Analytical method, analytical apparatus, and imaging system

Abstract

To provide an analysis method, analysis apparatus, and imaging system that can efficiently separate and identify multiple detection targets from the background, and that can determine "not found in the database" for objects not present in the database. [Solution] The analysis method according to the present disclosure is an analysis method for separating and identifying N detectable objects and a background in an image containing positional information and spectral information of at least two N detectable objects and a background, wherein the method involves acquiring an image, calculating the standard deviation of reflectance for each spectral field contained in the image, creating a histogram of the standard deviations, determining N to 2 × N standard deviation thresholds from the obtained histogram, dividing the region below the smallest threshold among the N to 2 × N thresholds as the background, and detecting N detectable objects using the N to 2 × N thresholds.

Description

[Technical Field]

[0001] The present disclosure relates to an analysis method, an analysis apparatus, and an imaging system, and particularly relates to an analysis method, an analysis apparatus, and an imaging system for separating and identifying a detection target object and a background included in a spectral spectrum image. [Background Art]

[0002] A spectral spectrum image represented by a hyperspectral image stores one piece of spectral spectrum data per pixel, and the distribution of chemical components such as water and oil can be visualized, for example, by selecting a wavelength used for drawing. In addition, analysis techniques that combine hyperspectral images with multivariate analysis and machine learning have been actively researched as methods for identifying chemical components and quantifying component concentrations.

[0003] Identification of a spectral spectrum of an unknown sample is performed by calculating spectral similarity with a database (library) of spectral spectra of a plurality of known samples acquired in advance, and fitting the result to the known sample having the highest similarity.

[0004] For example, Patent Document 1 employs a correlation coefficient as the spectral similarity, and Patent Document 2 employs Euclidean distance as the spectral similarity. [Prior Art Documents] [Patent Documents]

[0005] [Patent Document 1] Japanese Patent No. 7015579 [Patent Document 2] Japanese Patent No. 7419093 [Summary of the Invention] [Problem to be Solved by the Invention]

[0006] In spectral similarity-based discrimination, the similarity is calculated for each unknown sample, equal to the number of known sample types. Therefore, when applied to spectral images, the spectral similarity calculation needs to be performed a number of times equal to the product of the number of pixels and the number of known sample types. Furthermore, because spectral similarity-based discrimination is a relative evaluation, for example, a sample that does not exist in the database may be incorrectly identified as the known sample with the highest similarity.

[0007] Therefore, the higher the resolution of the spectral image and the greater the variety of known samples, the longer the computation time required for discriminant analysis. Furthermore, it may incorrectly classify samples that do not exist in the database.

[0008] This disclosure aims to solve the above-mentioned problems and to provide an analysis method, analysis apparatus, and imaging system that can efficiently separate and identify multiple detection targets from the background, and that can determine "not found in the database" for items not present in the database. [Means for solving the problem]

[0009] The analysis method according to this disclosure is an analysis method for separating and identifying N objects and a background in an image containing positional information and spectral information of at least two N objects and a background, wherein the method involves acquiring the image, calculating the standard deviation of reflectance for each spectral field included in the image, creating a histogram of the standard deviations, determining N to 2 × N threshold values ​​of the standard deviation from the obtained histogram, dividing the region where the value is less than or equal to the smallest threshold value among the N to 2 × N threshold values ​​as the background, and detecting the N objects using the N to 2 × N threshold values.

[0010] The spectral data used to calculate the standard deviation may include reflectance information in the range of 950 to 1650 nm. After determining the threshold for the standard deviation, the image may be subjected to binarization. The image may be a hyperspectral image containing multiple pixels, each containing a spectral data point.

[0011] The analytical apparatus according to this disclosure is an analytical apparatus for separating and identifying N objects and a background in an image containing positional information and spectral information of at least two N objects and a background, comprising: an acquisition unit for acquiring the image; a calculation unit for calculating the standard deviation of reflectance for each spectral spectrum included in the image and creating a histogram of the standard deviations; a determination unit for determining N to 2 × N threshold values ​​of the standard deviation from the histogram obtained by the calculation unit; and a processing unit for dividing the region where the value is less than or equal to the smallest threshold among the N to 2 × N threshold values ​​as the background and detecting the N objects using the N to 2 × N threshold values.

