JP2020144727A - Veneer determination system and veneer determination method - Google Patents
Veneer determination system and veneer determination method Download PDFInfo
- Publication number
- JP2020144727A JP2020144727A JP2019042135A JP2019042135A JP2020144727A JP 2020144727 A JP2020144727 A JP 2020144727A JP 2019042135 A JP2019042135 A JP 2019042135A JP 2019042135 A JP2019042135 A JP 2019042135A JP 2020144727 A JP2020144727 A JP 2020144727A
- Authority
- JP
- Japan
- Prior art keywords
- single plate
- determination
- grade
- image
- defect
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 18
- 230000007547 defect Effects 0.000 claims abstract description 98
- 238000013135 deep learning Methods 0.000 claims description 2
- 239000000284 extract Substances 0.000 abstract description 3
- 238000007689 inspection Methods 0.000 description 11
- 238000012545 processing Methods 0.000 description 7
- 239000002131 composite material Substances 0.000 description 6
- 230000002950 deficient Effects 0.000 description 5
- 238000005259 measurement Methods 0.000 description 5
- 238000005286 illumination Methods 0.000 description 4
- 238000001514 detection method Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 239000002023 wood Substances 0.000 description 2
- 235000002233 Penicillium roqueforti Nutrition 0.000 description 1
- 238000003705 background correction Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000005336 cracking Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000002845 discoloration Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Abstract
Description
本発明は、単板の欠点をＡＩを用いて分類でき、単板のグレードをＡＩを用いて判定できる単板判定システムおよび単板の判定方法に関する。 The present invention relates to a single plate determination system and a single plate determination method capable of classifying defects of a single plate using AI and determining the grade of a single plate using AI.
従来、単板を撮影し、撮影した画像を解析することによって、カビによる変色、穴、割れ、死節、抜け節等の欠点を抽出し、単板の品質を検査する技術が知られている。例えば、特許文献１には、検査対象の単板を撮影し、撮影したカラー画像の色分布を利用して欠点を検出する技術が記載されている。また、特許文献２には、検査対象の単板を撮影し、画像の円形度に基づいて節を検出し、検出した節周辺の画素数が閾値を超えた場合に死節を検出する技術が記載されている。 Conventionally, there is known a technique for inspecting the quality of a single plate by extracting defects such as discoloration due to mold, holes, cracks, dead nodes, missing nodes, etc. by photographing a single plate and analyzing the captured image. .. For example, Patent Document 1 describes a technique of photographing a single plate to be inspected and detecting defects by using the color distribution of the photographed color image. Further, Patent Document 2 describes a technique of photographing a single plate to be inspected, detecting nodes based on the circularity of the image, and detecting dead nodes when the number of pixels around the detected nodes exceeds a threshold value. Are listed.
しかし、単板は自然物であるため、節、穴、角欠け、割れ、カビ、腐れ等の多様な欠点を有し、しかも同じ種類の欠点であっても、異なる位置に異なる大きさ、深さで表れる。このような状況において、特許文献１，２の技術によれば、単板の画像から色情報や形状情報を抽出し、これらの情報に基づいて画一的な検査を実施するため、欠点を抽出し損なったり、欠点の種類を誤って判定したりする問題があった。 However, since a single plate is a natural object, it has various defects such as knots, holes, chipped corners, cracks, mold, and rot, and even if the defects are of the same type, they have different sizes and depths at different positions. Appears in. In such a situation, according to the techniques of Patent Documents 1 and 2, color information and shape information are extracted from the image of a single plate, and a uniform inspection is performed based on the information, so that defects are extracted. There was a problem that it failed or the type of defect was erroneously determined.
そこで、本発明の目的は、欠点の抽出精度および欠点の種類の分類精度を上げ、より適切なグレード判定を実施できる単板判定システムおよび単板の判定方法を提供することにある。 Therefore, an object of the present invention is to provide a single plate determination system and a single plate determination method capable of improving the defect extraction accuracy and the defect type classification accuracy and performing more appropriate grade determination.
上記課題を解決するために、本発明の単板判定システムおよび単板の選別方法は、単板を撮影して画像を生成する撮影装置と、画像に基づいて所定の判定パラメータを用いて単板のグレードを判定するグレード判定装置と、を備え、グレード判定装置は、ＡＩを用いて単板のグレードを判定するＡＩ判定手段と、を含み、ＡＩ判定手段は、画像について局所特徴量を算出し、局所特徴量に基づいて欠点領域を生成する欠点領域生成手段と、欠点領域に表された欠点を所定の教師情報により深層学習を施したＡＩを用いて分類する欠点分類手段と、を含み、欠点領域生成手段は、局所特徴量の強度に応じて局所特徴量に対応する特徴点を関連付け、関連付けられた全ての特徴点を含む欠点領域を生成することを特徴とする。 In order to solve the above problems, the single plate determination system and the single plate selection method of the present invention use a photographing device that photographs a single plate to generate an image, and a single plate using a predetermined determination parameter based on the image. The grade determination device includes an AI determination means for determining the grade of a single plate using AI, and the AI determination means calculates a local feature amount for an image. , A defect region generation means for generating a defect region based on a local feature amount, and a defect classification means for classifying defects represented in the defect region using AI subjected to deep learning based on predetermined teacher information. The defect region generation means is characterized in that feature points corresponding to the local feature amount are associated according to the intensity of the local feature amount, and a defect region including all the associated feature points is generated.
