Material Discriminating Apparatus utilizing 2-dimensional Photo Sensor

Abstract

A material identification device is provided having a structure in which optical signals with different passbands reach different light-receiving regions of a single image sensor. A filter unit fixed to the front of the image sensor has a plurality of filters with different passbands arranged continuously in a plane along the longitudinal direction along the direction in which the object to be identified is transported. The material of the subject can be identified from the image signals detected in the different light-receiving regions. Depending on additional aspects, an optical signal feature value is calculated from the average value of the image signal in consecutive image frames of a two-dimensional image sensor over a certain range on the time axis and a certain range on the spatial axis, and the material can be determined from this optical signal feature value.

Description

Technology Field

[0001] An apparatus for analyzing materials using optical means, particularly a technique for analyzing solids such as plastics, is disclosed. Background Technology

[0002] Devices for analyzing materials using optical means are known. For example, U.S. Patent No. 6,433,338, granted on August 13, 2002, discloses a device and method for determining the material of a plastic bottle that is wholly or at least partially transparent. This device is equipped with a disc-shaped filter in which a plurality of filters having different spectral characteristics are arranged in a circular manner. Light irradiated from an infrared radiation source passes through the subject and the disc-shaped filter and is detected by a detector; the signal output from the detector is processed by a microprocessor to determine the material based on signature correlation analysis related to a statistical model. A subject, such as a plastic bottle, is moved via a conveyor for identification.

[0003] Such devices utilize a single photodetector to collect characteristic values ​​for optical signals across multiple bands, thereby enabling reliable material identification. However, rotating disc filters suffer from poor durability due to mechanical drive and impose limitations on reducing the device's size. Furthermore, since the detector provides only a limited number of signals, there is a high probability of errors caused by foreign substances. The problem to be solved

[0004] The proposed invention aims to increase the reliability of material identification using optical means.

[0005] Furthermore, another objective of the proposed invention is to provide a material identification device that has improved durability and is robust against impact. means of solving the problem

[0006] According to one aspect of the proposed invention, a material identification device is provided having a structure in which optical signals having different passbands reach different light receiving regions of a single image sensor. A filter unit fixed to the front of the image sensor has a plurality of filters having different passbands arranged continuously in a plane along the longitudinal direction along the direction in which the object to be identified is transported. The material of the subject can be identified from the image signals detected in the different light receiving regions.

[0007] Depending on additional aspects, an optical signal feature value is calculated from the average value of the image signal in consecutive image frames of a two-dimensional image sensor over a certain range on the time axis and a certain range on the spatial axis, and the material can be determined from this optical signal feature value.

[0008] According to an additional aspect, a material identification algorithm is proposed that comprises multiple material identification models, each of which identifies a different material. Each material identification model can be designed based on a statistical analysis of transmission or absorption characteristics for multiple different frequencies. As another example, each material identification model may be an artificial intelligence-based classifier trained on transmission or absorption rate data for different frequencies of multiple objects to be identified having a specific material.

[0009] Depending on additional aspects, the proposed material identification algorithm may discard data acquired from a specific area of ​​the optical sensor where the detected area does not reflect detection characteristics, for example, a plastic bottle where the area is labeled rather than a transparent area. Effects of the invention

[0010] According to the proposed invention, a single two-dimensional image sensor can be spatially divided to acquire data on the transmission or absorption characteristics of a material for multiple different frequencies. The reliability of the data can be enhanced by averaging the two-dimensional sensor data over time and space in each light-receiving area. By employing filters with multiple different passbands fixed to the front of the image sensor, a material identification device without mechanical movement can be implemented, thereby increasing the durability of the device. Furthermore, by arranging multiple light-receiving areas along the direction in which the object to be identified moves, it is possible to acquire reliable sensing data even when foreign substances or labels are attached to the object to be identified. Brief explanation of the drawing

[0011] FIG. 1 illustrates the appearance of a material identification device according to one embodiment. FIG. 2 is a block diagram illustrating the configuration of a material identification device according to one embodiment. FIG. 3 is a block diagram illustrating the configuration of an optical signal separation unit and a material identification unit according to one embodiment. Specific details for implementing the invention