[0012] The imaging system according to this disclosure comprises a camera that detects electromagnetic waves in the near-infrared wavelength region and visualizes them as a monochrome image, an illumination system equipped with multiple LEDs of different wavelengths, an image processing unit that integrates multiple near-infrared images for each wavelength captured by the camera into a single multiband image and outputs it, and a transmission unit that transmits the output multiband image to the acquisition unit. [Effects of the Invention]

[0013] This disclosure provides an analysis method, an analysis apparatus, and an imaging system that can efficiently separate and identify multiple detection targets from the background, and can determine that objects not present in the database are "not found in the database." [Brief explanation of the drawing]

[0014] [Figure 1] This flowchart shows the analysis method used in this disclosure. [Figure 2]It is a diagram illustrating the analysis method of the present disclosure. [Figure 3A] It is a histogram of the standard deviation of reflectance obtained for each spectral spectrum included in an image in the analysis method of the present disclosure. [Figure 3B] It is an image recognized as a background in the analysis method of the present disclosure. [Figure 3C] It is an image detected as a detection target in the analysis method of the present disclosure. [Figure 4] It is a spectral image including a plurality of detection targets and a background according to the present disclosure. [Figure 5] It is a spectral spectrum at a specific position of the spectral image of the present disclosure. [Figure 6] It is a histogram of the standard deviation of reflectance obtained for each spectral spectrum included in the spectral image of the present disclosure. [Figure 7A] It is an image detected as a detection target in the analysis method of the present disclosure. [Figure 7B] It is an image detected as a detection target in the analysis method of the present disclosure. [Figure 7C] It is an image detected as a detection target in the analysis method of the present disclosure. [Figure 7D] It is an image detected as a detection target in the analysis method of the present disclosure. [Figure 7E] It is an image recognized as a background in the analysis method of the present disclosure. [Figure 8] It is a block diagram showing the analysis apparatus of the present disclosure. [Figure 9] It is a schematic diagram showing the imaging system of the present disclosure. [Figure 10] It is a visible image captured by the imaging system of the present disclosure. [Figure 11] It is a near-infrared image captured by the imaging system of the present disclosure. [Figure 12] It is a histogram of the standard deviation of luminance values at 8 wavelengths included in a near-infrared image in the analysis method of the present disclosure. [Figure 13A]This is an image detected as a target object in the analysis method of this disclosure. [Figure 13B] This is an image detected as a target object in the analysis method of this disclosure. [Figure 13C] This is an image detected as a target object in the analysis method of this disclosure. [Figure 13D] This is an image detected as a target object in the analysis method of this disclosure. [Figure 13E] In the analysis method described herein, this image was recognized as the background. [Figure 14] These are near-infrared images of polypropylene at various wavelengths. [Figure 15] Figure 14 shows the wavelength spectrum of the brightness value at the "+" position in the near-infrared image. [Modes for carrying out the invention]

[0015] The analysis method and analytical apparatus relating to this disclosure will be described below with reference to the drawings. First, the separation and identification of a single object to be detected and the background, which forms the basis of the image processing of this disclosure, will be explained using Figures 1 to 3C and Figure 8. Figure 1 is a flowchart of the analysis method relating to this disclosure. Figure 2 is a diagram illustrating the analysis method relating to this disclosure. Figure 3A is a histogram of the standard deviation of reflectance obtained for each spectral spectrum contained in the image, Figure 3B is an image recognized as the background, and Figure 3C is an image detected as an object to be detected. Figure 8 is a block diagram of the analytical apparatus relating to this disclosure.

[0016] As shown in Figure 8, the analytical apparatus 100 according to this disclosure comprises an acquisition unit 101, a calculation unit 102, a determination unit 103, and a processing unit 104.

[0017] In the analysis method described herein, first, the acquisition unit 101 acquires an image 200 (see Figure 1: S11). The image 200 shown in Figure 2A on the left is, for example, an image of a leaf pie placed on white paper with dimensions of 139 x 224 = 31136 pixels, and is a hyperspectral image containing multiple pixels, each containing a spectral data point. The image 200 includes positional information and spectral data of the object to be detected 201 (leaf pie) and the background 202 (white paper), and is rendered using a reflectance of 1300 nm. Therefore, the image 200 contains 31136 spectral data points.