撮影装置は、カラー画像および濃淡画像を生成する画像生成手段を含み、グレード判定装置は、カラー画像に基づいて単板のグレードを判定するカラー判定手段と、濃淡画像に基づいて単板のグレードを判定する濃淡判定手段と、所定の判定結果に基づいて単板の総合的なグレードを判定する総合判定手段を含むことが好ましい。 The photographing device includes an image generation means for generating a color image and a shade image, and the grade determination device includes a color determination means for determining the grade of the single plate based on the color image and the grade of the single plate based on the shade image. It is preferable to include a shading determination means for determining and a comprehensive determination means for determining the overall grade of the single plate based on a predetermined determination result.
このとき、教師情報として、単板の欠点を撮影した画像を採用できる。また、判定パラメータは、単板の原木が属するグループ毎に設定されたパラメータを選択できる。 At this time, as the teacher information, an image obtained by capturing the defect of the single plate can be adopted. Further, as the determination parameter, the parameter set for each group to which the single plate log belongs can be selected.
本発明の単板判定システムおよび単板の判定方法によれば、局所特徴量の強度に応じて局所特徴量に対応する特徴点を関連付け、関連付けられた全ての特徴点を含む欠点領域を生成するため、様々な位置・大きさの欠点を柔軟に無駄なく抽出し、ＡＩ判定手段による欠点の分類精度を高め、単板のグレード判定を適切に実施できるという優れた効果を有する。 According to the single plate determination system and the single plate determination method of the present invention, the feature points corresponding to the local feature amount are associated with each other according to the intensity of the local feature amount, and a defect region including all the associated feature points is generated. Therefore, it has an excellent effect that defects of various positions and sizes can be flexibly and efficiently extracted, the accuracy of classification of defects by the AI determination means can be improved, and the grade determination of a single plate can be appropriately performed.
以下、本発明を、単板判定システムおよび単板の判定方法に具体化した一実施形態を図面に基づいて説明する。 Hereinafter, an embodiment in which the present invention is embodied in a single plate determination system and a single plate determination method will be described with reference to the drawings.
図１に示すように、この実施形態の単板判定システム１は、単板Ｂを撮影してライン画像５１（図３（ａ）参照）を生成するカラーラインセンサ型のカメラ１２と、ライン画像５１に基づいて所定の判定パラメータを用いて単板Ｂのグレードを判定するグレード判定装置１１と、単板Ｂをカメラ１２の撮影位置に移動させる前コンベヤ１４ａと、単板Ｂを撮影位置からグレード別に選別する装置（図示なし）に移動させる後コンベヤ１４ｂと、コンベヤ１４ａ，１４ｂに載置された単板Ｂを検知して検知信号を出力する単板検知器１５と、単板検知器１５の検知結果に基づきコンベヤ１４ａ，１４ｂの搬送量に連動して単板Ｂの同期信号を出力するエンコーダ１６から構成されている。 As shown in FIG. 1, the single plate determination system 1 of this embodiment includes a color line sensor type camera 12 that captures the single plate B and generates a line image 51 (see FIG. 3A), and a line image. A grade determination device 11 that determines the grade of the single plate B using a predetermined determination parameter based on 51, a front conveyor 14a that moves the single plate B to the imaging position of the camera 12, and a grade of the single plate B from the imaging position. A rear conveyor 14b that is moved to a separate sorting device (not shown), a single plate detector 15 that detects the single plate B mounted on the conveyors 14a and 14b and outputs a detection signal, and a single plate detector 15. It is composed of an encoder 16 that outputs a synchronization signal of the single plate B in conjunction with the amount of transportation of the conveyors 14a and 14b based on the detection result.
カメラ１２と、単板Ｂの表面を照明する第１照明１３ａと、単板Ｂの裏面を照明する第２照明１３ｂは撮影装置として機能する。このとき、第１照明１３ａから出射した光は、単板Ｂの表面で反射して反射光として撮影され、第２照明１３ｂから出射した光は、単板Ｂの周囲を透過して透過光として撮影される。また、カメラ１２は単板Ｂの幅方向に複数台が並設されており、これらのカメラ１２が撮影したライン画像５１ａ，５１ｂは、グレード判定装置１１の画像処理部２１において、単板検知器１５の検知信号およびエンコーダ１６の同期信号を用いて一のカラー画像５２に合成される。 The camera 12, the first illumination 13a that illuminates the front surface of the single plate B, and the second illumination 13b that illuminates the back surface of the single plate B function as a photographing device. At this time, the light emitted from the first illumination 13a is reflected on the surface of the single plate B and photographed as reflected light, and the light emitted from the second illumination 13b passes around the single plate B and is used as transmitted light. Be photographed. Further, a plurality of cameras 12 are arranged side by side in the width direction of the single plate B, and the line images 51a and 51b captured by these cameras 12 are single plate detectors in the image processing unit 21 of the grade determination device 11. The detection signal of 15 and the synchronization signal of the encoder 16 are combined into one color image 52.