[0012] The foregoing and additional aspects are embodied through embodiments described with reference to the accompanying drawings. It is understood that the components of each embodiment may be combined in various ways within the embodiment or with components of other embodiments, unless otherwise stated or contradicted. Based on the principle that the inventor may appropriately define the concepts of terms to best describe his invention, the terms used in this specification and claims shall be interpreted in a meaning and concept consistent with the description or proposed technical idea. Preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0013] <Fig. 1 Description of the Invention>

[0014] FIG. 1 illustrates the external appearance of a material identification device according to one embodiment. An object to be identified, for example, a plastic bottle (40), moves in front of the material identification device. The object to be identified may be transported, for example, by a conveyor belt (90). As illustrated, the material identification device according to one embodiment includes a filter unit (30), a two-dimensional optical sensor (50), and an information processing device (10). The filter unit (30) has a first bandpass filter that passes near a first short wavelength band, a second bandpass filter that passes near a second short wavelength band, and a third bandpass filter that passes near a third short wavelength band arranged continuously in one direction on a plane. The two-dimensional optical sensor (50) detects light that has passed through the filter unit (30) in a single light receiving area. The information processing device (10) identifies the material from the optical signal output from the two-dimensional optical sensor (50). One or more light sources (70) may be installed in a position facing the filter unit (30) and the two-dimensional light sensor (50). The light source (70) may be, for example, a white light source. The white light emitted from the light source (70) passes through each filter of the filter unit (30), allowing only light with a specific wavelength range to pass through and reach the light receiving surface of the two-dimensional light sensor (50). In the illustrated embodiment, the information processing device (10) may be implemented as, for example, an industrial computer.

[0015] <Description of Claim 1>

[0016] FIG. 2 is a block diagram illustrating the configuration of a material identification device according to one embodiment. As illustrated, the material identification device according to one embodiment includes a filter unit (30), a two-dimensional optical sensor (50), an optical signal separation unit (300), and a material identification unit (500). In one embodiment, the filter unit (30) has a structure in which a first bandpass filter (31) that passes near a first short wavelength band, a second bandpass filter (33) that passes near a second short wavelength band, and a third bandpass filter (35) that passes near a third short wavelength band are arranged continuously in one direction on a plane.

[0017] These first to third short wavelengths can be selected based on the frequency-dependent light absorption or transmission characteristics of the multiple types of plastics to be identified. In order to distinguish PET (polyethylene terephthalate), which has economic value when recycled, from other plastics such as PP (polypropylene) and PC (polycarbonate), it is necessary to focus on the transmission characteristics at specific wavelengths where they have different optical properties, for example, around 3.1 μm, around 3.2 μm, and around 3.7 μm. For example, since PC exhibits high transmission characteristics at 3.2 μm while PP and PET exhibit low transmission characteristics, it is possible to identify and exclude PC based on the transmittance at 3.2 μm. Also, at around 3.1 μm and 3.7 μm, both PP and PET show higher transmittance than PC, but at around 3.1 μm, PP shows about twice the relative transmittance of PET, and at around 3.7 μm, PP shows slightly higher relative transmittance than PET.

[0018] The two-dimensional light sensor (50) detects light that has passed through a filter section in a single light receiving area. For example, the two-dimensional light sensor (50) may be implemented as a CMOS image sensor. In another example, the two-dimensional light sensor (50) may be a camera including an image sensor and an optical system.

[0019] In the illustrated embodiment, the optical signal separation unit (300) and the material identification unit (500) may be implemented as program instructions that are stored in internal memory, for example, in the information processing device (10) of FIG. 1 and loaded and executed by a computing unit, for example, a microprocessor.

[0020] The optical signal separation unit (300) may include a first signal extraction unit (310), a second signal extraction unit (330), and a third signal extraction unit (350). The first signal extraction unit (310) extracts a first image signal detected in an area (51) where light passing through a first bandpass filter (31) is received by a two-dimensional optical sensor (50). The second signal extraction unit (330) extracts a second image signal detected in an area (53) where light passing through a second bandpass filter (33) is received by a two-dimensional optical sensor (50). The third signal extraction unit (350) extracts a third image signal detected in an area (55) where light passing through a second bandpass filter (35) is received by a two-dimensional optical sensor (50).