[0018] The acquisition unit 101 transmits the spectral data of the acquired image 200 to the calculation unit 102. The calculation unit 102 then calculates the standard deviation of reflectance for each spectral data point in the image 200 (see Figure 1: S12) and creates a histogram of the standard deviations (see Figure 1: S13).

[0019] Figure 2, right-hand Figure B, shows the spectral spectra of the object 201 and background 202 at specific locations within image 200, with wavelength (nm) on the horizontal axis and reflectance on the vertical axis. By calculating the standard deviation of the reflectance of the spectral spectra, i.e., the degree of spread (variation) in the vertical axis direction of each spectral spectrum, the object 201 and background 202 can be accurately separated.

[0020] The spectral data shown in Figure 2, right-hand Figure B, was acquired at 5nm intervals between 950 and 1650nm, and therefore contains reflectance information from 141 points. However, the wavelength range and intervals are not limited to these. Furthermore, the standard deviation can be calculated using reflectance information from just three points, making it applicable to RGB images and RGB videos, for example.

[0021] The means and methods for creating histograms are not particularly limited. For example, the number of bars in a histogram may be determined as the square root of the number of data points (k = √n, where k is the number of bars in the histogram and n is the number of data points, i.e., the number of pixels). In the example above, since it contains 31,136 spectral data points, it would be appropriate to create a histogram with approximately 176 bars.

[0022] The calculation unit 102 transmits the obtained histogram to the determination unit 103. The determination unit 103 then determines the threshold for the standard deviation (see Figure 1: S14) and performs binarization processing (see Figure 1: S15).

[0023] After the determination unit 103 determines the threshold and performs binarization, the processing unit 104 separates the detection target object 201 and the background 202 in the image 200 (see Figure 1: S16). The background 202 has fewer irregularities in the shape of its spectral spectrum compared to the detection target object 201, so its standard deviation tends to be smaller. Therefore, as will be described later, by considering the region with a relatively small standard deviation as the background 202, the detection target object 201 and the background 202 in the image 200 can be separated.

[0024] The determination of the threshold, the binarization process, and the separation of the detected object 201 and the background 202 will be explained using Figures 3A to 3C. Figure 3A is a histogram of the standard deviation of reflectance obtained for each spectral spectrum contained in image 200 shown in Figure 2A on the left.

[0025] In the analysis method described herein, a threshold can be determined for the histogram using Otsu's binarization method. Otsu's binarization method assumes that the histogram forms two peaks, one corresponding to the detected object and the other to the background. It defines a class as a collection of the detected object and the background within the image, and sets a threshold so that the separation of the classes is highest. However, the determination of the threshold is not limited to this method; other binarization methods or manual threshold setting may also be used. Otsu's binarization method is described in detail in Otsu, N., “A Threshold Selection Method from Gray-Level Histograms.” IEEE Transactions on Systems, Man, and Cybernetics. Vol.9, No.1, 1979, pp.62-66.

[0026] In this disclosure, a threshold is set using Otsu's binarization method, and the region above the determined threshold is designated as the object to be detected 201, while the region below the threshold is designated as the background 202. The binarization process is performed by replacing the reflectance of the background 202 and the object to be detected 201 with 0 and 1, respectively, but is not limited to this. Alternatively, the reflectance may be replaced with NaN (Not a Number) and 1, or only the reflectance of the background 202 may be replaced with NaN.

[0027] Figure 3B shows the background 202, which is below the threshold (dotted line in Figure 3A) set by Otsu's binarization method, and Figure 3C shows the detected object 201, which is above the threshold. The separation score after binarization was 0.898. In this way, it is possible to efficiently and accurately remove background pixels.

[0028] The analysis method described above calculates the standard deviation of the reflectance of the spectral data, but is not limited to this. For example, instead of the standard deviation of the reflectance of the spectral data, variance, the first principal component score, the interquartile range, or the difference between the maximum and minimum values ​​may be used. Alternatively, absorbance may be used instead of the reflectance of the spectral data, or the luminance values ​​of an RGB image may be used.

[0029] Next, we will describe the separation and identification of multiple detection targets from the background. In this disclosure, "multiple" is defined as "N objects, at least two or more" (where N is an integer greater than or equal to 2).

