JP5433182B2 - Waste material discrimination method and material discrimination device - Google Patents

Description

  The present invention relates to a waste material discrimination method and a waste material discrimination device, and more particularly to an improvement in waste material discrimination based on an image obtained by photographing waste.

  Conventionally, construction waste discharged when dismantling constructions such as structures and buildings has a sharp shape or is heavy, so usually heavy machinery and robots (hereinafter referred to as heavy machinery) It is carried out from the dismantling site using.

  Then, when carrying out these wastes, heavy machinery, etc. grabs these wastes with an attachment fork attached to the tip of a boom, etc., and holds the wastes in another place (loading platform of the transport vehicle). And the designated storage area).

  By the way, when grabbing waste with these heavy machinery or robots, if the type of material of the target waste can be determined in advance, its gripping work (adjustment of gripping force, etc.) and subsequent work (separation of waste) Etc.).

  Here, as a general method for discriminating the type of material, for example, the material is irradiated with near-infrared rays, X-rays or the like (hereinafter referred to as near-infrared rays), and the reflected light spectrum or fluorescent X-ray energy Detecting and analyzing materials, electron microscope, laser scanner, STM (Scanning Tunneling Microscope), AFT (Atomic Force Microscope), etc. (hereinafter referred to as electron microscope) The method of discriminating the material by observing the contour shape of the material surface at the molecular level and the atomic level as in the above), and the method of discriminating the material by measuring the surface roughness of the material using a surface roughness meter and so on.

  However, the method of irradiating near-infrared rays and the like and the method using a surface roughness meter are expensive to detect, and the sensor itself needs to be close to or in contact with the material to be detected. It takes a long time to discriminate, and is not suitable for a construction waste disposal site where a large amount of waste is to be discriminated.

  In addition, when construction waste is targeted, it is often used outdoors, but electron microscopes, etc. that are supposed to be used in extremely limited environments such as laboratories are often used for outdoor vibration. Under the influence of noise and sound, the original highly accurate detection cannot be performed, and the reliability of the detection result may be lowered.

  Furthermore, there is a possibility that the physical property value of the target object cannot be detected properly due to the influence of rain, water spray or dust attached to the surface of the target object, and an accurate detection result cannot be obtained.

  On the other hand, in recent years, with the widespread use of digital cameras, methods and systems for performing various processes based on images obtained by cameras have been devised, and even in the field of waste discrimination processing, A method and system for performing such discrimination have been proposed (Patent Documents 1 and 2).

  That is, Patent Document 1 discloses an apparatus that extracts a contour shape of an object from an input image and discriminates a fragment containing a recyclable resin based on the obtained contour shape.

  Patent Document 2 discloses an apparatus that detects the hue of an object from an input image and further discriminates an object of a predetermined metal material based on the area ratio occupied by the obtained hue on the object surface. Has been.

In addition to these, when an image in which a plurality of objects are mixed is displayed on the display device, the user uses the touch panel provided in the display device to limit the target objects to be identified in advance. A method for improving the recognition accuracy of the target object has also been proposed (Patent Document 3).
JP 2001-229385 A JP-A-7-275802 JP-A-10-181858

  However, since the construction waste has various shapes and sizes depending on how it is crushed and deformed when the construction is crushed, the technology disclosed in Patent Document 1 that is intended for finely crushed materials of home appliances It is difficult to apply to the construction waste as it is, and since the construction waste is often placed outdoors, the wastes overlap each other and were photographed. It is difficult to detect the contour shape of each waste based on the image, and is applied in a situation where the contour shape of each waste is easily detected by leveling on a belt conveyor in a factory. With technology, accurate discrimination results cannot be obtained.

  In addition, since the color of the target object changes greatly due to the effects of sunlight, watering, rain, and the like outdoors, and the color also changes due to rust, neutralization, attached dirt, etc., it is disclosed in Patent Document 2. It is not possible to obtain an appropriate discrimination result by the discrimination process using the hue that has been made.

  Note that the above-mentioned problems are mainly in the case of construction waste, but even in the case of general waste such as industrial waste other than construction waste, It is desired to perform the discrimination with higher accuracy.

  The present invention has been made in view of the above circumstances, and an object thereof is to provide a waste material discrimination method and a waste material discrimination device capable of accurately discriminating the material of an object.

Material discrimination method and material discrimination device of wastes according to the present invention, waste based on the shape information, the three information of the color information and texture information obtained on the basis of including a color image of waste only the image information By discriminating the material of an object, the material can be discriminated with higher accuracy than before.

That is, the material discrimination method of waste according to the present invention has been input, based only on the image information including color images of waste, the shape information regarding the shape of the waste, the color information on the color of the waste Further, it is characterized in that texture information relating to the rough surface of the waste is acquired, and the material of the waste is determined based on the acquired shape information, color information, and texture information.

  Here, the above-mentioned waste includes construction waste, that is, waste discharged by dismantling or crushing buildings or other structures (bridges, etc.) (whether it is actually discarded or used for recycling). Not only in the state where these construction waste is legally or illegally dumped, but also in various industrial waste other than construction waste, Waste also applies.

  In addition to waste left unmanaged or managed, waste in the present invention also includes waste flowing on a belt conveyor or the like of a recycling factory or garbage factory. Can be applied as

  In addition, the image information including the waste image may be, for example, a camera that captures a scene including the waste (a digital camera including an image sensor such as a CCD or CMOS, or an analog camera). Or may be input from a storage medium that temporarily stores an image taken by such a camera or a signal processing device that performs various signal processing.

  When inputting image information from such a camera or the like, the image information may be transmitted and input from a camera or the like by wire, or may be transmitted and input wirelessly.

  The shape information, the color information, and the texture information based on the image information may be acquired by performing signal processing (image processing) that is adapted to the image information.

  For example, as signal processing for acquiring shape information, processing for extracting an edge in an image can be applied, and the outline of the waste is thereby extracted, so that shape information relating to the shape of the waste can be obtained. it can.

  The shape information is not only information representing the contour of the shape but also information representing a distinction between a long and narrow shape in a specific direction and a shape that does not have such a non-uniform length. Also good.

  The color information is information representing a color and a hue. Therefore, it is assumed that the input image information is image information representing a color image at least for a waste image.

  The texture information is information that makes the user feel what kind of material, and is information that represents the rough surface of the waste. Specifically, this rough feeling is evaluated by the state of unevenness on the surface of the object.

  According to the waste material discrimination method according to the present invention configured as described above, based on the three appearance information acquired based on the input image information, that is, shape information, color information, and texture information, In order to discriminate the material of the waste contained in the image represented by the image information, the discrimination accuracy is remarkably increased as compared with the material discrimination based on only the shape information and the color discrimination based on the color information as in the past. be able to.

  Moreover, since the image information including the waste image is information that has been input in the conventional material determination process such as the material determination based only on the shape information and the material determination based only on the color information, the image information Since the process of inputting information is not a process newly added in the waste material discrimination method according to the present invention, it does not cause a significant cost increase.

  In the waste material discrimination method according to the present invention and the waste material discrimination apparatus according to the present invention to be described later, the meaning of “discriminate waste material” means “preparation of waste material name in advance” Not only means "select one of multiple types of material names", but also means "associate waste material names with probabilities corresponding to multiple types of prepared materials" Including.

  That is, when the result of determining the material is displayed on a monitor or printed on a medium such as paper, only one material name may be output as the determination result. A plurality of material names may be output together with the corresponding probabilities in the order of high probability or in the order of material names in alphabetical order or alphabetical order.

  In the waste material determination method according to the present invention described above, the first material determination result is obtained by determining the waste material candidate (meaning a material candidate; the same applies hereinafter) based on the shape information. The waste material candidates are determined based on the color information to obtain a second determination result, and the waste material candidates are determined based on the texture information to obtain a third determination result. Preferably, the material of the waste is specified by performing predetermined weighting on the obtained first discrimination result, the second discrimination result, and the third discrimination result.

