor is calculated to be lower as the degree of having the feature of the roadside object is higher for each of the features of the roadside object.
続いて、Ｓ１２で算出した複数の区画線確信度を統合して、統合確信度を算出する（Ｓ１３）。続いて、車両５０の右側及び左側のそれぞれにおいて、Ｓ１３で算出した統合確信度が第１閾値よりも高く、且つ最も統合確信度が高い区画線候補を、走行区画線として認識し、走行区画線パラメータを算出する。走行区画線パラメータは、車線曲率、曲率変化量、車線位置、車線傾き、車線幅等である。そして、算出した走行区画線パラメータを、走行支援装置４０へ出力する。また、認識した走行区画線の幅を算出し、学習値を更新する。以上で本処理を終了する。 Subsequently, the plurality of dividing line certainty factors calculated in S12 are integrated to calculate an integrated certainty factor (S13). Subsequently, on each of the right side and left side of the vehicle 50, the division line candidate having the integrated certainty factor calculated in S13 higher than the first threshold and having the highest integrated certainty factor is recognized as a traveling parting line, Calculate the parameters. The travel division line parameters are a lane curvature, a curvature change amount, a lane position, a lane inclination, a lane width and the like. Then, the calculated travel lane parameter is output to the travel support device 40. Also, the width of the recognized travel lane is calculated, and the learning value is updated. This is the end of the process.
以上説明した第１実施形態によれば、以下の効果を奏する。 According to the first embodiment described above, the following effects can be obtained.
・区画線候補が路側物の特徴を有するほど、区画線候補の確信度が低く算出される。すなわち、区画線候補が所定対象物である可能性が高い場合には、区画線候補の確信度が比較的低く算出される。そして、路側物の特徴ごとに算出された区画線確信度と、走行区画線の特徴ごとに算出された区画線確信度とが統合されて、統合確信度が算出される。さらに、算出された統合確信度が第１閾値よりも高い区画線候補が、走行区画線として認識される。よって、走行区画線候補が路側物に対応するものである可能性が高い場合には、統合確信度が比較的低く算出されるため、路側物を走行区画線として認識することを抑制できる。したがって、走行区画線の認識精度を向上させることができる。 -As the lane line candidate has the feature of the roadside object, the certainty factor of the lane line candidate is calculated to be lower. That is, when the division line candidate is highly likely to be the predetermined target, the certainty factor of the division line candidate is calculated to be relatively low. Then, the dividing line certainty factor calculated for each feature of the roadside object and the dividing line certainty factor calculated for each feature of the traveling parting line are integrated to calculate an integrated certainty factor. Furthermore, a division line candidate whose calculated integrated certainty factor is higher than the first threshold value is recognized as a traveling division line. Therefore, when it is highly possible that the traveling lane line candidate corresponds to the roadside object, the integrated certainty factor is calculated to be relatively low, so that recognition of the roadside object as the traveling lane line can be suppressed. Therefore, the recognition accuracy of the traveling lanes can be improved.
・所定時間又は所定距離の間、路側物の特徴に対して算出された区画線確信度が保持される。そして、路側物の特徴に対して算出された区画線確信度が保持されている場合には、保持されている区画線確信度を用いて、統合確信度が算出される。よって、実際の走行区画線がかすれて、実際の走行区画線に対応する区画線候補が抽出されなくなった場合でも、路側物に対応する区画線候補の統合確信度は継続して比較的低く算出されるため、路側物を新たな走行区画線として認識することを好適に抑制できる。 The dividing line certainty factor calculated for the feature of the roadside object is held for a predetermined time or a predetermined distance. Then, when the dividing line certainty factor calculated for the feature of the roadside object is held, the integrated certainty factor is calculated using the held dividing line certainty factor. Therefore, even when the actual travel division line is blurred and the division line candidate corresponding to the actual travel division line is not extracted, the integrated certainty factor of the division line candidate corresponding to the roadside object is continuously calculated relatively low. As a result, recognition of the roadside object as a new travel division line can be suitably suppressed.