図１，２に示すように、グレード判定装置１１は、撮影装置から入力したライン画像５１から単板Ｂ全体を含む一枚のカラー画像５２を生成する画像処理部２１と、ＡＩを用いて単板のグレードを判定するＡＩ判定部２４と、カラー画像５２に基づいてグレードを判定するカラー判定部２２と、濃淡画像５３に基づいてグレードを判定する濃淡判定部２３と、カラー判定部２２および濃淡判定部２３の判定結果に基づいてグレードを判定する複合判定部２５と、判定部２２〜２５の判定結果に基づいて単板Ｂの総合的なグレードを判定する総合判定部２６と、総合判定部２６による判定結果を出力する表示部２８と、判定部２２〜２６の判定結果を記憶する記憶部２７から構成されている。 As shown in FIGS. 1 and 2, the grade determination device 11 uses an image processing unit 21 that generates a single color image 52 including the entire single plate B from the line image 51 input from the photographing device, and AI. The AI determination unit 24 that determines the grade of the plate, the color determination unit 22 that determines the grade based on the color image 52, the shade determination unit 23 that determines the grade based on the shade image 53, the color determination unit 22 and the shade A composite determination unit 25 that determines the grade based on the determination result of the determination unit 23, an overall determination unit 26 that determines the overall grade of the single plate B based on the determination results of the determination units 22 to 25, and an overall determination unit. It is composed of a display unit 28 that outputs a determination result by 26 and a storage unit 27 that stores the determination results of the determination units 22 to 26.
ＡＩ判定部２４は、カラー画像５２について欠点領域５５を生成する欠点領域生成部３１と、欠点領域に表された欠点をＡＩを用いて分類する欠点分類部３２を含み、欠点の分類毎に単板Ｂのグレードを判定する。欠点の分類に用いるＡＩは、単板Ｂの欠点を撮影した画像等を教師情報とする深層学習が施されていることが好ましい。 The AI determination unit 24 includes a defect region generation unit 31 that generates a defect region 55 for the color image 52, and a defect classification unit 32 that classifies the defects represented in the defect region using AI, and simply classifies each defect. Determine the grade of plate B. It is preferable that the AI used for classifying the defects is deep-learned by using an image obtained by photographing the defects of the single plate B as teacher information.
図３に示すように、画像処理部２１は、ライン画像５１からカラー画像５２を生成し、カラー画像５２から濃淡画像５３を生成する。また、そして、カラー画像５２に含まれる反射光と透過光の差分を抽出し、単板Ｂのみが撮影された領域を検査領域５４として抽出する。カラー画像５２全体ではなく、検査領域５４について欠点領域５５の算出を実施することにより、ＡＩ判定部２４内での計算量を低減することができる。 As shown in FIG. 3, the image processing unit 21 generates a color image 52 from the line image 51 and a shade image 53 from the color image 52. Further, the difference between the reflected light and the transmitted light included in the color image 52 is extracted, and the region in which only the single plate B is photographed is extracted as the inspection region 54. By calculating the defect region 55 for the inspection region 54 instead of the entire color image 52, the amount of calculation in the AI determination unit 24 can be reduced.
図４（ａ）に示すように、欠点領域生成部３１は、カラー画像５２のうち、検査領域５４について局所特徴量を算出する。局所特徴量は、位置情報（特徴点Ｐ）と、その特徴点Ｐにおけるベクトルおよび強度の情報を備え、検査領域５４上に複数箇所表れる。なお、ここで、図４（ｂ）、図５は、図４（ａ）の検査領域５４に含まれる範囲Ａについての部分拡大図である。 As shown in FIG. 4A, the defect region generation unit 31 calculates the local feature amount for the inspection region 54 in the color image 52. The local feature amount includes position information (feature point P) and vector and intensity information at the feature point P, and appears at a plurality of locations on the inspection area 54. Here, FIGS. 4 (b) and 5 are partially enlarged views of the range A included in the inspection area 54 of FIG. 4 (a).
図４（ｂ）に示すように、欠点領域生成部３１は、マーカされた特徴点Ｐを関連付け、関連付けられた全ての特徴点Ｐを含む欠点領域５５を生成する。特徴点Ｐの関連付けは、特徴点Ｐ間の距離および各特徴点Ｐにおける局所特徴量の強さに基づく。具体的には、欠点領域生成部３１は、検査領域５４上の特徴点Ｐ１〜ｎを中心として、各局所特徴量の強度に応じた大きさの中間領域Ｕ１〜ｎを生成する。そして、欠点領域生成部３１は、中間領域Ｕ同士が接触した場合または重複した場合（Ｕ１〜５）に、接触または重複した中間領域Ｕ同士を結合して一の欠点領域５５を生成する（図４（ｂ）の欠点領域５５ａ）。一方、中間領域Ｕ同士が接触しない場合（Ｕ６）には、中間領域Ｕ同士を結合することなく、欠点領域を生成する（図４（ｂ）の欠点領域５５ｂ）。 As shown in FIG. 4B, the defect region generation unit 31 associates the marked feature points P and generates a defect region 55 including all the associated feature points P. The association of feature points P is based on the distance between feature points P and the strength of local features at each feature point P. Specifically, the defect region generation unit 31 generates intermediate regions U 1 to n having a size corresponding to the intensity of each local feature amount, centering on the feature points P 1 to n on the inspection region 54. Then, when the intermediate regions U are in contact with each other or overlap (U 1 to 5 ), the defective region generation unit 31 combines the contacted or overlapping intermediate regions U with each other to generate one defective region 55 (). The defect area 55a in FIG. 4B). On the other hand, when the intermediate regions U do not come into contact with each other (U 6 ), a defective region is generated without connecting the intermediate regions U (defect region 55b in FIG. 4B).