  Here, in the process of specifying the material by weighting, the degree of importance should be added to the respective discrimination results of the first discrimination result, the second discrimination result, and the third discrimination result. Finally, any one material is specified. Specifically, each discrimination result (first discrimination result, second discrimination result, and third discrimination result) is Bayes-estimated ( This is a process that derives one discrimination result by integration based on Bayes theory) or integration based on Dempster & Shafer theory.

  According to the waste material discriminating method according to the preferred configuration of the present invention, when specifying one material based on three appearance information of shape information, color information and texture information, these three appearance information The first discrimination result (first material identification result) is obtained based on the shape information, and the second discrimination result (second material identification result is obtained based on the color information. ) And obtaining the third discrimination result (third material identification result) based on the texture information and the three discrimination results obtained in this first step It is appropriate for each situation by carrying out the second stage discrimination that is fixed in advance according to the type of construction, etc., or weighting set for each environmental condition such as weather and sunshine. Discrimination results that emphasize discrimination results Rukoto can, it is possible to increase the percentage of correct material discrimination further.

  In the waste material determination method according to the present invention described above, the color information is acquired based on color information representing the color of the waste image in the image information, and the texture information is the image information. Of these, it is preferable to obtain based on lightness difference variation information representing variation in lightness difference of the waste image.

  Here, as the color information representing the color of the image, for example, it is expressed by the a * value and b * value in the CIE color system L * a * b * (Elster, Aster, Baster; defined by JIS Z 8729). It is also possible to apply information expressed by various methods (color spaces) for expressing colors by a combination of numerical values.

  The brightness difference variation information representing the variation in brightness difference may be, for example, information expressed by variations in L * values in the CIE color system L * a * b *, or other known, It is possible to apply information expressed by various methods for expressing variations in brightness difference by numerical values.

  Note that the variation in brightness difference expresses the distribution of brightness within the region, and when the same brightness is distributed uniformly, the surface of the waste is in a smooth state with little roughness. If almost the same brightness is not evenly distributed, it can be said that the surface of the waste has a lot of shading due to the unevenness, and thus there is a lot of roughness.

  According to the waste material discriminating method according to the preferred configuration of the present invention, the color information can be acquired based on the color information representing the color of the waste image in the image information, and the texture information. Can be obtained on the basis of lightness difference variation information representing variation in lightness difference of waste images in the image information, so that color information and texture information can be obtained by simple signal processing.

  In the waste material discrimination method according to the present invention described above, the color information is defined by the a * value and the b * value when the image information is represented by the CIE color system L * a * b *. Is defined by an L * value when the image information is represented by a CIE color system L * a * b *, and brightness difference variation information is processed by a Laplacian filter for the coordinate system defined by the L * value. Is preferably defined by calculating a standard deviation in the coordinate system after the Laplacian filter processing.

  According to the waste material discrimination method according to the preferred configuration of the present invention, the a * value and the b * value when the color information is represented by the CIE color system L * a * b *. The brightness is defined by the L * value when expressed in the CIE color system L * a * b *, and the brightness difference variation information is processed by the Laplacian filter for the coordinate system defined by the L * value. Are obtained by calculating the standard deviation in the coordinate system after the Laplacian filter processing, thereby obtaining two pieces of information, color information and brightness difference variation information, based on the common CIE color system L * a * Therefore, it is possible to save the calculation processing.

  In the waste material discrimination method according to the present invention described above, the shape information is preferably obtained by performing an edge extraction process on the image information to detect a contour shape.

  According to the waste material discriminating method according to the preferred configuration of the present invention, shape information called a contour shape can be obtained by simple signal processing called edge extraction processing for image information.

  The obtained shape information is compared with a reference shape (reference shape) associated in advance for each material, and by selecting a close reference shape, waste close to the reference shape Can be specified (first discrimination result).

  In the waste material discrimination method according to the present invention described above, a region including the waste image to be discriminated from the image represented by the input image information is specified in advance and specified. Image information corresponding to the extracted area is extracted as specific area image information, and the shape information, the color information, and the texture information are obtained based on the extracted specific area image information, and the waste It is preferable to discriminate the material.

  According to the waste material discrimination method according to the preferred configuration of the present invention, by identifying and extracting in advance a waste subject to material discrimination from the visible image represented by the image information, It is possible to prevent incorrect extraction at the stage of processing to extract target waste from the entire image, and to prevent wrong material determination results based on erroneously extracted information from being derived. Accuracy can be further improved.

  In addition, as a method for specifying in advance an image of a target waste whose material is to be discriminated from images represented by image information, for example, an area specifying input such as a touch panel is displayed on a screen on which the image (visible image) is displayed. An operation unit may be provided, and a method for inputting an operation for specifying a target waste display area on the screen by the input operation unit may be applied, or a mouse may be used by using an input operation unit such as a mouse. You may make it specify with the cursor etc. on the screen linked | linked with the motion of.

  In addition, the waste display area may be specified using an input operation unit including a touch panel and a touch pen.

  In the waste material discrimination method according to the present invention described above, the waste is preferably construction waste.

  According to the waste material discrimination method according to the preferred configuration of the present invention, since the waste subject to material discrimination is limited to construction waste, it is possible to narrow down the types of materials to be discriminated. It is possible to further increase the accuracy of material discrimination.

  Construction waste, for example, aluminum, concrete, plastic, iron, wood, etc., is the main material. Therefore, in the material discrimination process, it is only necessary to discriminate one of these materials. The material discrimination accuracy can be improved as compared with the configuration for general waste including a wide variety of materials.

  In addition, since many construction wastes are characterized by the texture of the surface of the object, the material according to the method for determining the material of waste according to the present invention, which uses the texture information related to the texture as an element of material discrimination. The accuracy of discrimination can be obtained more effectively.

  In addition, since construction waste is often placed in the state of being stacked outdoors, the wastes overlap each other, and the shape of each waste is accurately detected based on the captured images. In addition, outdoors, the color of the target object changes greatly due to the effects of sunlight, watering, rain, etc., and only the color information because the color changes due to rust, neutralization, attached dirt, etc. However, the material discrimination method according to the present invention discriminates the material based on the three types of appearance information including the texture information in addition to the shape information and the color information. When construction waste is applied, a highly accurate discrimination result can be obtained more effectively.

  Construction waste generally refers to waste obtained by dismantling or crushing construction materials, but even if these are legally or illegally dumped wastes, Corresponds to waste.

The material discriminating apparatus of waste according to the present invention has been input, based only on the image information including color images of waste, the shape information acquiring unit that acquires shape information regarding the shape of the waste, the image A color information acquisition unit that acquires color information related to the color of the waste based on the information; a texture information acquisition unit that acquires texture information related to a rough surface of the waste based on the image information; A material for determining the material of the waste based on the shape information acquired by the shape information acquisition unit, the color information acquired by the color information acquisition unit, and the texture information acquired by the texture information acquisition unit And a determination unit.

  According to the waste material discrimination device according to the present invention configured as described above, the shape information acquired based on the image information input to the shape information acquisition unit, the image information input to the color information acquisition unit, Based on the three appearance information of the color information acquired based on and the texture information acquired based on the image information input to the texture information acquisition unit, the material determination unit is included in the image represented by the image information Since the material of the waste is discriminated, the discrimination accuracy can be remarkably improved as compared with the material discrimination based only on the shape information and the material discrimination based only on the color information as in the prior art.