・路側物に近接する所定領域では、画像の色情報又は輝度情報のばらつきが大きくなる。よって、区画線候補に近接する所定領域の色情報又は輝度情報のばらつきが大きいほど、区画線候補の区画線確信度が低く算出されることにより、路側物を走行区画線として認識することを好適に抑制できる。 In a predetermined area close to a roadside object, variation in color information or luminance information of an image becomes large. Therefore, it is preferable to recognize the roadside object as a travel division line by calculating the division line certainty factor of the division line candidate to be lower as the variation in color information or luminance information of the predetermined area closer to the division line candidate is larger. Can be suppressed.
・走行区画線が複合線の場合、走行区画線に近接する所定領域において、画像の色情報又は輝度情報のばらつきが大きくなることがある。よって、区画線候補が複合線に含まれるものと判定された場合には、画像の色情報又は輝度情報のばらつきが、路側物の特徴とされない。これにより、複合線に含まれる区画線候補を認識対象から除外することを防止し、複合線を適切に認識できる。 In the case where the traveling division line is a compound line, variation in color information or luminance information of an image may be large in a predetermined area close to the traveling division line. Therefore, when it is determined that the dividing line candidate is included in the compound line, the variation in the color information or the luminance information of the image is not considered as the feature of the roadside object. Thereby, it is possible to prevent the division line candidate included in the compound line from being excluded from the recognition target, and the compound line can be appropriately recognized.
・路側物は高さがあるため、画像の幅方向において、実際の路側物の位置よりも車両５０から離れた位置の区画線候補として抽出される。そのため、区画線候補が路側物に対応するものである場合、走行区画線に対応する区画線候補であった場合よりも太く見える。よって、画像の幅方向の中央から区画線候補までの距離に対して、区画線候補の幅の割合が大きいほど、区画線候補の区画線確信度が低く算出される。これにより、路側物を走行区画線として認識することを好適に抑制できる。 Since the roadside object has a height, it is extracted as a dividing line candidate at a position farther from the vehicle 50 than the actual position of the roadside object in the width direction of the image. Therefore, when the lane line candidate corresponds to a roadside object, it looks thicker than the lane line candidate corresponding to the traveling lane line. Therefore, the division line certainty factor of the division line candidate is calculated to be lower as the ratio of the width of the division line candidate is larger than the distance from the center of the image in the width direction to the division line candidate. Thereby, it can suppress suitably recognizing a roadside thing as a run division line.
・区画線候補が路側物に対応するものである場合、区画線候補の幅は走行区画線の幅からかけ離れていることが多い。よって、算出された区画線候補の幅と認識された走行区画線の幅の学習値との偏差が大きいほど、区画線候補の区画線確信度が低く算出される。これにより、路側物を走行区画線として認識することを好適に抑制できる。 If the lane line candidate corresponds to a roadside object, the width of the lane line candidate is often far from the width of the traveling lane line. Therefore, the division line certainty factor of the division line candidate is calculated to be lower as the deviation between the calculated division line candidate width and the learning value of the recognized travel division line width is larger. Thereby, it can suppress suitably recognizing a roadside thing as a run division line.
・左右の区画線候補の少なくとも一方が路側物に対応するものである場合、左右の区画線候補は、車線の中央に対して非対称となることが多い。よって、左右の区画線候補が車線の中央に対して非対称である場合に、区画線確信度が第３閾値よりも低く算出される。これにより、路側物を走行区画線として認識することを好適に抑制できる。このとき、車線の中央の位置は走行区画線の認識した履歴情報などから算出する。 If at least one of the left and right lane line candidates corresponds to a roadside object, the left and right lane line candidates are often asymmetrical with respect to the center of the lane. Therefore, when the left and right dividing line candidates are asymmetric with respect to the center of the lane, the dividing line certainty factor is calculated to be lower than the third threshold. Thereby, it can suppress suitably recognizing a roadside thing as a run division line. At this time, the central position of the lane is calculated from the history information or the like of the recognized traveling lanes.