つまり、特徴点Ｐ１〜ｎにおいて、点間距離が短く、これらの特徴点Ｐにおける局所特徴量が強いほど、特徴点Ｐ同士が関連付けられ易くなり、欠点領域５５ａのように大きく生成される。一方、点間距離が長く、これらの特徴点Ｐにおける局所特徴量が小さいほど、特徴点Ｐ同士は関連付けられ難くなり、欠点領域５５ｂのように小さく生成される。一般に、欠点領域５５は、一つの単板Ｂについて複数生成される。また、局所特徴量は、木目に反応し難い性質を有するため、正常な木目は、欠点として抽出され難くなる。 That is, the feature point P 1 ~ n, the distance between the points is short, the stronger the local features in these feature points P, easily associated feature point P between greater are generated as in disadvantages regions 55a. On the other hand, the longer the distance between points and the smaller the local feature amount at these feature points P, the more difficult it is for the feature points P to be associated with each other, and the smaller the feature points P are generated as in the defect region 55b. In general, a plurality of defect regions 55 are generated for one single plate B. In addition, since the local feature amount has a property of being difficult to react with the grain of wood, it is difficult to extract a normal grain of wood as a defect.
図５（ａ）に示すように、欠点分類部３２は、欠点領域５５を入力し、欠点領域５５に表された欠点をＡＩを用いて分類する。ＡＩは事前に教師情報を用いて深層学習が施され、欠点を抽象化して記憶しているため、欠点領域５５に表れた情報を適切に取捨選択しつつ欠点の分類を実施する。図５（ａ）の例では、欠点領域５５ａは「死節」、欠点領域５５ｂは「小さい節」と分類されている。 As shown in FIG. 5A, the defect classification unit 32 inputs the defect area 55 and classifies the defects represented in the defect area 55 using AI. Since AI is deep-learned using teacher information in advance and abstracts and memorizes defects, the defects appearing in the defect area 55 are appropriately selected and the defects are classified. In the example of FIG. 5A, the defective region 55a is classified as a “dead node” and the defective region 55b is classified as a “small node”.
図５（ｂ）に示すように、ＡＩ判定部２４は、欠点分類部３２による分類に従って欠点領域５５に含まれる欠点の範囲を測定し、測定結果を判定パラメータに基づいて評価し、単板のグレードを判定する。測定内容は、例えば、横幅ｄ１や縦幅ｄ２、または欠点の面積等が挙げられる。 As shown in FIG. 5B, the AI determination unit 24 measures the range of defects included in the defect area 55 according to the classification by the defect classification unit 32, evaluates the measurement result based on the determination parameters, and evaluates the single plate. Determine the grade. The measurement contents include, for example, the width d 1 and the height d 2 , the area of the defect, and the like.
ここで、判定パラメータは、判定部２２〜２６それぞれにおいて、欠点の種類毎に設けられている。例えば、ＡＩの判定項目としては、「節判定」、「死節判定」、「穴判定」、「ヤニツボ判定」等が存在し、判定項目毎に判定パラメータが設けられている。また、判定パラメータは、単板Ｂの原木が属するグループ毎に設定されており、原木の種類に応じた細やかなグレード判定を実施することが可能である。 Here, the determination parameters are provided for each type of defect in each of the determination units 22 to 26. For example, as the determination items of AI, there are "section determination", "dead section determination", "hole determination", "yarn pot determination" and the like, and determination parameters are provided for each determination item. Further, the determination parameters are set for each group to which the log of the single plate B belongs, and it is possible to carry out detailed grade determination according to the type of log.
カラー判定部２２は、カラー画像からＨＳＶ値またはＲＧＢ値を等の色情報を取得し、色情報等を所定の閾値と比較して欠点の分類を実施する。また、欠点の種類毎に欠点の横幅、縦幅、面積等の所定の項目について測定を実施する。そして、色情報等および測定値を、判定パラメータに基づいて評価し、単板Ｂのグレードを判定する。カラー判定部２２では、特に、「青カビ」、「腐れ・皮」、「角欠け」、「穴」、「貫通穴数」等の欠点についてグレード判定を実施する。 The color determination unit 22 acquires color information such as an HSV value or an RGB value from a color image, compares the color information or the like with a predetermined threshold value, and classifies defects. In addition, measurement is performed for predetermined items such as the width, height, and area of the defect for each type of defect. Then, the color information and the measured values are evaluated based on the determination parameters, and the grade of the single plate B is determined. In particular, the color determination unit 22 performs grade determination on defects such as "blue mold", "rot / skin", "corner chipping", "hole", and "number of through holes".