  In the waste material discrimination device according to the present invention described above, the material discrimination unit means the waste material candidate (meaning a material candidate) based on the shape information acquired by the shape information acquisition unit. , The same), and a first determination unit that obtains a first determination result, and a second determination result by determining a material candidate for the waste based on the color information acquired by the color information acquisition unit. A second determination unit that obtains a third determination result by determining a material candidate for the waste based on the texture information acquired by the texture information acquisition unit, and the first determination unit Performing predetermined weighting on the first discrimination result obtained by the above, the second discrimination result obtained by the second discrimination unit, and the third discrimination result obtained by the third discrimination unit. To specify the material of the waste When, it is preferable that with a.

  According to the waste material discriminating apparatus according to the preferred configuration of the present invention, when the material discriminating unit specifies one material based on three appearance information such as shape information, color information, and texture information, Instead of directly specifying one material from these three appearance information, the first determination unit of the material determination unit obtains the first determination result (first material candidate) based on the shape information, and the material determination unit The second determination unit obtains the second determination result (second material candidate) based on the color information, and the third determination unit of the material determination unit determines the third determination result (third material candidate) based on the texture information. ), The material specifying unit performs predetermined weighting on the three determination results (first determination result, second determination result, and third determination result) thus obtained, In order to identify the material of the waste, appropriate judgment results are overlapped for each situation. Determination result can be obtained that can increase the percentage of correct material discrimination further.

  In the waste material discrimination apparatus according to the present invention described above, the color information acquisition unit acquires the color information based on color information representing the color of the waste image in the image information. And the texture information acquisition unit acquires the brightness difference variation information representing the variation in brightness difference based on the brightness information representing the brightness of the waste image in the image information, It is preferable that the texture information is acquired based on the brightness difference variation information.

  According to the waste material discrimination apparatus according to the preferred configuration of the present invention, the color information is acquired by the color information acquisition unit based on color information representing the color of the waste image in the image information. Since the texture information can be acquired by the texture information acquisition unit based on the brightness difference variation information representing the variation in the brightness difference of the waste image in the image information, the color information acquisition unit and the texture information Each acquisition unit can obtain color information and texture information by simple signal processing.

  In the waste material discrimination apparatus according to the present invention described above, the color information acquisition unit includes the color information, an a * value when the image information is represented by a CIE color system L * a * b *, and The texture information acquisition unit defines the brightness by an L * value when the image information is represented by a CIE color system L * a * b *, and the brightness difference variation is defined by a b * value. It is preferable that the information is defined by performing a Laplacian filter process on the coordinate system defined by the L * value and calculating a standard deviation in the coordinate system after the Laplacian filter process.

  According to the waste material discriminating apparatus according to the preferred configuration of the present invention, the color information acquisition unit is configured so that the color information is represented by a when the image information is represented by the CIE color system L * a * b *. The texture information acquisition unit defines the brightness by the L * value when expressed in the CIE color system L * a * b *, and the brightness difference variation information is expressed by the L * value. The Laplacian filter process is performed on the coordinate system defined in (2), and the standard deviation in the coordinate system after the Laplacian filter process is calculated. It can be obtained on the basis of the color system L * a * b *, and the calculation process can be saved.

  In the waste material discrimination apparatus according to the present invention described above, it is preferable that the shape information acquisition unit obtains the shape information by performing an edge extraction process on the image information.

  According to the waste material discriminating apparatus according to the preferred configuration of the present invention, the shape information acquisition unit can obtain shape information called a contour shape by simple signal processing called edge extraction processing on image information. it can.

  Then, the material discriminating unit compares the obtained shape information with a reference shape (reference shape) previously associated with each material, and selects a reference shape that is close to the reference shape. It is possible to identify the waste material close to the shape (first discrimination result).

  In the waste material discrimination apparatus according to the present invention described above, an image display unit that displays an image represented by the input image information, and an arbitrary region is specified among the images displayed on the image display unit. A region specifying input operation unit for inputting an operation to be performed, and a specific region extracting unit for extracting image information corresponding to a region specified by the region specifying input operation unit from among the input image information as specific region image information And, based on the specific region image information extracted by the specific region extraction unit, the shape information acquisition unit acquires waste shape information included in the specific region image information, and the color The information acquisition unit acquires color information of the waste included in the specific area image information, the texture information acquisition unit acquires the texture information of the waste included in the specific area image information, and the material Another portion is preferably to determine the material of the waste contained in the specific area image information.

  According to the waste material discrimination apparatus according to the preferred configuration of the present invention, the waste to be subjected to material discrimination among the visible images represented by the image information is specified in advance by the area specifying input operation unit, and specified. The area extracting unit extracts image information (specific area image information) corresponding to the specified area from the entire image information, and thereby erroneously extracts the target waste at the stage of processing. This can prevent the occurrence of an erroneous material discrimination result based on erroneously extracted information, thereby further improving the accuracy of the material discrimination.

  Of the images represented by the image information, as the region specifying input operation unit, for example, a touch panel provided so as to overlap the screen of the image display unit on which the image (visible image) is displayed can be applied. Then, the user may input an operation for specifying the display area of the target waste on the screen on the touch panel, or by using an area specifying input operation means such as a mouse. Alternatively, it may be specified by a cursor on the screen linked to the movement of the mouse.

  An area specifying input operation unit including a touch panel and a touch pen can also be applied, and the waste display area may be specified by the area specifying input operation unit.

  In the waste material discrimination apparatus according to the present invention described above, the shape information acquisition unit, the color information acquisition unit, and the texture are obtained by capturing an image including the waste and capturing image information including the waste image. It is preferable that the camera further includes a camera that inputs to the information acquisition unit.

  According to the waste material discrimination apparatus according to the preferred configuration of the present invention, the camera captures an image including the waste, and the image information including the waste image is obtained as a shape information acquisition unit. The information can be input to the color information acquisition unit and the texture information acquisition unit.

  The camera may be a digital camera provided with an image sensor such as a CCD or CMOS, or an analog camera.

  In the waste material discrimination apparatus according to the present invention described above, the waste is preferably construction waste.

  According to the waste material discriminating apparatus according to the preferred configuration of the present invention, since the waste subject to material discrimination is limited to construction waste, it is possible to narrow down the types of materials to be discriminated. It is possible to further increase the accuracy of material discrimination.

  In addition, since construction waste is often placed in the state of being stacked outdoors, the wastes overlap each other, and the shape of each waste is accurately detected based on the captured images. In addition, outdoors, the color of the target object changes greatly due to the effects of sunlight, watering, rain, etc., and only the color information because the color changes due to rust, neutralization, attached dirt, etc. However, the material discrimination device of the present invention discriminates the material based on the three types of appearance information including the texture information in addition to the shape information and the color information. When construction waste is applied, a highly accurate discrimination result can be obtained more effectively.

  Construction waste generally refers to waste obtained by dismantling or crushing construction materials, but even if these are legally or illegally dumped wastes, Corresponds to waste.

  According to the waste material discrimination method and the waste material discrimination device according to the present invention, the discrimination accuracy is higher than that of the material discrimination based only on the shape information and the color information based on the color information as in the prior art. It can be remarkably improved, and the material of the object (waste) can be distinguished with high accuracy.

BEST MODE FOR CARRYING OUT THE INVENTION Preferred embodiments of a waste material discrimination method and a material discrimination device according to the present invention will be described below with reference to the drawings.
(Embodiment 1)
FIG. 1 is a block diagram showing a basic configuration of a waste material discrimination apparatus 100 according to an embodiment of the present invention, and FIG. 2 is a disposal subject to material discrimination by the material discrimination apparatus 100 shown in FIG. FIG. 2 is a schematic diagram showing construction waste 200 (hereinafter simply referred to as waste 200), which is an example of a product, where (a) is a concrete piece, (b) is an aluminum bar piece, and (c) is a plastic pipe. A piece, (d) shows a reinforcing bar piece, and (d) shows a piece of wood. FIG. 3 is a flowchart for explaining the operation of the material discriminating apparatus 100 shown in FIG.