・分岐地点や合流地点などの車線幅が大きく変化する地点では、左右の走行区画線が車線の中央に対して非対称となっている。よって、予め有している車線幅と所定幅との偏差が所定閾値よりも大きい場合には、左右の区画線候補が非対称であることが路側物の特徴とされない。これにより、走行区画線の幅が大きく変わる地点における区画線候補を適切に認識できる。 • At points where the lane width changes significantly, such as junctions and junctions, the left and right travel division lines are asymmetric with respect to the center of the lanes. Therefore, when the deviation between the lane width and the predetermined width held in advance is larger than the predetermined threshold value, it is not considered that the left and right dividing line candidates are asymmetric. Thus, it is possible to appropriately recognize the parting line candidate at the point where the width of the traveling parting line changes significantly.
（第２実施形態）
次に、第２実施形態に係る走行区画線認識装置について、第１実施形態に係る走行区画線認識装置と異なる点を説明する。
Second Embodiment
Next, with respect to the travel lane line recognition apparatus according to the second embodiment, differences from the travel lane line recognition device according to the first embodiment will be described.
第２実施形態に係る走行区画線認識装置では、路側物確信度算出部２４は、抽出された区画線候補が路側物の特徴を有するほど、区画線候補の区画線確信度を低く算出するとともに、区画線候補の路側物確信度を高く算出する。路側物確信度（対象物確信度）は、区画線候補が路側物である確からしさを表す。さらに、路側物確信度算出部２４は、路側物の特徴ごとに算出した路側物確信度を統合して、統合的な路側物確信度を算出する。 In the travel lane line recognition apparatus according to the second embodiment, the roadside object certainty factor calculation unit 24 calculates the lane line certainty factor of the lane line candidate lower as the extracted lane line candidate has the feature of the roadside object. The roadside object certainty factor of the dividing line candidate is calculated to be high. The roadside object certainty factor (object certainty factor) represents the certainty that the lane line candidate is a roadside thing. Furthermore, the roadside object certainty factor calculation unit 24 integrates the roadside object certainty factors calculated for each feature of the roadside thing, and calculates an integrated roadside item certainty factor.
また、第２実施形態に係る走行区画線認識装置は、図２に示すように、路側物認識部３１の機能を備える。路側物認識部３１（対象物認識手段）は、路側物確信度算出部２４により算出された路側物確信度が第２閾値よりも高い区画線候補を、路側物として認識する。 Further, as shown in FIG. 2, the traveling lane marking recognition apparatus according to the second embodiment includes the function of the roadside object recognition unit 31. The roadside object recognition unit 31 (target object recognition means) recognizes a division line candidate having a roadside object certainty factor calculated by the roadside object certainty factor calculation unit 24 higher than a second threshold value as a roadside thing.
また、区画線認識部２９は、走行区画線を認識しなくなった場合、すなわち統合確信度が第１閾値を超える区画線候補がなくなった場合に、路側物認識部３１により路側物と認識された区画線候補を、新たな走行区画線として認識しない。区画線候補の路側物確信度が高いほど、区画線確信度は低い。よって、本来の走行区画線が認識されなくなった場合でも、路側物と認識された区画線候補が走行区画線として認識されることはない。 In addition, when the lane marking recognition unit 29 no longer recognizes the traveling lane marking, that is, when there is no marking line candidate whose integrated certainty factor exceeds the first threshold, the roadside article recognition unit 31 recognizes the roadside article. The lane line candidate is not recognized as a new traveling lane line. The higher the roadside object certainty factor of the lane line candidate, the lower the lane line certainty factor. Therefore, even when the original travel division line is not recognized, the division line candidate recognized as the roadside object is not recognized as the travel division line.
次に、本実施形態に係る走行区画線を認識する処理手順について、図７のフローチャートを参照して説明する。本処理手順は、ＥＣＵ２０が、車載カメラ１０により１フレームの画像が撮影された都度、繰り返し実行する。 Next, the processing procedure for recognizing the traveling lane lines according to the present embodiment will be described with reference to the flowchart of FIG. The processing procedure is repeatedly executed by the ECU 20 each time an image of one frame is taken by the on-vehicle camera 10.