[0030] Figure 4 shows spectral images of four types of plastic test pieces (objects to be detected) arranged on an anti-reflective material (background), with the reflectances at 950 nm, 1300 nm, and 1650 nm displayed in pseudo-color. This spectral image is a hyperspectral image with 188 x 196 = 36848 pixels, where each pixel stores a spectral value. These spectral values ​​were acquired at 5 nm intervals between 950 nm and 1650 nm, similar to Figure 2, right B.

[0031] Figure 5 shows graphs of the spectral distribution of the four types of plastics that are the target of detection, as well as the anti-reflective background material, at specific positions a to e as shown in Figure 4. In the spectral distribution images, the horizontal axis of this graph represents wavelength (nm) and the vertical axis represents reflectance for the four types of plastic test pieces and the anti-reflective background material at specific positions a to e.

[0032] The raw materials for the four types of plastics are as follows: the region containing specific position a is PS (Polystyrene), the region containing specific position b is Nylon 6 (PA6: Polyamide), the region containing specific position c is Polypropylene (PP: Polypropylene), and the region containing specific position d is Acrylonitrile Butadiene Styrene (ABS: Acrylonitrile Butadiene Styrene).

[0033] In Figure 4, PS (specific position a) and ABS (specific position d) are brightly visible in the image, while PA6 (specific position b) and PP (specific position c) are dark and blend into the background. Furthermore, from the results in Figure 5, the baseline of the spectral curves for PS (specific position a) and ABS (specific position d) is at a high position, while the baseline of the spectral curves for PA6 (specific position b) and PP (specific position c) is at a low position. From these results, it is considered that the difference in appearance of the spectral images changes depending on the baseline of the spectral curve.

[0034] Figure 6 is a histogram of the standard deviation of reflectance obtained for each spectral spectrum in the spectral image of Figure 4. In the histogram shown in Figure 6, peaks appear at standard deviations of 0.001, 0.015, 0.033, 0.095, and 0.105.

[0035] By setting thresholds between each of these peaks, the four types of plastics that are the target objects and the anti-reflective material that forms the background can be separated and identified. In the example shown in Figure 6, five thresholds of 0.007, 0.023, 0.040, 0.088, and 0.098, indicated by the dotted lines, are set to perform separation and identification.

[0036] Figure 7A is a spectral image showing a standard deviation peak between thresholds of 0.088 and 0.098. From this, it can be seen that the region where the PS containing a specific position a is located is clearly separated and identified.

[0037] Figure 7B is a spectral image showing a standard deviation peak between thresholds of 0.023 and 0.040. From this, it can be seen that the region where PA6 is located, including a specific position b, is clearly separated and identified.

[0038] Figure 7C is a spectral image with a standard deviation peak between thresholds of 0.007 and 0.023. From this, it can be seen that the region where the PP containing a specific position c is located is clearly separated and identified.

[0039] Figure 7D is a spectral image showing peaks with a standard deviation of 0.098 or higher. From this, it can be seen that the region where the ABS is located, including the specific position d, is clearly separated and identified.

[0040] Figure 7E is a spectral image with a peak having a standard deviation below a threshold of 0.007. From this, it can be seen that the background (anti-reflective material) containing a specific position e is clearly separated and identified.

[0041] Note that the upper right corner of the area shown in Figures 7A to 7D is partially missing because each plastic test piece has a polyethylene terephthalate (PET) label attached to it.

[0042] In the example shown in Figure 6, the interval between the standard deviation peaks of 0.033 and 0.095 is large, so two thresholds, 0.040 and 0.088, are set between these points. The analytical method of this disclosure is not limited to this. For example, if the number of detected objects is known, one threshold can be set at any point between the standard deviation peaks of 0.033 and 0.095. Alternatively, two thresholds may be set for each detected object, and the peak of the detected object may fall between these two thresholds.

[0043] Therefore, by setting N to 2 × N thresholds, it is possible to separate and identify N objects and the background.

[0044] When using the analytical methods disclosed in Patent Documents 1 and 2, for example, in a spectral image with 36,848 pixels as shown in Figure 4, 36,848 × 4 = 147,392 similarity calculations are required to separate and identify four types of objects. Furthermore, because it is a relative comparison of similarity, pixels of the anti-reflective material in the background may also be mistakenly identified as the plastic with the highest similarity.

[0045] On the other hand, in the analysis method of this disclosure, the standard deviation of reflectance over the entire wavelength range of the spectral image is calculated and used for separation and identification, so there is no need to calculate spectral similarity for each type of known sample. In addition, since the region below the smallest threshold among N to 2 × N thresholds is divided as the background, the background is not misidentified.