  In the illustrated material discriminating apparatus 100, an image P including any one of these wastes 200 is captured by an external camera 110 that is not configured by the material discriminating apparatus 100, and the image P captured by the camera 110 is captured by the image P. Is converted into image information S as a signal (data) representing the image and output from the camera 110, and the output image information S is input to the material discrimination device 100.

  The material discriminating apparatus 100 obtains three pieces of appearance information, ie, shape information, color information, and texture information, based on the input image information S, and the image information S is obtained based on the obtained three pieces of appearance information. The material of the waste 200 contained in the image P to be represented is discriminated.

  Here, the camera 110 is treated as not being a component of the material discrimination device 100 in the description of the present embodiment, but may be handled as a component of the material discrimination device 100 of the present embodiment.

  The material discriminating apparatus 100 is based on image information S including an image of the waste 200 input from at least one of the two cameras 110 provided at slightly different positions and having different parallaxes. The shape information acquisition unit 10 that acquires the shape information M1 related to the shape of 200, the color information acquisition unit 20 that acquires the color information M2 related to the color of the waste 200 based on the image information S, and the waste based on the image information S The texture information acquisition unit 30 that acquires the texture information M3 related to the surface roughness of the surface 200, the shape information M1 acquired by the shape information acquisition unit 10, the color information M2 acquired by the color information acquisition unit 20, and the texture information acquisition unit 30, a material determination unit 40 that determines the material of the waste 200 based on the texture information M <b> 3 acquired by 30.

  In FIG. 1, reference numeral 190 denotes a position detection device that detects the presence position of the photographed waste 200 based on the image information S input from the camera 110, and reference numeral 180 denotes the position detection device 190. Information L related to the position detected by the information determining unit 40 and information K related to the material determined by the material determining unit 40 are input. Based on the input information L related to the position and information K related to the material, This is an operation processing device that performs some operation.

  The camera 110, the position detection device 190, and the operation processing device 180, together with the material discrimination device 100 of the present embodiment, as an example of an operation processing system for the waste 200, for example, holds the waste 200 and sorts the materials by type. You may make it comprise as a heavy machine etc. which perform a classification process.

  In addition, when inputting the image information S from the camera 110 to the material determination apparatus 100, the image information S may be transmitted from the camera 110 by wire or may be input by wireless transmission.

  Specifically, the material determination unit 40 of the material determination apparatus 100 determines the material candidate of the waste 200 based on the shape information M1 acquired by the shape information acquisition unit 10, and obtains the first determination result K1. A second determination unit 60 that determines a material candidate of the waste 200 based on the color information M2 acquired by the color information acquisition unit 20 and the color information acquisition unit 20, and obtains a second determination result K2, and a texture information acquisition unit 30. Based on the acquired texture information M3, a third determination unit 70 that determines a material candidate of the waste 200 and obtains a third determination result K3; a first determination result K1 obtained by the first determination unit 50; By performing predetermined weighting on the second discrimination result K2 obtained by the second discrimination unit 60 and the third discrimination result K3 obtained by the third discrimination unit 70, the material of the waste 200 is reduced. Specific and specific And the material identifying unit 80 for outputting information about a single material K, a configuration in which a.

  In detail, the material specifying unit 80, for example, by performing predetermined weighting on the first determination result K1, the second determination result K2, and the third determination result K3, respectively, for a plurality of types of materials, Probabilities representing the possibility of corresponding to the material of the waste 200 that is the object of the material discrimination are associated, and one material having the highest probability is specified as the material of the waste 200.

  However, if the probabilities for two or more materials are similar, and one of them shows the highest probability, then for some materials that have similar probabilities, either material is the correct material. If the determination is simply based on the magnitude of the probability, there is a high risk that the wrong material will be specified.In such a case, specify multiple types of materials whose probability values are approximated as one. Instead, present the multiple material names together with the probability value, or change the weighting conditions, etc., such as "There are multiple approximate material candidates." Or the like may be presented.

  Further, in the material discriminating apparatus 100 for the waste 200 described above, the shape information acquisition unit 10 obtains the shape information M1 by performing an edge extraction process on the image information S, and the color information acquisition unit 20 The color information M2 is acquired based on the color information representing the color of the image of the waste 200 in the image information S, and the texture information acquisition unit 30 determines the brightness of the image of the waste 200 in the image information S. Lightness difference variation information representing variations in lightness difference is obtained based on the lightness information that is represented, and texture information is obtained based on the lightness difference variation information.

Here, the acquisition process of the shape information M1 by the shape information acquisition unit 10 includes, for example, a threshold value appropriately set for brightness and the like in addition to the process of performing the edge extraction process on the image information S to obtain the contour shape. The binarization process used, the “pattern matching” process to compare with pre-stored / registered reference shapes, the “labeling” process to compare with pre-stored / registered data such as waveforms and numerical values, and noise components The shape information M1 may be obtained by combining a plurality of various processes as necessary, such as a “smoothing” process that erases the necessary shape information.

  More specifically, the color information acquisition unit 20 defines the color information based on the a * value and the b * value when the image information S is expressed in the CIE color system L * a * b * (JIS Z 8729). The texture information acquisition unit 30 defines brightness by the L * value when the image information S is represented by the CIE color system L * a * b *, and the coordinate system defined by the L * value ( The Laplacian filter process is performed on a large number of pixels constituting the image information S (represented by the corresponding L * values), and the standard deviation in the coordinate system after the Laplacian filter process is calculated, thereby obtaining a brightness difference. Specifies variation information (standard deviation of brightness difference variation).

  Next, the operation of the waste 200 material discrimination apparatus 100 according to the present embodiment (the embodiment of the waste material discrimination method according to the present invention) will be described with reference to the flowchart of FIG.

  First, an image P including any waste 200 in FIGS. 2A to 2E is captured by the camera 110, and the camera 110 converts the captured image P into image information S that is a signal. The image information S is output to the material discrimination device 100 and the position detection device 190.

  Here, as a method in which the position detection device 190 detects the existence position of the photographed waste 200 based on the image information S, for example, the same object is obtained by two cameras 110 provided at slightly different positions. By photographing (waste 200), two slightly different pieces of image information S1 and S2 with different parallaxes are obtained, and the installation position interval (known) of both cameras 110 and 110 and the two pieces of image information S1 and S2 Based on the principle of triangulation based on the correspondence relationship, a method of detecting the distance from the cameras 110, 110, that is, the position where the waste 200 exists can be applied.

  The position detection device 190 is an application example (operation processing for waste) in which the operation by the operation processing device 180 for the target waste 200 whose material is to be determined corresponds to the position where the waste 200 exists. Needless to say, it is only necessary when it is described as a (system), and is not essential as a configuration of the material discrimination device 100 of the present embodiment.

  Image information S (step 10 (# 10 in FIG. 3)) input from the camera 110 to the material discrimination device 100 is input to the shape information acquisition unit 10, the color information acquisition unit 20, and the texture information acquisition unit 30, respectively.

  The shape information acquisition unit 10 to which the image information S is input performs an edge extraction process on the image information S (# 21), and after the edge extraction process (# 21), the waste 200 included in the image information S is obtained. The contour shape of the image is recognized (# 22), and the recognized contour shape (whether it is a long and narrow shape in a specific direction or a block-like shape without such a non-uniform length) In general, aluminum bar pieces, plastic pipe pieces, rebar pieces, etc. are elongated, and concrete pieces and wood pieces are block-like)) as shape information M1. (A series of processing for obtaining and outputting the shape information M1 is expressed as step 20 (# 20)).