[0021] The material determination unit (500) determines the material of the subject as one of at least two different materials from the image signals output from the optical signal separation unit (300). In one embodiment, the material determination unit (500) determines PET plastic from PP and PC plastics from the first to third image signals detected in regions (51, 53, 55) where light is received after passing through filters (31, 33, 35) that each have bandpass characteristics for wavelengths around 3.1 μm, around 3.2 μm, and around 3.7 μm.

[0022] <Description of Claim 2>

[0023] Depending on additional aspects, the material identification device may further include a frame sampling unit (100). The frame sampling unit (100) extracts some of the image frames sequentially output from the two-dimensional optical sensor (50) and outputs them to the optical signal separation unit (300). In one embodiment, the frame sampling unit (100) receives a frame synchronization signal and a pixel synchronization signal output from the driving circuit of the two-dimensional optical sensor (50), ignores some of the frame synchronization signals at a constant ratio, takes only some of them, and transmits the image signals output from the two-dimensional optical sensor (50) according to the pixel synchronization signal to the first signal extraction unit (310) to the third signal extraction unit (350).

[0024] <Description of the Invention of Claims 3, 4, and 5>

[0025] According to an additional aspect, a material identification algorithm is proposed that includes a plurality of material identification models, each of which identifies a different material. FIG. 3 is a block diagram illustrating the configuration of an optical signal separation unit and a material identification unit according to one embodiment. As illustrated, the material identification unit (500) may include a first material identification model calculation unit (581) and a second material identification model calculation unit (583). The first material identification model calculation unit (581) calculates a first material identification value by inputting a first optical signal feature value calculated from a first image signal, a second optical signal feature value calculated from a second image signal, and a third optical signal feature value calculated from a third image signal into the first material identification model. The second material identification model calculation unit (583) calculates a second material identification value by inputting a first optical signal feature value calculated from a first image signal, a second optical signal feature value calculated from a second image signal, and a third optical signal feature value calculated from a third image signal into the second material identification model. The material determination unit (590) determines the material of the subject as one of at least two different materials from the first material determination value and the second material determination value.

[0027] According to an additional aspect, each material identification model may be designed based on a statistical analysis of transmission or absorption characteristics for multiple different frequencies. In the illustrated embodiment, a first material identification model is generated based on a statistical analysis of transmission or absorption characteristics for multiple different frequencies of a material having the first material, and a second material identification model may be generated based on a statistical analysis of transmission or absorption characteristics for multiple different frequencies of a material having the second material.

[0028] For example, the first material identification model may be defined as the distance value from the PET average value vector calculated for three-dimensional vectors composed of relative transmittance values ​​at wavelengths of 3.1 μm, 3.2 μm, and 3.7 μm for PET plastic, and the second material identification model may be defined as the distance value from the PE average value vector calculated for three-dimensional vectors composed of relative transmittance values ​​at wavelengths of 3.1 μm, 3.2 μm, and 3.7 μm for PE plastic.

[0029] At this time, the first material determination value output by the first material determination model calculation unit (581) is the distance value between the input data and the PET average value vector, and the second material determination value is the distance value between the input data and the PE average value vector. In this embodiment, the material determination unit (590) compares the first material determination value and the second material determination value and determines the material of the object to be determined as the material corresponding to the smaller value.

[0031] According to another aspect, each material identification model may be an artificial intelligence-based classifier trained with transmittance or absorption rate data for different frequencies of multiple objects to be identified having a specific material. In the illustrated embodiment, the first material identification model may be a first artificial intelligence-based classifier trained with transmittance or absorption characteristic data for multiple different frequencies of a material having the first material, and the second material identification model may be a second artificial intelligence-based classifier trained with transmittance or absorption characteristic data for multiple different frequencies of a material having the second material.

[0032] For example, the first material identification model may be a multilayer neural network model trained with relative transmittance data of PET plastic at wavelengths around 3.1 μm, 3.2 μm, and 3.7 μm, and the second material identification model may be a multilayer neural network model trained with relative transmittance data of PE plastic at wavelengths around 3.1 μm, 3.2 μm, and 3.7 μm.