まず、Ｓ２０及びＳ２１において、Ｓ１０及びＳ１１と同様の処理を行う。続いて、Ｓ２１で抽出した区画線候補について、路側物の特徴に基づいて、路側物確信度を算出する（Ｓ２２）。続いて、Ｓ２３及びＳ２４において、Ｓ１２及びＳ１３と同様の処理を行う。 First, in S20 and S21, the same processing as S10 and S11 is performed. Subsequently, the roadside object certainty factor is calculated based on the characteristics of the roadside object for the lane line candidates extracted in S21 (S22). Subsequently, in S23 and S24, the same processing as S12 and S13 is performed.
続いて、Ｓ２２で算出した路側物確信度が第２閾値よりも高い区画線候補を、路側物として認識する（Ｓ２５）。すなわち、Ｓ２１で抽出した区画線候補から、路側物の特徴に基づいて選択した区画線候補を、路側物として認識する。 Subsequently, a lane line candidate whose roadside object certainty factor calculated in S22 is higher than the second threshold value is recognized as a roadside object (S25). That is, from the parting line candidates extracted in S21, the parting line candidate selected based on the feature of the roadside object is recognized as the roadside object.
続いて、Ｓ２４で算出した区画線確信度が第１閾値よりも高い区画線候補を、走行区画線として認識する（Ｓ２６）。すなわち、Ｓ２１で抽出した区画線候補から、走行区画線の特徴及び路側物の特徴に基づいて選択した区画線候補を、走行区画線として認識する。そして、認識した走行区画線に基づいて、走行区画線パラメータを算出し、算出した走行区画線パラメータを、走行支援装置４０へ出力する。また、認識した走行区画線の幅を算出し、学習値を更新する。以上で本処理を終了する。 Subsequently, a division line candidate whose division line certainty factor calculated in S24 is higher than the first threshold value is recognized as a traveling division line (S26). That is, from the parting line candidates extracted in S21, the parting line candidate selected based on the features of the traveling parting line and the features of the roadside object is recognized as the traveling parting line. Then, based on the recognized travel division lines, the travel division line parameters are calculated, and the calculated travel division line parameters are output to the travel support device 40. Also, the width of the recognized travel lane is calculated, and the learning value is updated. This is the end of the process.
以上説明した第２実施形態によれば、以下の効果を奏する。 According to the second embodiment described above, the following effects can be obtained.
・路側物として認識された区画線候補は、統合確信度が比較的低く算出される。そのため、走行区画線の特徴及び路側物の特徴に基づいて走行区画線が認識されなくなった場合に、路側物と認識された区画線候補が、新たな走行区画線として認識されない。よって、実際の走行区画線がかすれて、実際の走行区画線に対応する区画線候補が抽出されない場合でも、路側物に対応する区画線候補が、新たな走行区画線として認識されることがない。したがって、走行区画線の認識精度を向上させることができる。 The lane line candidate recognized as a roadside object has a relatively low integrated certainty factor. Therefore, when the traveling parting line is not recognized based on the characteristics of the traveling parting line and the characteristics of the roadside thing, the parting line candidate recognized as the roadside thing is not recognized as a new traveling parting line. Therefore, even when the actual travel division line is blurred and the division line candidate corresponding to the actual travel division line is not extracted, the division line candidate corresponding to the roadside object is not recognized as a new travel division line. . Therefore, the recognition accuracy of the traveling lanes can be improved.
（他の実施形態）
・路側物確信度算出部２４は、少なくとも１つの路側物の特徴に基づいて、区画線確信度又は路側物確信度を算出すればよい。路側物の特徴は、適宜組み合わせて用いればよい。
(Other embodiments)
The roadside object certainty factor calculation unit 24 may calculate the lane line certainty factor or the roadside object certainty factor based on the feature of at least one roadside thing. The features of the roadside may be used in combination as appropriate.