濃淡判定部２３は、濃淡画像から濃淡のエッジ情報等を取得し、エッジ情報等を所定の閾値と比較して欠点の分類を実施する。また、欠点の種類毎に欠点の横幅、縦幅、面積、射影幅のＸＹ比等の所定の項目について測定を実施する。そして、勾配情報等および測定値を、判定パラメータに基づいて評価し、単板Ｂのグレードを判定する。濃淡判定部２３では、特に、「単板Ｂの長さ」、「幅」、「割れ」、「節穴数」、「凹み穴数」等の欠点についてグレード判定を実施する。 The shading determination unit 23 acquires the shading edge information and the like from the shading image, compares the edge information and the like with a predetermined threshold value, and classifies the defects. In addition, measurement is performed for predetermined items such as the width, height, area, and XY ratio of the projection width for each type of defect. Then, the gradient information and the measured values are evaluated based on the determination parameters, and the grade of the single plate B is determined. The shading determination unit 23 particularly determines the grade of defects such as "length of single plate B", "width", "cracking", "number of knot holes", and "number of recessed holes".
複合判定部２５は、カラー判定部２２のグレード判定結果と、濃淡判定部２３のグレード判定結果に基づいて、判定パラメータに基づいて評価を実施し、単板Ｂのグレードを判定する。複合判定部２５では、特に、カラー判定部２２と濃淡判定部２３の共通する判定項目である、例えば、「節穴数および節数」等の判定結果に基づいてグレード判定を実施する。 The composite determination unit 25 performs evaluation based on the determination parameters based on the grade determination result of the color determination unit 22 and the grade determination result of the shade determination unit 23, and determines the grade of the single plate B. In particular, the composite determination unit 25 performs grade determination based on determination results such as “number of knot holes and number of nodes”, which are determination items common to the color determination unit 22 and the shade determination unit 23.
総合判定部２６は、判定部２２〜２５のグレード判定結果に基づいて、判定パラメータに基づいて評価を実施し、単板Ｂのグレードを判定する。 The comprehensive determination unit 26 performs evaluation based on the determination parameters based on the grade determination results of the determination units 22 to 25, and determines the grade of the single plate B.
次に、上記構成の単板判定システム１の動作として表れる単板の判定方法について図６に基づいて説明する。まず、単板Ｂが前コンベヤ１４ａにより撮影位置に配置されると、カメラ１２が単板Ｂを撮影し、ライン画像５１を取得する（Ｓ１）。カメラ１２は、エンコーダ１６からの同期信号に合わせてライン画像５１にシェーディング補正を施し、撮影装置の記憶部（図示なし）に保管する。その後、単板Ｂは、後コンベヤ１４ｂによって撮影位置から単板をグレード別に選別する装置に搬送される。 Next, a method for determining a single plate that appears as an operation of the single plate determination system 1 having the above configuration will be described with reference to FIG. First, when the single plate B is arranged at the photographing position by the front conveyor 14a, the camera 12 photographs the single plate B and acquires the line image 51 (S1). The camera 12 applies shading correction to the line image 51 in accordance with the synchronization signal from the encoder 16, and stores the line image 51 in a storage unit (not shown) of the photographing apparatus. After that, the single plate B is conveyed by the rear conveyor 14b to an apparatus that sorts the single plates by grade from the photographing position.
グレード判定装置１１の画像処理部２１は、撮影装置の記憶部に保存されたライン画像５１を読み出し、複数台のカメラ１２が撮影した複数のライン画像５１（５１ａ，５１ｂ）にキャリブレーションを施して位置座標を合わせ込み、これらのライン画像５１を合成して一枚のカラー画像５２を生成する（Ｓ２）。 The image processing unit 21 of the grade determination device 11 reads out the line image 51 stored in the storage unit of the photographing device, and calibrates the plurality of line images 51 (51a, 51b) photographed by the plurality of cameras 12. The position coordinates are adjusted, and these line images 51 are combined to generate one color image 52 (S2).
ＡＩ判定部２４は、カラー画像５２に基づいて単板Ｂが撮影された領域を抽出し、検査領域５４として設定する（Ｓ６）。その後、ＡＩ判定部２４の欠点領域生成部３１は、検査領域５４について局所特徴量を算出し、各特徴点Ｐが示す局所特徴量の強度に応じて各特徴点Ｐを関連付け、関連付けられた全ての特徴点Ｐを含む領域を欠点領域５５として生成する（Ｓ７）。 The AI determination unit 24 extracts the area where the single plate B is photographed based on the color image 52 and sets it as the inspection area 54 (S6). After that, the defect region generation unit 31 of the AI determination unit 24 calculates the local feature amount for the inspection area 54, associates each feature point P with the intensity of the local feature amount indicated by each feature point P, and associates all the associated features. The region including the feature point P of is generated as the defect region 55 (S7).
欠点領域５５が生成されると、欠点分類部３２は、欠点領域５５に表された欠点をＡＩを用いて分類する（Ｓ８）。 When the defect region 55 is generated, the defect classification unit 32 classifies the defects represented in the defect region 55 by using AI (S8).
欠点の分類が終わると、ＡＩ判定部２４は、欠点の分類に基づいて欠点領域５５に含まれる欠点を測定し、測定値を判定パラメータに基づいて評価し、単板Ｂのグレードを判定する（Ｓ９）。ＡＩ判定部２４によるグレード判定処理が終了すると、判定結果は記憶部２７に格納される。 When the classification of the defects is completed, the AI determination unit 24 measures the defects included in the defect region 55 based on the classification of the defects, evaluates the measured values based on the determination parameters, and determines the grade of the single plate B ( S9). When the grade determination process by the AI determination unit 24 is completed, the determination result is stored in the storage unit 27.