[0033] At this time, the first material determination value output by the first material determination model calculation unit (581) is a probability value that the material of the object to be determined having input data is PET plastic, and the third material determination value output by the second material determination model calculation unit (583) may be a probability value that the material of the object to be determined having input data is PE plastic. In this embodiment, the material determination unit (590) compares the first material determination value and the second material determination value and determines the material of the object to be determined as the material corresponding to the larger value.

[0034] The illustrated embodiment includes three optical signal feature value calculation units (571, 573, 575) and two material identification model calculation units (581, 583) that receive the three outputs and calculate material identification values ​​for two materials, respectively, but the present invention is not limited thereto, and it is obvious to those skilled in the art who understand the technical concept of the present invention that embodiments having more or fewer optical signal feature value calculation units and / or more or fewer material identification model calculation units can also be derived from the presented embodiment.

[0035] <Description of the Invention of Claims 6 and 7>

[0036] Depending on additional aspects, an optical signal feature value is calculated from the average value of the image signal in consecutive image frames of a two-dimensional image sensor over a certain range on the time axis and a certain range on the spatial axis, and the material can be determined from this optical signal feature value.

[0037] In an embodiment illustrated according to this aspect, the material determination unit (500) may include a first buffer (511) and a first optical signal feature value calculation unit (571), a second buffer (513) and a second optical signal feature value calculation unit (573), and a third buffer (515) and a third optical signal feature value calculation unit (575).

[0038] The first buffer (511) stores the first image signal extracted from the first signal extraction unit (310) frame by frame. As previously described, the first image signal is an image signal corresponding to the light receiving area (51) of the two-dimensional image sensor in FIG. 2. The second buffer (513) stores the second image signal extracted from the second signal extraction unit (330) frame by frame. As previously described, the second image signal is an image signal corresponding to the light receiving area (53) of the two-dimensional image sensor in FIG. 2. The third buffer (515) stores the third image signal extracted from the third signal extraction unit (350) frame by frame. As previously described, the third image signal is an image signal corresponding to the light receiving area (55) of the two-dimensional image sensor in FIG. 2. The first buffer (511) to the third buffer (515) may be a FIFO (First-In-First-Out) buffer that stores video signals extracted from each region of five video frames, for example.

[0039] The first optical signal feature value calculation unit (571) calculates the average image signal of the first image signals stored in the first buffer (511) and outputs it as the first optical signal feature value. The first optical signal feature value calculation unit (571) may include a first inter-frame averaging part (531) and a first intra-frame averaging part (551). The first inter-frame averaging part (531) obtains the first inter-frame average image of the first image signals. The first inter-frame average image is an image in which each pixel has an average value of the pixel values ​​at the corresponding positions of the first image signals stored in the first buffer (511). The first intra-frame average part (551) obtains the intra-frame average pixel value of the first inter-frame average image output by the first inter-frame averaging part (531) and outputs it as the first optical signal feature value.

[0040] Similarly, the second optical signal feature value calculation unit (573) calculates the average image signal of the second image signals stored in the second buffer (513) and outputs it as the second optical signal feature value. The second optical signal feature value calculation unit (573) may include a second inter-frame average unit (533) and a second frame average unit (553). The second inter-frame average unit (533) obtains the second inter-frame average image of the second image signals. The second inter-frame average image is an image in which each pixel has an average value of the pixel values ​​at the corresponding positions of the second image signals stored in the second buffer (513). The second frame average unit (553) obtains the frame-in-frame average pixel value of the second inter-frame average image output by the second inter-frame average unit (533) and outputs it as the second optical signal feature value.

[0041] Similarly, the third optical signal feature value calculation unit (575) calculates the average image signal of the third image signals stored in the third buffer (515) and outputs it as the third optical signal feature value. The third optical signal feature value calculation unit (575) may include a third inter-frame average unit (535) and a third frame average unit (555). The third inter-frame average unit (535) obtains the third frame average image of the third image signals. The third frame average image is an image in which each pixel has an average value of the pixel values ​​at the corresponding positions of the third image signals stored in the third buffer (515). The third frame average unit (555) obtains the frame average pixel value of the third frame average image output by the third frame average unit (535) and outputs it as the third optical signal feature value.