・第２実施形態において、路側物確信度算出部２４は、路側物の特徴に基づいて、路側物確信度のみを算出し、区画線確信度を算出しなくてもよい。この場合、路側物と認識された区画線候補を、走行区画線として認識する認識対象から除外するようにする。このようにすれば、実際の走行区画線がかすれて、実際の走行区画線に対応する区画線候補が抽出されない場合でも、路側物に対応する区画線候補が、新たな走行区画線として認識されることがない。よって、走行区画線の認識精度を向上させることができる。 In the second embodiment, the roadside object certainty factor calculation unit 24 may calculate only the roadside thing certainty factor based on the characteristics of the roadside thing, and may not calculate the dividing line certainty factor. In this case, the lane line candidate recognized as a roadside object is excluded from the recognition targets recognized as the traveling lane lines. In this way, even if the actual travel division line is blurred and no division line candidate corresponding to the actual travel division line is extracted, the division line candidate corresponding to the roadside object is recognized as a new travel division line. There is no Therefore, the recognition accuracy of the traveling lane lines can be improved.
・走行区画線を認識する際に外乱要因となる所定対象物は、路側物に限らない。所定対称物は、例えば、路面のひび割れ部分に埋められたコールタールや、路面上の影などの、路面上の外乱要因でもよい。コールタールは、画像で光って見えるため、区画線候補として抽出される。また、図３に示すようなパイプ状のガードレールの影が路面上にある場合、パイプの影部分は、路面よりも暗くなる。そのため、平行な２本のパイプの影に挟まれた路面が区画線候補として抽出される。 -The predetermined target object which becomes a disturbance factor when recognizing a traveling division line is not limited to a roadside object. The predetermined symmetrical object may be, for example, coal tar buried in a cracked portion of the road surface or a disturbance on the road surface such as a shadow on the road surface. Coal tar is extracted as a dividing line candidate because it appears to glow in the image. Further, when the shadow of the pipe-shaped guardrail as shown in FIG. 3 is on the road surface, the shadowed portion of the pipe becomes darker than the road surface. Therefore, the road surface sandwiched by the shadows of two parallel pipes is extracted as the dividing line candidate.
路面上の外乱要因も所定対象物とする場合、所定対象物の特徴として、路側物の特徴以外に、走行区画線の直線形状、走行区画線の色情報、走行区画線の輝度情報等を含めるとよい。ただし、所定対象物を路面上の外乱要因のとする場合は、所定対象物は路面から高い位置にないので、画面の幅方向の中央から区画線候補までの距離に対する区画線候補の幅の割合を、所定対象物の特徴としない。 When the disturbance factor on the road surface is also a predetermined object, in addition to the features of the roadside, the straight line shape of the traveling division line, the color information of the traveling division line, the luminance information of the traveling division line, etc. It is good. However, when the predetermined object is a disturbance factor on the road surface, the predetermined object is not at a high position from the road surface, so the ratio of the width of the division line candidate to the distance from the center in the width direction of the screen to the division line candidate Is not a feature of the predetermined object.
１０…車載カメラ、２０…ＥＣＵ、５０…車両。 10: In-vehicle camera, 20: ECU, 50: Vehicle.
Claims (11)
前記区画線の特徴ごとに、前記候補抽出手段により抽出された前記区画線候補の前記区画線である区画線確信度を算出する第１確信度算出手段と、
前記区画線候補が、前記区画線を認識する際に外乱要因となる所定対象物の特徴を有するほど、前記区画線候補の前記区画線である区画線確信度を低く算出する第２確信度算出手段と、
前記第１確信度算出手段により算出された前記区画線確信度、及び前記第２確信度算出手段により算出された前記区画線確信度を統合して、統合確信度を算出する統合手段と、
前記統合手段により算出された前記統合確信度が第１閾値よりも高い前記区画線候補を前記区画線として認識する区画線認識手段と、
を備える走行区画線認識装置。 From an image taken by the vehicle (50) camera (10) mounted on, and candidate extracting means for extracting the lane line candidate that is a candidate of lane marking defining a lane of a road,
A first certainty factor calculating unit that calculates a certainty factor of the dividing line that is the dividing line of the dividing line candidate extracted by the candidate extraction unit for each feature of the dividing line;
Second confidence calculation that calculates the certainty of the parting line candidate of the parting line candidate lower as the parting line candidate has the feature of the predetermined target that is a disturbance factor when recognizing the parting line Means,
Integrating means for integrating the certainty factor by integrating the dividing line certainty factor calculated by the first certainty factor calculating means and the dividing line certainty factor calculated by the second certainty factor calculating means;
Partition line recognition unit that recognizes the partition line candidate having the integrated certainty factor calculated by the integration unit higher than a first threshold value as the partition line;