一方、カラー判定部２２は、カラー画像５２を判定パラメータに基づいて評価し、単板Ｂのグレードを判定する（Ｓ３）。カラー判定部２２によるグレード判定が終了すると、判定結果は記憶部２７に格納される。 On the other hand, the color determination unit 22 evaluates the color image 52 based on the determination parameters and determines the grade of the single plate B (S3). When the grade determination by the color determination unit 22 is completed, the determination result is stored in the storage unit 27.
このとき、画像処理部２１は、カラー画像５２に基づいて濃淡画像５３も生成する（Ｓ２）。濃淡画像５３が生成されると、濃淡判定部２３は、濃淡画像５３を判定パラメータに基づいて評価し、単板Ｂのグレードを判定する（Ｓ４）。カラー判定部２２によるグレード判定が終了すると、判定結果は記憶部２７に格納される。 At this time, the image processing unit 21 also generates a shade image 53 based on the color image 52 (S2). When the shade image 53 is generated, the shade determination unit 23 evaluates the shade image 53 based on the determination parameters and determines the grade of the single plate B (S4). When the grade determination by the color determination unit 22 is completed, the determination result is stored in the storage unit 27.
複合判定部２５は、カラー判定部２２および濃淡判定部２３の判定結果を記憶部２７から読み出し、判定パラメータに基づいてさらに評価し、複合的にグレードを判定する（Ｓ５）。複合的なグレード判定が終了すると、判定結果は記憶部２７に格納される。 The composite determination unit 25 reads the determination results of the color determination unit 22 and the shade determination unit 23 from the storage unit 27, further evaluates them based on the determination parameters, and determines the grade in a complex manner (S5). When the combined grade determination is completed, the determination result is stored in the storage unit 27.
総合判定部２６は、ＡＩ判定部２４の判定結果と、カラー判定部２２の判定結果と、濃淡判定部２３の判定結果と、複合判定部２５の判定結果を記憶部２７から読み出し、読み出した判定結果を判定パラメータに基づいて評価し、単板Ｂの総合的なグレードを判定する（Ｓ１０）。 The comprehensive determination unit 26 reads out the determination result of the AI determination unit 24, the determination result of the color determination unit 22, the determination result of the shading determination unit 23, and the determination result of the composite determination unit 25 from the storage unit 27, and the determination is read out. The result is evaluated based on the determination parameter, and the overall grade of the single plate B is determined (S10).
最後に、グレード判定装置１１は、結果表示用の画像を生成し、表示部２８に判定結果を表示する（Ｓ１１）。この後、グレード判定装置１１は単板をグレード別に選別する装置に制御信号を送信し、処理は該装置に引き継がれる。 Finally, the grade determination device 11 generates an image for displaying the result, and displays the determination result on the display unit 28 (S11). After that, the grade determination device 11 transmits a control signal to a device that sorts single plates by grade, and the processing is taken over by the device.
以上の構成の単板判定システム１および単板の判定方法によれば、欠点領域生成部３１が、局所特徴量の強度に応じて特徴点Ｐを関連付けるため、単板Ｂに表れた欠点の位置および大きさを正確に抽出できるという優れた効果を有する。また、欠点領域５５を正確に抽出できた結果、ＡＩが欠点を分類し、グレードを判定する際に、より正確な判定を実施できるという効果も有する。さらに、カラー画像に基づくカラー判定、濃淡画像に基づく濃淡判定、カラー判定および濃淡判定結果を合わせた複合判定をＡＩによる判定と組み合わせることにより、各判定部の弱点を補うことができるという優れた効果を有する。 According to the single plate determination system 1 and the single plate determination method having the above configuration, since the defect region generation unit 31 associates the feature points P according to the intensity of the local feature amount, the position of the defect appearing on the single plate B And it has an excellent effect that the size can be extracted accurately. Further, as a result of being able to accurately extract the defect region 55, there is also an effect that the AI can classify the defects and perform a more accurate determination when determining the grade. Further, by combining the color judgment based on the color image, the shading judgment based on the shading image, the combined judgment including the color judgment and the shading judgment result with the judgment by AI, the weak point of each judgment unit can be compensated. Have.
その他、本発明は、上記実施形態に限定されるものではなく、発明の趣旨を逸脱しない範囲で、各部の構成を任意に変更して実施することも可能である。例えば、判定部２２〜２６を同じ装置内に設けることも、別体に設けることも可能であり、さらに、別体に設けた各判定部２２〜２６について、各々、ＣＰＵ、画像処理部、記憶部、表示部を設けることも可能である。 In addition, the present invention is not limited to the above-described embodiment, and the configuration of each part can be arbitrarily changed and implemented without departing from the spirit of the invention. For example, the determination units 22 to 26 can be provided in the same device or in a separate body, and each of the determination units 22 to 26 provided in the separate body can be provided with a CPU, an image processing unit, and a storage unit, respectively. It is also possible to provide a unit and a display unit.