[0046] The analysis method described herein classifies samples by absolute comparison of standard deviations. As an example, consider a known sample A with a standard deviation peak of 0.4-0.6 and a known sample B with a standard deviation peak of 0.1-0.3. In this case, an unknown sample C with a standard deviation peak of 0.2 can be classified as known sample B, and an unknown sample D with a standard deviation peak of 0.8 can be easily classified as "not applicable".

[0047] On the other hand, the analytical methods disclosed in Patent Documents 1 and 2 classify samples by comparing their similarity to that of known samples A and B. Therefore, if the similarity between unknown sample C and known sample A and B is 0.99 and 0.98, respectively, the classification must be made based on a difference of only 0.01, requiring precise measurement. Conversely, if the similarity between unknown sample D and known sample A and B is 0.51 and 0.49, respectively, unknown sample D will be classified as known sample A despite its low similarity. While it is possible to set a threshold (for example, similarity = 0.50) below which samples are classified as "not applicable," the method of determining this threshold is arbitrary.

[0048] Furthermore, since the analysis method described herein only involves the calculation of the standard deviation, it can reduce the computational load even in the case of high-resolution and multi-type discriminant analysis.

[0049] Next, an imaging system for performing the analytical apparatus of this disclosure will be described with reference to Figure 9. The imaging system 300 is a system for imaging a sample 304, including the object to be detected 305 and the background 306, at each wavelength using a camera 301 and illumination 302.

[0050] Camera 301 is a monochrome camera that detects electromagnetic waves in the visible to near-infrared wavelength range of 400nm to 1700nm and visualizes them as a monochrome image. Illumination 302 is a multiband illumination system equipped with multiple LEDs of different wavelengths.

[0051] As an example, the illumination 302 of this disclosure is configured to sequentially light up eight LED lights with wavelengths of 850nm, 940nm, 1050nm, 1200nm, 1300nm, 1450nm, 1550nm, and 1650nm.

[0052] The camera 301 images the sample 304 for each wavelength and acquires multiple near-infrared images. The eight acquired images are output as a single integrated multiband image by an image processing unit (not shown). Furthermore, a transmission unit (not shown) may be provided to transmit the output multiband image to the acquisition unit 101 of the analysis device 100. The analysis method described above is applied to the transmitted multiband image by the analysis device 100.

[0053] Hyperspectral cameras, which capture hyperspectral images, are expensive and require several seconds to tens of seconds per image for data acquisition. On the other hand, the imaging system 300 is less expensive than a hyperspectral camera, and data acquisition can be performed in about one-tenth the time of a hyperspectral camera. Therefore, for example, the imaging system 300 of this disclosure can be applied to the inspection line of industrial products.

[0054] Figure 10 is a visible image of sample 304, which includes multiple detection targets 305 and a background 306, captured by an RGB camera. Sample 304 shown in Figure 10 consists of four types of plastics, i.e., detection targets 305, arranged on a background 306 which is an anti-reflective material: polystyrene (PS), nylon 6 (PA6), polypropylene (PP), and acrylonitrile butadiene styrene (ABS).

[0055] Figure 11 is a near-infrared image of a sample 304, including multiple detection targets 305 and a background 306, captured by the imaging system 300. Within this near-infrared image, the ROI (region of interest) enclosed by the dotted line represents the specific area to be analyzed. The black circular pattern visible in the center of the near-infrared image is due to the dome shape of the illumination 302 and the shape of the opening for the camera 301.

[0056] Figure 12 is a histogram of the standard deviations of the luminance values ​​at eight wavelengths included in the near-infrared image of Figure 11. In the histogram shown in Figure 12, peaks appear at standard deviations of 20, 38, 46, 87, and 997.

[0057] By setting thresholds between each of these peaks, the object 305 and the background 306 can be separated and identified. In the example shown in Figure 12, six thresholds, 30, 40, 50, 85, 90, and 95, are set as indicated by the dotted lines to perform separation and identification.

[0058] Figure 13A is a near-infrared image with a peak in standard deviation between thresholds of 30 and 40. From this, it can be seen that the region where polypropylene (PP) is placed is clearly separated and identified.