Road marking recognition device provided with
前記区画線の特徴に基づいて、前記候補抽出手段により抽出された前記区画線候補から選択した前記区画線候補を前記区画線として認識する区画線認識手段と、
前記区画線を認識する際に外乱要因となる所定対象物の特徴に基づいて、前記候補抽出手段により抽出された前記区画線候補から選択した前記区画線候補を前記所定対象物として認識する対象物認識手段と、を備え、
前記区画線認識手段は、前記区画線を認識しなくなった場合に、前記対象物認識手段により前記所定対象物と認識された前記区画線候補を、新たな前記区画線として認識しないことを特徴とする走行区画線認識装置。 From an image taken by the vehicle (50) camera (10) mounted on, and candidate extracting means for extracting the lane line candidate that is a candidate of lane marking defining a lane of a road,
Division line recognition unit that recognizes the division line candidate selected from the division line candidates extracted by the candidate extraction unit as the division line based on the characteristic of the division line;
An object that recognizes the division line candidate selected from the division line candidates extracted by the candidate extraction unit as the predetermined object, based on the characteristics of the predetermined object that is a disturbance factor when recognizing the division line And a recognition means,
The marking line recognition means does not recognize the marking line candidate recognized as the predetermined target by the object recognition means as a new marking line when the marking line is no longer recognized. Road marking line recognition device.
前記区画線候補が前記所定対象物の特徴を有するほど、前記区画線候補の前記区画線である区画線確信度を低く算出するとともに、前記区画線候補の前記所定対象物である対象物確信度を高く算出する第２確信度算出手段と、
前記第１確信度算出手段により算出された前記区画線確信度、及び前記第２確信度算出手段により算出された前記区画線確信度を統合して、統合確信度を算出する統合手段と、
前記区画線認識手段は、前記統合手段により算出された前記統合確信度が第１閾値よりも高い前記区画線候補を、前記区画線として認識し、
前記対象物認識手段は、前記第２確信度算出手段により算出された前記対象物確信度が第２閾値よりも高い前記区画線候補を、前記所定対象物として認識する請求項２に記載の走行区画線認識装置。 A first certainty factor calculating unit that calculates a certainty factor of the dividing line that is the dividing line of the dividing line candidate extracted by the candidate extraction unit for each feature of the dividing line;
As the division line candidate has the feature of the predetermined object, the division line certainty factor which is the division line of the division line candidate is calculated to be lower, and the object certainty factor which is the predetermined object of the division line candidate Second confidence calculation means for calculating
Integrating means for integrating the certainty factor by integrating the dividing line certainty factor calculated by the first certainty factor calculating means and the dividing line certainty factor calculated by the second certainty factor calculating means;
The parting line recognition means recognizes the parting line candidate having the integrated certainty factor calculated by the integration means higher than a first threshold as the parting line.
The travel according to claim 2, wherein the object recognition means recognizes the division line candidate having the object certainty factor calculated by the second certainty factor calculation unit higher than a second threshold as the predetermined object. Lane marking recognition device.
前記統合手段は、前記保持手段により前記区画線確信度が保持されている場合に、前記保持手段により保持されている前記区画線確信度を用いて、前記統合確信度を算出する請求項１又は３に記載の走行区画線認識装置。 A holding unit configured to hold the dividing line certainty factor calculated by the second certainty factor calculation unit for a predetermined time or a predetermined distance;
The integrated means calculates the integrated certainty factor using the parting line certainty factor held by the holding means when the parting line certainty factor is held by the holding means. The travel lane line recognition device according to 3.
前記第２確信度算出手段は、前記ばらつき算出手段により算出された前記ばらつきが大きいほど、前記区画線確信度を低く算出する請求項１、３、５のいずれか１項に記載の走行区画線認識装置。 The image includes variation calculation means for calculating variation in color information or variation in luminance information of a predetermined area adjacent to the division line candidate extracted by the candidate extraction means in the image;
The travel division line according to any one of claims 1, 3 and 5, wherein the second certainty factor calculation means calculates the division line certainty factor lower as the variation calculated by the variation calculation means is larger. Recognition device.