１ 単板判定システム
１１ グレード判定装置
１２ カメラ
１３ 照明
１４ コンベヤ
１５ 単板検知器
１６ エンコーダ
２１ 画像生成部
２２ カラー判定部
２３ 濃淡判定部
２４ ＡＩ判定部
２５ 複合判定部
２６ 総合判定部
２７ 記憶部
２８ 表示部
２９ ＣＰＵ
３１ 欠点領域生成部
３２ 欠点分類部
５１ ライン画像
５２ カラー画像
５３ 濃淡画像
５４ 検査領域
５５ 欠点領域
Ａ 範囲
Ｂ 単板
Ｐ 特徴点
Ｕ 中間領域
1 Single plate judgment system 11 Grade judgment device 12 Camera 13 Lighting 14 Conveyor 15 Single plate detector 16 Encoder 21 Image generation unit 22 Color judgment unit 23 Darkness judgment unit 24 AI judgment unit 25 Composite judgment unit 26 Comprehensive judgment unit 27 Storage unit 28 Display 29 CPU
31 Defect area generation unit 32 Defect classification unit 51 Line image 52 Color image 53 Light and shade image 54 Inspection area 55 Defect area A Range B Single plate P Feature point U Intermediate area
Claims (8)
前記グレード判定装置は、ＡＩを用いて単板のグレードを判定するＡＩ判定手段と、を含み、
前記ＡＩ判定手段は、前記画像について局所特徴量を算出し、局所特徴量に基づいて欠点領域を生成する欠点領域生成手段と、欠点領域に表された欠点を所定の教師情報により深層学習を施したＡＩを用いて分類する欠点分類手段と、を含み、
前記欠点領域生成手段は、前記局所特徴量の強度に応じて前記局所特徴量に対応する特徴点を関連付け、関連付けられた全ての特徴点を含む欠点領域を生成することを特徴とする単板判定システム。 It is provided with a photographing device that photographs a single plate and generates an image, and a grade determination device that determines the grade of the single plate using a predetermined determination parameter based on the image.
The grade determination device includes an AI determination means for determining the grade of a single plate using AI.
The AI determination means calculates a local feature amount for the image and performs deep learning on the defect area generation means that generates a defect area based on the local feature amount and the defect represented by the defect area by predetermined teacher information. Including defect classification means for classifying using the AI
The single plate determination means that the defect region generation means associates feature points corresponding to the local feature amount according to the intensity of the local feature amount, and generates a defect region including all the associated feature points. system.
前記グレード判定装置は、前記カラー画像に基づいて単板のグレードを判定するカラー判定手段と、前記濃淡画像に基づいて単板のグレードを判定する濃淡判定手段と、所定の判定結果に基づいて単板の総合的なグレードを判定する総合判定手段と、を含む請求項１に記載の単板の判定システム。 The photographing apparatus includes an image generation means for generating a color image and a light and shade image.
The grade determination device includes a color determination means for determining the grade of a single plate based on the color image, a shade determination means for determining the grade of the single plate based on the shade image, and a single plate based on a predetermined determination result. The single plate determination system according to claim 1, further comprising a comprehensive determination means for determining the overall grade of the plate.
前記単板のグレードを判定する工程が、ＡＩを用いて単板のグレードを判定する工程を含み、
ＡＩを用いて単板のグレードを判定する工程が、前記画像について局所特徴量を算出し、局所特徴量に基づいて欠点領域を生成する工程と、前記欠点領域に表された欠点を所定の教師情報により深層学習を施したＡＩを用いて分類する工程と、を含み、
前記欠点領域を生成する工程において、前記局所特徴量の強度に応じて前記局所特徴量に対応する特徴点を関連付け、関連付けられた全ての特徴点を含む欠点領域を生成することを特徴とする単板の判定方法。 It includes a step of photographing a single plate and generating an image, and a step of determining the grade of the single plate using a predetermined determination parameter based on the image.
The step of determining the grade of the single plate includes a step of determining the grade of the single plate using AI.
The step of determining the grade of the single plate using AI is the step of calculating the local feature amount for the image and generating the defect area based on the local feature amount, and the step of generating the defect area based on the local feature amount, and the process of determining the defect expressed in the defect area by a predetermined teacher. Including the process of classifying using AI that has been deep-learned by information.
In the step of generating the defect region, the feature points corresponding to the local feature amount are associated with each other according to the intensity of the local feature amount, and a defect region including all the associated feature points is generated. How to judge the board.
前記単板のグレードを判定する工程が、前記カラー画像に基づいて単板のグレードを判定する工程と、前記濃淡画像に基づいて単板のグレードを判定する工程と、所定の判定結果に基づいて単板の総合的なグレードを判定する工程と、を含む請求項５に記載の単板の判定方法。 The step of photographing the single plate and generating an image includes a step of generating a color image and a shade image.