  Further, the color information acquisition unit 20 to which the image information S is input acquires RGB data for each pixel constituting the image information S based on the image information S (# 31), and the obtained RGB data of each pixel. Are converted into CIE color system L * a * b * (# 32), respectively, and a color difference (color information) a * value and b * value are acquired from this CIE color system L * a * b * ( In step # 33, the obtained a * value and b * value for each pixel are output as color information M2 to the material discriminating unit 40 (a series of processing for obtaining and outputting the color information M2 is performed in step 30 (# 30). ).

  Furthermore, the texture information acquisition unit 30 to which the image information S has been input acquires RGB data for each pixel constituting the image information S based on the image information S (# 41), and the obtained RGB data is CIE. It converts to the color system L * a * b * (# 42), acquires the lightness L * value of this CIE color system L * a * b * (# 43), and for each pixel, the surrounding 8 pixel L A Laplacian filter process using the * value is performed (the Laplacian filter process is performed on the coordinate system represented by the L * value), and the standard deviation value σ in the coordinate system after the Laplacian filter process is calculated (# 44) Then, the obtained standard deviation value σ of the variation in brightness difference is output to the material discriminating unit 40 as the texture information M3 (a series of processing for obtaining and outputting the texture information M3 is expressed as step 40 (# 40). ).

  Here, in step 43 (# 43), a Laplacian filter process is performed on the coordinate system in which each pixel is expressed by an L * value. Further, in step 44 (# 44), the standard deviation value in the coordinate system after the Laplacian filter process is applied. The process for calculating σ will be described in detail.

  FIG. 4 is a schematic diagram showing a state in which the waste material 200 of a plurality of types of materials shown in FIG. 2 is included in one image P, and is used in the material discrimination device 100 of the embodiment shown in FIG. It is not an image (an image including only one waste 200) but an image used for a material discrimination device according to another embodiment to be described later, but the Laplacian filter processing is performed in the coordinate system expressed by the L * value described above. The processing for performing the above and the processing for obtaining the standard deviation value σ are the same in this embodiment and the embodiments described later, and will be described with reference to FIG.

  FIG. 4B is an enlarged view of a portion P ′ of the plastic tube piece C extracted from the image P shown in FIG. 4A. The image P used in the material discrimination device 100 of this embodiment is shown in FIG. It can be considered as equivalent, ie an image containing a single waste 200.

  Therefore, RGB data for each pixel is acquired for a part P ′ of this image (# 41), and the obtained RGB data is converted into the CIE color system L * a * b * (# 42). FIG. 5A shows a coordinate system in which the lightness L * value is acquired from the CIE color system L * a * b * (# 43) and each pixel r (i, j) is expressed by the L * value. .

  As shown in FIG. 4B, since the surface of the plastic tube piece C is relatively smooth, the distribution of L * values shows a smooth distribution as shown in FIG.

Then, for the L * value (L * (r (i, j))) of the specific pixel of interest r (i, j), the surrounding 8 pixels (r (i-1, j-1), r (i, j) -1), r (i + 1, j-1), r (i-1, j), r (i + 1, j), r (i-1, j + 1), r (i, j + 1), r (i + 1, j + 1) )) When the Laplacian filter processing using each L * value is performed, the L * ′ value (L * ′ (r (i, j))) of each target pixel r (i, j) after processing is
L * ′ (r (i, j))
= L * (r (i-1, j-1)) + L * (r (i, j-1))
+ L * (r (i + 1, j-1)) + L * (r (i-1, j))
+ L * (r (i + 1, j)) + L * (r (i-1, j + 1))
+ L * (r (i, j + 1)) + L * (r (i + 1, j + 1))
−8 × L * (r (i, j))
It becomes.

  Then, the distribution of the L * ′ value after the target pixel is sequentially shifted and the Laplacian filter processing is performed on all the pixels is as shown in FIG. 5B, as can be understood from FIG. 5B. In the waste 200 made of a material (aluminum material, plastic material, non-rusted iron, etc.) having a smooth surface such as the plastic pipe piece C, the distribution is smoother than the distribution shown in FIG. Become.

  By calculating the standard deviation value σ of the distribution of the lightness L * ′ value after the Laplacian filter processing (# 44), the lightness variation in the image P ′ can be defined quantitatively, In the waste material 200 having a smooth texture, the standard deviation value σ is a relatively small value.

  On the other hand, in the case of a concrete piece which is a kind of other material that is not a plastic pipe piece, for example, a part P ′ of the concrete piece B is extracted and enlarged from the image P shown in FIG. As can be seen from (b), since the surface has irregularities, it has a rough texture, and the distribution of L * values is the distribution shown in FIG. 7 (a).

  Then, the distribution of L * ′ values after performing the same Laplacian filter processing as in FIG. 5B is as shown in FIG. 7B, and the lightness L * ′ value shown in FIG. 7B. The standard deviation value σ (# 44) of the distribution state of the material is larger than the standard deviation value σ of the waste 200 of a material (aluminum material, plastic material, non-rusted iron, etc.) whose surface has a smooth texture. It becomes.

  That is, in the waste material 200 having a rough texture on the surface (concrete material, wood, rusted iron, etc.), the standard deviation value σ is the standard for the waste material 200 having a smooth texture on the surface. The value is larger than the deviation value σ.

  The process of step 31 (# 31) by the color information acquisition unit 20 is the same as the process of step 41 (# 41) by the texture information acquisition unit 30, and the process of step 32 (# 32) by the color information acquisition unit 20 Since the process is the same as the process of step 42 (# 42) by the texture information acquisition unit 30, it is preferable to reduce waste of performing these common processes in duplicate.

  Therefore, the color information acquisition unit 20 and the texture information acquisition unit 30 in FIG. 1 are integrated, and step 41 (# 41) and step 42 (# 42) by the texture information acquisition unit 30 are omitted, and the color information acquisition unit Step 31 (# 31) and step 32 (# 32) according to 20 may be used as alternative processing for step 41 (# 41) and step 42 (# 42).

  Next, the shape information M1 of the waste 200 obtained by the shape information acquisition unit 10 is input to the first determination unit 50 of the material determination unit 40. The first determination unit 50 is not illustrated. A storage unit is provided, and a reference model shape (such as an elongated shape or a block shape, hereinafter referred to as a reference shape) corresponding to each material of the waste 200 is stored in advance in the storage unit. Each material is associated with a reference shape corresponding to it.

  That is, for example, concrete and wood are associated with each other as a block-shaped reference shape, and aluminum material, plastic material, reinforcing bar material, and the like are associated with each other as an elongated reference shape.

  The first determination unit 50 compares the contour shape represented by the input shape information M1 with the reference shape stored in the storage unit, and selects the closest reference shape (step 51a (# in FIG. 8) 51a)), the material associated with the selected reference shape is output as a candidate K1 to the material specifying unit 80 (step 51b (# 51b)).

  Note that a series of these shape comparison processing (# 51a) and material candidate K1 output processing (# 51b) is denoted as step 51 (# 51) in FIG.

  In addition, the color information M2 (a combination of a * value and b * value) for each pixel of the waste 200 obtained by the color information acquisition unit 20 is input to the second determination unit 60 of the material determination unit 40. However, the second discriminating unit 60 includes a storage unit (not shown), and the storage unit includes an orthogonal coordinate system in which the a * value and the b * value shown in FIG. A reference position corresponding to each material of the waste 200 in (hereinafter referred to as a * b * coordinate system) is stored in advance.

  Note that the reference position (combination of a * value and b * value) in the a * b * coordinate system for each material shown in FIG. 9 is experimentally determined in advance for each material of typical construction waste. In rusted iron and wood, the color of the material is red or orange, so both a * and b * values show a relatively large value on the positive side. Since the color of is a yellow or gray color, the b * value shows a relatively large value, while the a * value shows a relatively small value.