[0042] In FIG. 2, if the filters (31, 33, 35) have uniform characteristics within the region, the light source has the same luminance value on all illuminated surfaces, and the image sensor has the same sensitivity on all light-receiving surfaces, then all pixels included in the image output from the first buffer (511) have the same value, all pixels included in the image output from the second buffer (513) have the same value, and all pixels included in the image output from the third buffer (515) have the same value. However, since this is not the case due to imperfections in the manufacturing process and the characteristics of the filters or image sensors, errors resulting from these imperfections can be improved by calculating the average value.

[0043] <Description of Claim 8>

[0044] Additionally, the average sections between the first and third frames, and the average sections within the first and third frames, can each calculate average values ​​only for a central portion of the image signals output from the first to third buffers. Generally, since filters or lenses may not have uniform characteristics at the edges or may be heavily influenced by ambient light at the edges, peripheral data can be discarded, and an area with uniform characteristics and less influence from ambient light can be selected and processed only for that area.

[0045] <Description of Invention of Claims 9 and 10>

[0046] Referring to FIG. 1, a plastic bottle (40) has a portion (43) with a label attached and a transparent portion (41) that is not. Unlike other portions (41), light transmission characteristic data that can determine the material of the plastic cannot be obtained from the portion (43) with the label attached. In addition, the proposed material determination algorithm may discard data obtained from a portion where the portion detected in a specific area of ​​the optical sensor does not reflect detection characteristics, for example, a portion with a label attached rather than a transparent portion of the plastic bottle.

[0047] According to this aspect, the material identification unit (500) may include a first signal classification unit (521), a second signal classification unit (523), and a third signal classification unit (525). The first signal classification unit (521) determines whether the first image signal is an abnormal signal and controls storage in the first buffer (511) according to the result. The second signal classification unit (523) determines whether the second image signal is an abnormal signal and controls storage in the second buffer (513) according to the result. The third signal classification unit (525) determines whether the third image signal is an abnormal signal and controls storage in the third buffer (513) according to the result. For example, the signal classification units (521, 523, 525) can determine whether the image signal is a signal of the transparent area of ​​the plastic bottle or a signal of the label area.

[0049] Since 2D image sensors typically capture about 30 frames per second, even if the object to be identified moves at a considerably high speed, each area (51, 53, 55) of the image sensor in FIG. 2 can acquire data of the transparent area by avoiding the label for a sufficient number of frames.

[0050] Although the present invention has been described above with reference to embodiments with reference to the accompanying drawings, it is not limited thereto and should be interpreted to encompass various variations that can be obviously derived from them by those skilled in the art. The claims are intended to encompass such variations. Explanation of the symbols

[0051] 10 : Information processing device 30 : Filter section 31,33,35 : Filter 40 : Plastic bottle 50 : 2D image sensor 70 : Light source 90 : Conveyor belt 100 : Frame sampling section 300 : Optical signal separation unit 310: 1st signal extraction unit 330: 2nd signal extraction unit 350 : 3rd signal extraction unit 500 : Material identification unit 511,513,515 : 1st to 3rd buffers 521,523,525 : 1st to 3rd signal classification units 531,533,535 : Average portion between the 1st to 3rd frames 551,553,555 : Average portion within the 1st to 3rd frames 571,573,575 : 1st to 3rd optical signal feature value calculation unit 581 : 1st Material Identification Model Calculation Unit 583 : 2nd Material Identification Model Calculation Unit 590 : Material determination section