The step of determining the grade of the single plate is based on the step of determining the grade of the single plate based on the color image, the step of determining the grade of the single plate based on the shading image, and the predetermined determination result. The method for determining a single plate according to claim 5, further comprising a step of determining the overall grade of the single plate.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2019042135A JP7318907B2 (en) | 2019-03-08 | 2019-03-08 | Veneer Judgment System and Veneer Judgment Method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2019042135A JP7318907B2 (en) | 2019-03-08 | 2019-03-08 | Veneer Judgment System and Veneer Judgment Method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| JP2020144727A true JP2020144727A (en) | 2020-09-10 |
| JP7318907B2 JP7318907B2 (en) | 2023-08-01 |
Family
ID=72354341
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| JP2019042135A Active JP7318907B2 (en) | 2019-03-08 | 2019-03-08 | Veneer Judgment System and Veneer Judgment Method |
Country Status (1)
| Country | Link |
|---|---|
| JP (1) | JP7318907B2 (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2009074967A (en) * | 2007-09-21 | 2009-04-09 | Meinan Mach Works Inc | Method and apparatus for inspecting wood and inspection program of wood |
| JP2014215233A (en) * | 2013-04-26 | 2014-11-17 | 橋本電機工業株式会社 | Device for sorting and stacking and method of sorting and stacking veneer |
| JP2016110290A (en) * | 2014-12-03 | 2016-06-20 | 株式会社リコー | Information processing apparatus and information processing method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP5520006B2 (en) | 2009-10-29 | 2014-06-11 | パナソニック株式会社 | Wood defect detection apparatus and method |
-
2019
- 2019-03-08 JP JP2019042135A patent/JP7318907B2/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2009074967A (en) * | 2007-09-21 | 2009-04-09 | Meinan Mach Works Inc | Method and apparatus for inspecting wood and inspection program of wood |
| JP2014215233A (en) * | 2013-04-26 | 2014-11-17 | 橋本電機工業株式会社 | Device for sorting and stacking and method of sorting and stacking veneer |
| JP2016110290A (en) * | 2014-12-03 | 2016-06-20 | 株式会社リコー | Information processing apparatus and information processing method |
Non-Patent Citations (1)
| Title |
|---|
| LI,MENG-XIN,ET AL.: "Automatic visual inspection and classification based on rough sets and neural network", PROCEEDINGS OF THE 2003 INTERNATIONAL CONFERENCE ON MACHINE LEARNING AND CYBERNETICS (IEEE CAT. NO.0, JPN6023008201, 5 November 2003 (2003-11-05), US, pages 3095 - 3099, XP010682271, ISSN: 0005005459, DOI: 10.1109/ICMLC.2003.1260110 * |
Also Published As
| Publication number | Publication date |
|---|---|
| JP7318907B2 (en) | 2023-08-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR101298957B1 (en) | Wood knot detecting method, device, and program | |
| JP7391173B2 (en) | Food inspection aid system, food inspection aid device, and computer program | |
| JP5713376B1 (en) | Fish species discrimination device and fish species discrimination method | |
| KR101415628B1 (en) | Lumber inspection method, device and program | |
| CN103072718B (en) | Machine vision detection system and method for food packaging production line | |
| JP5085860B2 (en) | Wood inspection method, apparatus and program | |
| KR101762165B1 (en) | Appearance inspection device | |
| JP6357617B2 (en) | Defect inspection equipment | |
| CN111239142A (en) | Paste appearance defect detection device and method | |
| JP2009293999A (en) | Wood defect detector | |
| JP2018179698A (en) | Sheet inspection device | |
| JP2015114226A (en) | Visual inspection device and visual inspection method | |
| JP4708904B2 (en) | Wood inspection method, apparatus and program | |
| JP5006147B2 (en) | Wood inspection method, apparatus and program | |
| JP6699694B2 (en) | Inspection system, inspection method | |
| JP2012108026A (en) | Detection method of foreign material-contaminated product | |
| JP7318907B2 (en) | Veneer Judgment System and Veneer Judgment Method | |
| KR20230139166A (en) | Inspection Method for Wood Product | |
| TWI753424B (en) | Appearance inspection management system, appearance inspection management device, appearance inspection management method, and program | |
| TWI767229B (en) | Appearance inspection management system, appearance inspection management device, appearance inspection management method, and program | |
| CN114813761A (en) | Double-light-stroboscopic-based film pinhole and bright spot defect identification system and method | |
| TWI510776B (en) | Bubble inspection processing method for glass | |
| KR20090121612A (en) | Method and apparatus for automatic detecting quality defects for corrugated board | |
| CN212180649U (en) | Paste appearance defect detection equipment | |
| KR102094314B1 (en) | Machine vision based crop selection apparatus |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| A621 | Written request for application examination |
Free format text: JAPANESE INTERMEDIATE CODE: A621 Effective date: 20220209 |
|
| A977 | Report on retrieval |
Free format text: JAPANESE INTERMEDIATE CODE: A971007 Effective date: 20230222 |
|
| A131 | Notification of reasons for refusal |
Free format text: JAPANESE INTERMEDIATE CODE: A131 Effective date: 20230307 |
|
| A521 | Request for written amendment filed |
Free format text: JAPANESE INTERMEDIATE CODE: A523 Effective date: 20230508 |
|
| 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: 20230627 |
|
| A61 | First payment of annual fees (during grant procedure) |
Free format text: JAPANESE INTERMEDIATE CODE: A61 Effective date: 20230712 |
|
| R150 | Certificate of patent or registration of utility model |
Ref document number: 7318907 Country of ref document: JP Free format text: JAPANESE INTERMEDIATE CODE: R150 |