  In concrete materials (watering), aluminum materials, and plastic materials that have changed color due to moisture, the a * and b * values are relatively small because the material color is mainly gray. In the case of iron that is not rusted or aluminum material (with dust) to which dust of concrete powder adheres, the color of the material is more gray and close to blue than the plastic material, and the b * value is even smaller.

  And the 2nd discrimination | determination part 60 is the coordinate position M2 specified by the color information (a * value, b * value) for each input pixel, and the reference position of each material memorize | stored in the memory | storage part. Distance G (distance G1 to wood, distance G2 to rusted iron, distance G3 to concrete material (dry state), distance G4 to concrete material (during watering), distance G5 to aluminum material, distance to plastic material Measure (calculate) the distance G6, the distance G7 to the rust-free iron, and the distance G8 to the aluminum material (with dust) (step 52a (# 52a) in FIG. 10), and for all the pixels, up to each material ΣG1, ΣG2, ΣG3, ΣG4, ΣG5, ΣG6, ΣG7, ΣG8.

  Here, the material of the reference position with the shortest sum ΣG1, ΣG2, ΣG3, ΣG4, ΣG5, ΣG6, ΣG7, and ΣG8 of the distances for each material is the closest material candidate K2 from the viewpoint of the color information M2. Therefore, the probability corresponding to each material is calculated based on the sum ΣGi of each distance according to a preset arithmetic expression or the like, and the material having the highest probability, that is, the sum ΣGi of the distance is shortest. The material at the reference position is selected as the candidate K2 and output to the material specifying unit 80 (step 52b (# 52b)).

  When the candidate K2 is output to the material specifying unit 80, the probability corresponding to each material according to the above-described arithmetic expression is also output to the material specifying unit 80.

  A series of the distance measurement process (# 52a) and the material candidate K2 output process (# 52b) is represented as step 52 (# 52) in FIG.

  In addition, the texture information M3 of the waste 200 obtained by the texture information acquisition unit 30 is input to the third determination unit 70 of the material determination unit 40. The third determination unit 70 stores a memory (not shown). The standard value (standard standard deviation value) of the standard deviation value σ corresponding to each material of the waste 200 shown in FIG. 11 is stored in advance in the storage unit.

  Note that the simple average value Ave of the reference standard deviation value σ and the L * value for each material shown in FIG. 11 is a value obtained experimentally and empirically in advance.

  Then, the third determination unit 70 compares the standard deviation value σ represented by the input texture information M3 with the standard standard deviation value of each material stored in the storage unit (step 53a (# 53a in FIG. 12). )), The material corresponding to the reference standard deviation value with the smallest difference is selected as the candidate K3 and output to the material specifying unit 80 (step 53b (# 53b)).

  A series of processing of comparison processing (# 53a) with these reference standard deviation values and output processing (# 53b) of material candidate K3 is denoted as step 53 (# 53) in FIG.

  As described above, the first determination result (material candidate) K1 from the first determination unit 50, the second determination result (material candidate) K2 from the second determination unit 60, and the third determination result from the third determination unit 70. (Material Candidate) The material specifying unit 80 to which K3 is respectively input performs a predetermined weighting on the first determination result K1, the second determination result K2, and the third determination result K3, thereby determining these determinations. The results K1, K2, and K3 are integrated to finally identify the material of the waste 200 as one, and information on the identified single material K is output as a discrimination result (# 54).

  A specific method for integrating these three discrimination results K1, K2, and K3 to finally identify one discrimination result is Bayesian estimation (estimation based on Bayes theory) or Dempster & Shafer. Integration based on theory can be applied.

  As described above in detail, according to the material discriminating apparatus 100 for the waste 200 and the material discriminating method that is the function thereof according to the present embodiment, it is acquired based on the image information S input to the shape information acquiring unit 10. The shape information M1, the color information M2 acquired based on the image information S input to the color information acquisition unit 20, and the texture information M3 acquired based on the image information S input to the texture information acquisition unit 30. Based on the three pieces of appearance information, the material discrimination unit 40 discriminates the material of the waste 200 included in the image P represented by the image information S, so that the material discrimination based only on the shape information M1 as in the prior art. Compared to material discrimination based only on the color information M2, the discrimination accuracy can be significantly improved.

  FIG. 13 is a graph showing an example of demonstrating this effect. The waste 200 of each material of aluminum, concrete, plastic, iron (with rust), wood, and iron (without rust) is shown in this embodiment. The accuracy rate of the discrimination result of the material is expressed by the material discrimination device 100 and the material discrimination method.

  In the figure, the material discrimination based on the a * b * color difference represents a conventional material discrimination process that depends only on color information (corresponding to the color information M2 in the present embodiment), and is based on Bayesian estimation (estimation based on Bayes theory) and Material determination based on integration based on Dempster & Shafer theory is based on Bayesian estimation and Dempster Shafer theory in the processing (# 54) of the material specifying unit 80 in the material determination apparatus 100 and the material determination method of this embodiment. Represents the material discrimination process to which each is applied.

  As can be seen from this graph, for all materials, the material discrimination processing of this embodiment using Bayesian estimation or the material discrimination processing of this embodiment using Dempster-Shafer theory is based on the conventional material discrimination processing. It was also demonstrated that the correct answer rate is high.

  Depending on the material, both the material discrimination process using Bayesian estimation and the material discrimination process using Dempster-Shafer theory may be less than the accuracy rate of the conventional material discrimination process. A discrimination process and a material discrimination process using the Dempster-Shafer theory may be performed, respectively, and a material with higher accuracy may be selected.

  Information regarding the material K output from the material discrimination device 100 in this embodiment is input to the operation processing device 180 outside the material discrimination device 100, and the operation processing device 180 is input from the position detection device 190. Depending on the information L on the position of the waste 200 and the material K input from the material discriminating apparatus 100, the waste 200 is gripped in accordance with, for example, the gripping force suitable for the material K (the gripping suitable for the brittleness and weight of each material). Force) and transport it to a place sorted by the type of material K.

  At this time, the material discriminating apparatus 100 according to the present embodiment can discriminate the material K of the waste 200 with higher accuracy than before, so that the gripping force when gripping the waste 200 is set appropriately than before. be able to.

  Further, when the waste 200 is rotated while being held, a centrifugal force acts on the waste 200, or when an inertial force acts on the waste 200 due to acceleration at the start of movement or deceleration when stopped, the waste 200 The mass of the waste 200 can be appropriately estimated by accurately discriminating the material, and the rotation speed and the relationship with the gripping force gripping the waste 200 can be estimated. It is also possible to set the acceleration / deceleration appropriately.

  Furthermore, according to the material discriminating apparatus 100 and the material discriminating method for the waste 200 according to the present embodiment, the material discriminating unit 40 has one piece of information based on the three pieces of appearance information of the shape information M1, the color information M2, and the texture information M3. When the material K is specified, one material is not directly specified from these three appearance information, but the first determination unit 50 of the material determination unit 40 uses the first determination result (the first determination result based on the shape information M1). Material candidate) K1 is obtained, and the second discrimination unit 60 of the material discrimination unit 40 obtains a second discrimination result (second material candidate) K2 based on the color information M2, and the third discrimination unit 70 of the material discrimination unit 40 is obtained. Obtains a third discrimination result (third material candidate) K3 based on the texture information M3, and the three discrimination results thus obtained (first discrimination result K1, second discrimination result K2, 3 for the discrimination result K3) In order to identify the material K of the waste 200 by the unit 80 performing a predetermined weighting, a determination result K that places importance on an appropriate determination result (any one of K1, K2, and K3) for each situation such as the surrounding environment. It is possible to obtain the correct answer rate of the material discrimination.