Claims

Claim 1 A filter unit comprising a first bandpass filter that passes near a first short wavelength band, a second bandpass filter that passes near a second short wavelength band, and a third bandpass filter that passes near a third short wavelength band, all arranged continuously in one direction on a plane; a two-dimensional optical sensor that detects light passing through the filter unit in a single light receiving area; an optical signal separation unit comprising a first signal extraction unit that extracts a first image signal detected in an area where light passing through the first bandpass filter is received by the two-dimensional optical sensor, a second signal extraction unit that extracts a second image signal detected in an area where light passing through the second bandpass filter is received by the two-dimensional optical sensor, and a third signal extraction unit that extracts a third image signal detected in an area where light passing through the third bandpass filter is received by the two-dimensional optical sensor; and a material determination unit that determines the material of a subject as one of at least two different materials from the image signals output from the optical signal separation unit; wherein the material determination unit comprises: a first optical signal calculated from the first image signal A material determination device using a two-dimensional optical sensor comprising: a first material determination model calculation unit that inputs a feature value, a second optical signal feature value calculated from the second image signal, and a third optical signal feature value calculated from the third image signal into a first material determination model to calculate a first material determination value; a second material determination model calculation unit that inputs a first optical signal feature value calculated from the first image signal, a second optical signal feature value calculated from the second image signal, and a third optical signal feature value calculated from the third image signal into a second material determination model to calculate a second material determination value; and a material determination unit that determines the material of an object as one of at least two different materials from the first material determination value and the second material determination value. Claim 2 The material identification device according to claim 1, wherein the material identification device further comprises: a frame sampling unit that extracts a portion of image frames sequentially output from a two-dimensional optical sensor and outputs them to the optical signal separation unit; a material identification device using a two-dimensional optical sensor. Claim 3 A material identification device using a two-dimensional optical sensor according to claim 1, wherein the first material identification model is generated based on a statistical analysis of the transmission or absorption characteristics of a material having the first material for different multiple frequencies, and the second material identification model is generated based on a statistical analysis of the transmission or absorption characteristics of a material having the second material for different multiple frequencies. Claim 4 A material identification device using a two-dimensional optical sensor according to claim 1, wherein the first material identification model is a first artificial intelligence-based classifier trained with transmission or absorption characteristic data for different multiple frequencies of a material having the first material, and the second material identification model is a second artificial intelligence-based classifier trained with transmission or absorption characteristic data for different multiple frequencies of a material having the second material. Claim 5 The material determination unit of claim 1 further comprises: a first buffer for storing a first image signal per frame; a second buffer for storing a second image signal per frame; a third buffer for storing a third image signal per frame; a first optical signal feature value calculation unit for calculating an average image signal of the first image signals stored in the first buffer and outputting it as a first optical signal feature value; a second optical signal feature value calculation unit for calculating an average image signal of the second image signals stored in the second buffer and outputting it as a second optical signal feature value; and a third optical signal feature value calculation unit for calculating an average image signal of the third image signals stored in the third buffer and outputting it as a third optical signal feature value; a material determination device using a two-dimensional optical sensor. Claim 6 In claim 5, the first optical signal feature value calculation unit comprises a first inter-frame averaging part that calculates a first inter-frame average image of the first image signals and an first intra-frame averaging part that calculates an average pixel value within the first frame of the first inter-frame average image and outputs it as a first optical signal feature value; the second optical signal feature value calculation unit comprises a second inter-frame averaging part that calculates a second inter-frame average image of the second image signals and an second intra-frame averaging part that calculates an average pixel value within the second frame of the second inter-frame average image and outputs it as a second optical signal feature value; and the third optical signal feature value calculation unit comprises a third inter-frame averaging part that calculates a third inter-frame average image of the third image signals and an intra-frame part that calculates an average pixel value within the third frame of the third inter-frame average image and outputs it as a third optical signal feature value. A material identification device using a two-dimensional optical sensor including an intra-frame averaging part. Claim 7 A material identification device using a two-dimensional optical sensor according to claim 6, wherein the first-to-third frame average portion and the first-to-third frame average portion each calculate an average value only for a central portion of the image signals output from the first buffer to the third buffer. Claim 8 A material determination device using a two-dimensional optical sensor according to claim 5, wherein the material determination unit further comprises: a first signal classification unit that determines whether a first image signal is an abnormal signal and controls storage in the first buffer according to the result; a second signal classification unit that determines whether a second image signal is an abnormal signal and controls storage in the second buffer according to the result; and a third signal classification unit that determines whether a third image signal is an abnormal signal and controls storage in the third buffer according to the result. Claim 9 In claim 8, the signal classification units are a material identification device using a two-dimensional optical sensor that determines whether an image signal is a signal of a transparent area of ​​a plastic bottle or a signal of a label area. Claim 10 delete

Images