前記第２確信度算出手段は、前記判定手段により前記区画線候補が複合線と判定された場合に、前記色情報のばらつき又は輝度情報のばらつきを前記所定対象物の特徴としない請求項６に記載の走行区画線認識装置。 It has a determination means for determining whether the parting line candidate extracted by the candidate extraction means is a compound line,
The second certainty factor calculation means does not use the variation in color information or the variation in luminance information as the feature of the predetermined object when the determination means determines that the dividing line candidate is a compound line. The traffic lane line recognition device as described.
前記候補抽出手段により抽出された前記区画線候補の幅を算出する候補幅算出手段を備え、
前記第２確信度算出手段は、前記画像の幅方向の中央から前記区画線候補までの距離に対して、前記候補幅算出手段により算出された前記区画線候補の幅の割合が大きいほど、前記区画線確信度を低く算出する請求項１、３、５〜７のいずれか１項に記載の走行区画線認識装置。 The predetermined object is a roadside object of the lane,
A candidate width calculating unit configured to calculate the width of the dividing line candidate extracted by the candidate extracting unit;
The second certainty factor calculating means is configured such that the ratio of the width of the parting line candidate calculated by the candidate width calculating means to the distance from the center in the width direction of the image to the parting line candidate is larger. The travel lane line recognition apparatus according to any one of claims 1 to 5, wherein the lane line certainty factor is calculated to be low.
前記区画線認識手段により認識された前記区画線の幅を学習する学習手段と、を備え、
前記第２確信度算出手段は、前記候補幅算出手段により算出された前記区画線候補の幅と、前記学習手段により学習された前記区画線の幅の学習値との偏差が大きいほど、前記区画線確信度を低く算出する請求項１、３、５〜８のいずれか１項に記載の走行区画線認識装置。 Candidate width calculating means for calculating the width of the dividing line candidate extracted by the candidate extracting means;
And learning means for learning the width of the dividing line recognized by the dividing line recognition means,
The second certainty factor calculating means is configured to calculate the division as the deviation between the width of the dividing line candidate calculated by the candidate width calculating means and the learning value of the width of the dividing line learned by the learning means increases. The travel lane line recognition apparatus according to any one of claims 1, 3, 5 and 8, wherein the line certainty factor is calculated to be low.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2015053084A JP6426512B2 (en) | 2015-03-17 | 2015-03-17 | Driving lane recognition system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2015053084A JP6426512B2 (en) | 2015-03-17 | 2015-03-17 | Driving lane recognition system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| JP2016173711A JP2016173711A (en) | 2016-09-29 |
| JP6426512B2 true JP6426512B2 (en) | 2018-11-21 |
Family
ID=57008868
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| JP2015053084A Active JP6426512B2 (en) | 2015-03-17 | 2015-03-17 | Driving lane recognition system |
Country Status (1)
| Country | Link |
|---|---|
| JP (1) | JP6426512B2 (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11068735B2 (en) | 2017-12-05 | 2021-07-20 | Denso Corporation | Reliability calculation apparatus |
| US10223614B1 (en) * | 2018-09-04 | 2019-03-05 | StradVision, Inc. | Learning method, learning device for detecting lane through classification of lane candidate pixels and testing method, testing device using the same |
| US10311338B1 (en) * | 2018-09-15 | 2019-06-04 | StradVision, Inc. | Learning method, learning device for detecting lanes on the basis of CNN and testing method, testing device using the same |
| JP7417466B2 (en) * | 2020-05-07 | 2024-01-18 | 株式会社トヨタマップマスター | Information processing device, information processing method, and information processing program |
| CN114264310B (en) * | 2020-09-14 | 2024-06-21 | 阿里巴巴集团控股有限公司 | Positioning and navigation method, device, electronic equipment and computer storage medium |