  Further, according to the material determination apparatus 100 according to the present embodiment, the color information M2 is obtained by the color information acquisition unit 20 from the color information (a * value and b *) representing the color of the image of the waste 200 in the image information S. The texture information M3 can be acquired by the texture information acquisition unit 30 by the texture information acquisition unit 30. The brightness difference variation information (standard deviation) indicating the variation in the brightness difference of the image of the waste 200 in the image information S. Since the color information acquisition unit 20 and the texture information acquisition unit 30 can obtain the color information M2 and the texture information M3 by simple signal processing, respectively.

  Furthermore, according to the material discriminating apparatus 100 according to the present embodiment, the color information acquisition unit 20 uses the a * value when the color information M2 is represented by the CIE color system L * a * b *. The texture information acquisition unit 30 defines the brightness by the L * value when expressed in the CIE color system L * a * b *, and the brightness difference variation information by the L * value. Since the Laplacian filter process is performed on the defined coordinate system and the standard deviation value σ in the coordinate system after the Laplacian filter process is calculated, the two types of information, the color information and the brightness difference variation information, are shared. It can be obtained on the basis of the CIE color system L * a * b *, and the calculation processing can be saved.

  Moreover, according to the material discrimination device 100 according to the present embodiment, the shape information acquisition unit 10 can obtain the shape information M1 called the contour shape by the simple signal processing called edge extraction processing on the image information S.

  And the material discrimination | determination part 40 compares this obtained shape information M1 with the reference | standard shape (reference | standard shape) previously matched for every material, By selecting the thing of a near reference | standard shape, The material of the waste 200 close to the reference shape can be specified (first determination result).

  Further, according to the material discrimination device 100 according to the present embodiment, the waste subject to material discrimination is the construction waste 200 obtained by dismantling or crushing the construction, and therefore the material of the waste 200 Can be limited in advance to several types such as concrete, plastic, aluminum, wood, iron, etc., and the material discrimination accuracy can be further enhanced.

  Furthermore, since the construction waste 200 is often placed outdoors, the color of the construction waste 200 changes greatly due to the effects of sunlight, watering, rainfall, etc., and rust and neutralization are caused. However, it is difficult to accurately determine the material using only the color information M2 because the color changes depending on the attached dirt and the like. However, the material determination device and the material determination method according to the present embodiment use the shape information M1 and the color information M2. In addition, since the material is discriminated based on the three types of appearance information including the texture information M3, when the construction waste 200 is applied as the material discrimination target, a more accurate discrimination result can be obtained more effectively. .

In addition, in the material discriminating apparatus 100 for the waste 200 and the material discriminating method as the operation thereof according to the present embodiment, the first discrimination result K1 is obtained based on the shape information M1, and the second discriminating based on the color information M2. A first step (# 51, # 52, # 53) in which a discrimination result K2 is obtained and a third discrimination result K3 is obtained based on the texture information M3, and then the first discrimination result K1, second The second step (# 54) in which the discrimination result K2 and the third discrimination result K3 are integrated to obtain one discrimination result K is a separate step, but the waste according to the present invention The material discriminating apparatus and the material discriminating method are not limited to this form, and any material can be used as long as it identifies a single material based on three pieces of appearance information including shape information M1, color information M2, and texture information M3. Well, wood The determination unit 40, the first determination unit 50, the second determination unit 60, nor is it intended to be limited to the third determination unit 70 and forms a separate component that material identifying unit 80.
(Embodiment 2)
In the first embodiment described above, the image P represented by the input image information S includes only an image of one waste 200 (only one of (a) to (e) in FIG. 2). However, the waste material discrimination device according to the present invention is not limited to this embodiment.

  That is, FIG. 14 targets an image P including a plurality of wastes 200 as shown in FIG. 4A or FIG. 6A, for example, and any one waste 200 included in the image P is displayed. It is a block diagram which shows the structure of the waste material discrimination | determination apparatus 100 based on one Embodiment of this invention which makes a material discrimination | determination object.

  The material discrimination apparatus 100 shown in the drawing has the same basic configuration as the material discrimination apparatus 100 of the first embodiment shown in FIG.

  The material discriminating apparatus 100 according to the second embodiment shown in FIG. 14 has a monitor 92 (image display unit) that displays the image P represented by the input image information S as a visible image with respect to the material discriminating apparatus 100 shown in FIG. ) And a touch panel 93 (region specifying input operation unit) for inputting an operation for specifying an arbitrary region in the image P displayed on the monitor 92 and the touch panel 93 among the input image information S. For example, an extraction unit 94 (specific region extraction unit) that extracts image information corresponding to the region P ′ (see FIGS. 4A and 6A) as specific region image information S ′, a monitor 92, and an extraction unit The display control unit 91 further controls the transmission of the image information S to 94.

  Moreover, although the material discrimination | determination apparatus 100 of Embodiment 1 is the structure which acquires the shape information M1, the color information M2, and the texture information M3 based on the image information S showing the whole image P input from the camera 110, implementation Based on the specific area image information S ′ extracted by the extraction unit 94, the shape information acquisition unit 10 of the form 2 material determination apparatus 100 uses the shape information M1 of the waste 200 included in the specific area image information S ′. The color information acquisition unit 20 acquires the color information M2 of the waste 200 included in the specific area image information S ′, and the texture information acquisition unit 30 acquires the waste 200 of the waste 200 included in the specific area image information S ′. The material quality information M3 is acquired, and the material determination unit 40 determines the material of the waste 200 included in the specific area image information S ′.

  That is, the material discrimination device 100 according to the second embodiment is configured so that an image P ′ including only one waste 200 among the images P including a large number of wastes 200 is a target of the material discrimination process. The configuration is basically the same as that of the material discrimination device 100 of the first embodiment except that a configuration for processing the target image information S is added.

  According to the material discrimination device 100 of the second embodiment configured as described above, as shown in the flowchart of FIG. 15, the image information S representing the entire image P is input from the camera 110 to the display control unit 91 (# 10 The display control unit 91 displays the entire image P on the monitor 92 (# 11), and among the entire image P displayed on the monitor 92, one waste 200 that the user intends to make a material discrimination target. Is waited for input to the touch panel 93 superimposed on the monitor 92 (# 12).

  Then, when the area including the waste 200 that is the object of material discrimination is specified by the touch panel 93, the specified area P ′ is input to the extraction unit 94, and the extraction unit 94 is input from the display control unit 91. The image information (specific area image information) S ′ corresponding to the specified area P ′ is extracted from the image information S representing the entire image P (# 13), and the shape information acquisition unit 10 The shape information M1 of the waste 200 included in the image information S ′ is acquired (# 20), and the color information acquisition unit 20 acquires the color information M2 of the waste 200 included in the specific area image information S ′ (#). 30), the texture information acquisition unit 30 acquires the texture information M3 of the waste 200 included in the specific area image information S ′ (# 40), and the material determination unit 40 receives the shape information M1, color information M2, and Based on material information M3 The material of the waste 200 contained in the specific area image information S 'ultimately determine (# 50).

  As described above, according to the material discrimination device 100 of the second embodiment and the material discrimination method that is the function thereof, the shape information acquired based on the specific area image information S ′ input to the shape information acquisition unit 10. M1, texture information acquired based on the color information M2 acquired based on the specific area image information S ′ input to the color information acquisition section 20 and the specific area image information S ′ input to the texture information acquisition section 30 Based on the three appearance information M3, the material determination unit 40 determines the material of the waste 200 included in the image P ′ represented by the specific area image information S ′. Compared with material discrimination based only on color and material discrimination based only on color information M2, the discrimination accuracy can be remarkably improved.

  Further, the material discriminating apparatus 100 according to the second embodiment can exhibit the other effects of the material discriminating apparatus 100 according to the first embodiment as well.