[0059] Figure 13B is a near-infrared image with a peak in standard deviation between threshold values ​​of 40 and 50. From this, it can be seen that the region where nylon 6 (PA6) is placed is clearly separated and identified.

[0060] Figure 13C is a near-infrared image with a peak in standard deviation between threshold values ​​of 85 and 90. From this, it can be seen that the region where polystyrene (PS) is placed is clearly separated and identified.

[0061] Figure 13D is a near-infrared image with a standard deviation peak above a threshold of 95. From this, it can be seen that the region where acrylonitrile butadiene styrene (ABS) is placed is clearly separated and identified.

[0062] Figure 13E is a near-infrared image with a peak in the standard deviation below a threshold of 30. From this, it can be seen that the background (anti-reflective material) is clearly separated and identified.

[0063] Figure 14 shows how near-infrared images captured at each wavelength are displayed. As an example, Figure 14 shows near-infrared images of polypropylene (PP) at eight wavelengths: 850nm, 940nm, 1050nm, 1200nm, 1300nm, 1450nm, 1550nm, and 1650nm. The black circular pattern in the center of the near-infrared image is due to the dome shape of the illumination 302 and the shape of the opening for the camera 301.

[0064] Figure 15 is a graph of the wavelength spectrum of the luminance values ​​at the "+" position in the near-infrared image shown in Figure 14. The horizontal axis of this graph represents wavelength (nm), and the vertical axis represents the luminance value at the "+" position. Based on the luminance values ​​obtained in this way, the standard deviations of the luminance values ​​at the eight wavelengths shown in Figure 12 are plotted.

[0065] Thus, we can provide an analytical method, analytical apparatus, and imaging system that can efficiently separate and identify multiple detection targets from the background based on spectral information, and that can determine "not found in the database" for those not present in the database. The analytical method of this disclosure is expected to have applications, for example, in the identification of plastic raw materials in recycling plants and in the inspection of foreign matter in agricultural raw materials for processed foods.

[0066] This disclosure is not limited to the embodiments described above, and may be modified as appropriate without departing from its spirit. [Explanation of Symbols]

[0067] 100 Analyzer 101 Acquisition Department 102 Calculation Unit 103 Decision Section 104 Processing Unit 200 images 201 Detected object 202 Background 300 Imaging Systems 301 Camera 302 Lighting 303 Irradiation light 304 samples 305 Detected object 306 Background

Claims

1. An analysis method for separating and identifying N detectable objects and the background in an image containing positional information and spectral information of at least two N detectable objects and the background, The aforementioned image was obtained, The standard deviation of reflectance is calculated for each spectral spectrum contained in the aforementioned image. The aforementioned standard deviation is then plotted as a histogram. From the obtained histogram, N to 2 × N standard deviation thresholds are determined. The region whose value is less than or equal to the smallest threshold among the N to 2 × N thresholds is divided as the background. The N objects to be detected are detected using the N to 2 × N thresholds. Analysis method.

2. The spectral data used to calculate the standard deviation includes reflectance information in the range of 950 to 1650 nm. The analytical method according to claim 1.

3. Furthermore, after determining the threshold for the standard deviation, the image is subjected to a binarization process. The analytical method according to claim 1.

4. The aforementioned image is a hyperspectral image containing multiple pixels, each storing a spectral area. The analytical method according to any one of claims 1 to 3.

5. An analytical device for separating and identifying N detection targets and a background in an image containing positional information and spectral information of at least two N detection targets and a background, The acquisition unit acquires the aforementioned image, A calculation unit calculates the standard deviation of reflectance for each spectral spectrum included in the aforementioned image and generates a histogram of the standard deviations. A determination unit that determines N to 2 × N standard deviation thresholds from the histogram obtained by the calculation unit, The system includes a processing unit that divides the region below the smallest threshold value among the N to 2 × N threshold values ​​into the background, and detects the N objects using the N to 2 × N threshold values. Analyzer.

6. A camera that detects electromagnetic waves in the near-infrared wavelength range and visualizes them as a monochrome image, Lighting equipped with multiple LEDs of different wavelengths, An image processing unit that integrates multiple near-infrared images for each wavelength captured by the aforementioned camera into a single multiband image and outputs it, The device comprises a transmitting unit that transmits the output multiband image to the acquisition unit of the analysis apparatus according to claim 5, Imaging system.

Images