| KR102405472B1 (en) * | 2021-05-25 | 2022-06-08 | 포티투닷 주식회사 | Method and apparatus for determining a semantic lane in cluttered environments |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE602006011789D1 (en) * | 2005-12-27 | 2010-03-04 | Honda Motor Co Ltd | VEHICLE AND STEERING DEVICE FOR VEHICLE |
| JP5007840B2 (en) * | 2009-05-22 | 2012-08-22 | トヨタ自動車株式会社 | Driving assistance device |
| JP2013190842A (en) * | 2012-03-12 | 2013-09-26 | Alpine Electronics Inc | Lane mark detector and lane mark detection method |
| JP5805619B2 (en) * | 2012-12-26 | 2015-11-04 | 株式会社日本自動車部品総合研究所 | Boundary line recognition device |
| JP6087687B2 (en) * | 2013-03-26 | 2017-03-01 | 株式会社日本自動車部品総合研究所 | Driving support device |
| JP5986949B2 (en) * | 2013-04-08 | 2016-09-06 | 株式会社日本自動車部品総合研究所 | Boundary line recognition device |
-
2015
- 2015-03-17 JP JP2015053084A patent/JP6426512B2/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| JP2016173711A (en) | 2016-09-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP6426512B2 (en) | Driving lane recognition system | |
| CN107719233B (en) | Lane departure warning system and method associated therewith | |
| US20180225529A1 (en) | Hazard detection from a camera in a scene with moving shadows | |
| US10891738B2 (en) | Boundary line recognition apparatus and branch road determination apparatus | |
| JP5966965B2 (en) | Lane boundary departure control device and lane boundary departure suppression method | |
| US9697421B2 (en) | Stereoscopic camera apparatus | |
| JP5441549B2 (en) | Road shape recognition device | |
| JP6697522B2 (en) | Marking line recognition device | |
| US11024051B2 (en) | Object detection device | |
| US10870450B2 (en) | Vehicle control apparatus | |
| US20180005073A1 (en) | Road recognition apparatus | |
| JP6449627B2 (en) | Traveling line recognition device | |
| JP2016218539A (en) | Vehicle runway recognition device | |
| JP6408935B2 (en) | Traveling line recognition device | |
| JP7007856B2 (en) | White line recognition device for vehicles | |
| JP2018181093A (en) | Mark line recognition device | |
| US11010618B2 (en) | Apparatus for identifying line marking on road surface | |
| JP2014026519A (en) | On-vehicle lane marker recognition device | |
| US20170124880A1 (en) | Apparatus for recognizing vehicle location | |
| JP2014067320A (en) | Stereo camera device | |
| JP6404748B2 (en) | Traveling line recognition device | |
| JP5452518B2 (en) | Vehicle white line recognition device | |
| WO2019013253A1 (en) | Detection device | |
| JP6453571B2 (en) | 3D object recognition device | |
| KR20130042335A (en) | Method for discerning lane of input image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| A621 | Written request for application examination |
Free format text: JAPANESE INTERMEDIATE CODE: A621 Effective date: 20171106 |
|
| A977 | Report on retrieval |
Free format text: JAPANESE INTERMEDIATE CODE: A971007 Effective date: 20180813 |
|
| A131 | Notification of reasons for refusal |
Free format text: JAPANESE INTERMEDIATE CODE: A131 Effective date: 20180821 |
|
| A521 | Request for written amendment filed |
Free format text: JAPANESE INTERMEDIATE CODE: A523 Effective date: 20180913 |
|
| TRDD | Decision of grant or rejection written | ||
| A01 | Written decision to grant a patent or to grant a registration (utility model) |
Free format text: JAPANESE INTERMEDIATE CODE: A01 Effective date: 20181002 |
|
| A61 | First payment of annual fees (during grant procedure) |
Free format text: JAPANESE INTERMEDIATE CODE: A61 Effective date: 20181025 |
|
| R150 | Certificate of patent or registration of utility model |
Ref document number: 6426512 Country of ref document: JP Free format text: JAPANESE INTERMEDIATE CODE: R150 |
|
| R250 | Receipt of annual fees |
Free format text: JAPANESE INTERMEDIATE CODE: R250 |
|
| R250 | Receipt of annual fees |
Free format text: JAPANESE INTERMEDIATE CODE: R250 |
|
| R250 | Receipt of annual fees |
Free format text: JAPANESE INTERMEDIATE CODE: R250 |