  Furthermore, the material discriminating apparatus 100 according to the second embodiment is different from the material discriminating apparatus 100 according to the first embodiment that targets only the image information S representing the image P including only one waste 200. The image information S representing the image P containing the waste 200 can also be targeted, and higher versatility than the material discrimination device 100 of the first embodiment can be realized.

  The material discrimination device 100 of each of the first and second embodiments described above is not limited to the construction waste 200 as a material discrimination target, but is an industrial waste other than construction waste, It goes without saying that waste discharged from households or waste discharged for recycling purposes can also be targeted.

It is a block diagram which shows the structure of the material discrimination | determination apparatus which concerns on one Embodiment (Embodiment 1) of this invention. It is a schematic diagram which shows the construction system waste which is an example of the waste used as the object of material discrimination | determination by the material discrimination | determination apparatus shown in FIG. 1, (a) is a concrete piece, (b) is an aluminum bar piece, (c ) Is a plastic pipe piece, (d) is a reinforcing bar piece, and (d) is a piece of wood. It is a flowchart explaining the effect | action of the material determination apparatus shown in FIG. FIG. 3 is a schematic diagram (part 1) showing a state in which wastes of a plurality of types of materials shown in FIG. 2 are included in one image, (a) is an entire image, and (b) is a part of (a). Each represents an enlarged image. 4A is a schematic diagram expressing the distribution of brightness of the image of FIG. 4B in a matrix coordinate system, and FIG. 4B is a schematic diagram expressed in a matrix coordinate system after Laplacian filter processing is performed. FIG. FIG. 3 is a schematic diagram (part 2) showing a state in which wastes of a plurality of types of materials shown in FIG. 2 are included in one image, (a) is an entire image, and (b) is a part of (a). Each represents an enlarged image. FIG. 6A is a schematic diagram expressing the lightness distribution of the image of FIG. 6B in a matrix coordinate system, and FIG. 6B is a schematic diagram expressed in a matrix coordinate system after Laplacian filter processing. FIG. It is a flowchart explaining the detail of a 1st discrimination | determination process. It is a schematic diagram showing the reference | standard position in the a * b * coordinate system by the a * value corresponding to every material of a waste, and b * value. It is a flowchart explaining the detail of a 2nd discrimination | determination process. It is a graph which shows the standard value (standard standard deviation value) of the standard deviation corresponding to every material of waste. It is a flowchart explaining the detail of a 3rd discrimination | determination process. It is a graph which shows the example which proved the effect by the material discrimination device of Embodiment 1. It is a block diagram which shows the structure of the material discrimination | determination apparatus which concerns on one Embodiment (Embodiment 2) of this invention. It is the simplified flowchart explaining the effect | action of the material determination apparatus shown in FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Shape information acquisition part 20 Color information acquisition part 30 Texture information acquisition part 40 Material discrimination | determination part 100 Material discrimination | determination apparatus 110 Camera 200 Construction waste (waste)
M1 Shape information M2 Color information M3 Texture information

Claims (11)

Based on only the input image information including the color image of the waste, shape information on the shape of the waste, color information on the color of the waste, and texture information on the rough surface of the waste are acquired. ,
The acquired shape information, based on the color information and the texture information, a material discriminating method waste you determine the material of the waste,
The color information is acquired based on color information representing a color of the waste image in the image information,
The texture information is acquired based on brightness difference variation information representing variation in brightness difference of the waste image in the image information,
The color information is defined by a * value and b * value when the image information is represented by CIE color system L * a * b *.
The brightness is defined by an L * value when the image information is represented by a CIE color system L * a * b *, and the brightness difference variation information is relative to a coordinate system defined by the L * value. A waste material discrimination method characterized in that it is defined by performing a Laplacian filter process and calculating a standard deviation in the coordinate system after the Laplacian filter process . Based on the shape information, the material candidate for the waste is determined to obtain a first determination result,
Based on the color information, the material candidate for the waste is determined to obtain a second determination result,
Based on the texture information, the material candidate for the waste is determined to obtain a third determination result,
The material of the waste is specified by performing predetermined weighting on the obtained first discrimination result, the second discrimination result, and the third discrimination result. The method for discriminating the material of waste as described in 1. The waste shape determination method according to claim 1 or 2 , wherein the shape information is obtained by performing an edge extraction process on the image information to detect a contour shape. Among the images represented by the input image information, specify in advance a region including an image of the waste to be identified as the material,
Extract image information corresponding to the specified area as specific area image information,
Any based on the extracted said specific area image information, said shape information, and acquires the color information and the texture information, among the claims 1 to 3, characterized in that to determine the material of the waste Or the waste material discrimination method according to claim 1. The waste material determination method according to any one of claims 1 to 4 , wherein the waste material is construction waste material. A shape information acquisition unit that acquires shape information related to the shape of the waste based on only the input image information including a color image of the waste;
Based on the color image information, a color information acquisition unit that acquires color information related to the color of the waste,
Based on the color image information, a texture information acquisition unit that acquires texture information on the rough surface of the waste,
The material of the waste is determined based on the shape information acquired by the shape information acquisition unit, the color information acquired by the color information acquisition unit, and the texture information acquired by the texture information acquisition unit. a material discrimination system waste with a material discriminating portion,
The color information acquisition unit acquires the color information based on color information representing the color of the waste image in the image information,
The texture information acquisition unit acquires brightness difference variation information representing variations in brightness difference based on brightness information representing brightness of the waste image in the image information, and the texture information is converted into the brightness difference. It is acquired based on variation information,
The color information acquisition unit defines the color information by an a * value and a b * value when the image information is represented by a CIE color system L * a * b *.
The texture information acquisition unit defines the brightness by an L * value when the image information is represented by a CIE color system L * a * b *, and the brightness difference variation information is defined by the L * value. A waste material discrimination apparatus characterized in that a Laplacian filter process is performed on a coordinate system and a standard deviation in the coordinate system after the Laplacian filter process is calculated . The material discrimination unit is
A first determination unit for determining a material candidate for the waste based on the shape information acquired by the shape information acquisition unit and obtaining a first determination result;
A second determination unit for determining a material candidate for the waste based on the color information acquired by the color information acquisition unit and obtaining a second determination result;
A third determination unit for determining a material candidate for the waste based on the texture information acquired by the texture information acquisition unit and obtaining a third determination result;
Predetermined with respect to the first discrimination result obtained by the first discrimination unit, the second discrimination result obtained by the second discrimination unit, and the third discrimination result obtained by the third discrimination unit; The waste material determination apparatus according to claim 6 , further comprising: a material specifying unit that specifies the material of the waste by performing weighting. The waste shape determination apparatus according to claim 6 or 7 , wherein the shape information acquisition unit obtains the shape information by performing an edge extraction process on the image information. An image display unit for displaying an image represented by the input image information;
An area specifying input operation section for inputting an operation for specifying an arbitrary area of the image displayed on the image display section;
A specific area extracting unit that extracts, as specific area image information, image information corresponding to the area specified by the area specifying input operation unit among the input image information;
Based on the specific region image information extracted by the specific region extraction unit, the shape information acquisition unit acquires waste shape information included in the specific region image information, and the color information acquisition unit The color information of the waste included in the specific area image information is acquired, the texture information acquisition unit acquires the texture information of the waste included in the specific area image information, and the material discrimination unit is the specific area image The waste material discrimination apparatus according to any one of claims 6 to 8 , wherein the waste material contained in the information is discriminated. The image processing apparatus further includes a camera that captures an image including the waste and inputs image information including the image of the waste to the shape information acquisition unit, the color information acquisition unit, and the texture information acquisition unit. The waste material discrimination device according to any one of claims 6 to 9 . The waste material discrimination system waste according to any one of claims 6 to 10, characterized in that the construction waste.

Images