WO2023276853A1 - 画像処理装置、方法及び画像処理システム - Google Patents
画像処理装置、方法及び画像処理システム Download PDFInfo
- Publication number
- WO2023276853A1 WO2023276853A1 PCT/JP2022/025146 JP2022025146W WO2023276853A1 WO 2023276853 A1 WO2023276853 A1 WO 2023276853A1 JP 2022025146 W JP2022025146 W JP 2022025146W WO 2023276853 A1 WO2023276853 A1 WO 2023276853A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- image
- image processing
- feature amount
- feature
- identification
- Prior art date
Links
- 238000012545 processing Methods 0.000 title claims abstract description 143
- 238000000034 method Methods 0.000 title claims description 40
- 238000000605 extraction Methods 0.000 claims abstract description 157
- 239000000284 extract Substances 0.000 claims abstract description 15
- 238000001514 detection method Methods 0.000 claims description 101
- 238000012795 verification Methods 0.000 claims description 41
- 230000011218 segmentation Effects 0.000 claims description 30
- 238000003384 imaging method Methods 0.000 claims description 19
- 238000003672 processing method Methods 0.000 claims description 13
- 230000009467 reduction Effects 0.000 claims description 12
- 238000004458 analytical method Methods 0.000 claims description 8
- 238000004159 blood analysis Methods 0.000 claims description 7
- 239000003153 chemical reaction reagent Substances 0.000 claims 1
- 238000010586 diagram Methods 0.000 description 21
- 238000004364 calculation method Methods 0.000 description 9
- 239000003814 drug Substances 0.000 description 6
- 229940079593 drug Drugs 0.000 description 5
- 238000010801 machine learning Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 3
- 241000894006 Bacteria Species 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 238000001000 micrograph Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 238000012742 biochemical analysis Methods 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 239000000470 constituent Substances 0.000 description 1
- 230000008602 contraction Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007717 exclusion Effects 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 238000003018 immunoassay Methods 0.000 description 1
- 230000000873 masking effect Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/771—Feature selection, e.g. selecting representative features from a multi-dimensional feature space
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Definitions
- the present invention relates to an image processing device, method, and image processing system for identifying an object in an image.
- Japanese Patent Laid-Open No. 2002-200000 describes that “an image with a low resolution is segmented into regions, and a region near the boundary in the processing result of the segmentation is set as a processing target region, and an image with a high resolution is segmented”. .
- the boundary of the region segmentation result for the low-resolution image is a method for improving accuracy using the feature amount extracted from the high-resolution image.
- a high resolution image is required to correctly segment or classify an input image into a particular class.
- abnormalities that cannot be determined only by low-resolution images may be included.
- the class of the entire area may be determined depending on the presence or absence of locally existing characters or patterns. In these cases, it is difficult to divide the region into correct classes using only the low-resolution image, and it is possible to specify the class only by using the features of the high-resolution image.
- an object of the present invention to provide an image processing apparatus, method, and system that perform region segmentation and identification processing with high accuracy and speed by extracting the minimum feature amount necessary for identification at high speed.
- An image processing apparatus disclosed in the present invention to achieve the above object is an image processing apparatus for identifying an object in an image, comprising: an image input unit for receiving an input image; an image generating unit for generating an image of, a feature amount extracting unit for extracting a feature amount from the generated image generated by the image generating unit, and an object in the image using the feature amount output by the feature amount extracting unit an output unit for outputting the identification result output by the identification unit; and based on the feature amount output by the feature amount extraction unit, the image generation unit generates a new generated image as necessary. and a feature quantity map generation unit for generating a feature quantity map indicating extraction conditions for the feature quantity for the newly generated image and outputting it to the feature quantity extraction unit.
- An image processing method disclosed in the present invention is an image processing method for identifying an object in an image, comprising: an image input step of receiving an input image; and an image for feature extraction from the input image.
- An image generation step a feature amount extraction step of extracting a feature amount from the generated image generated by the image generation step, and detecting or identifying an object in the image using the feature amount calculated in the feature amount extraction step. and an output step of outputting the identification result calculated in the identification step, wherein a new generated image is generated as necessary based on the feature amount output by the feature amount extraction step.
- It further includes a feature quantity map generating step, which is performed by the generating step, and which generates a feature quantity map indicating extraction conditions of the feature quantity for the new generated image and reflects it in the processing by the feature quantity extracting step.
- FIG. 1 is a diagram illustrating an example of a hardware configuration of an image processing apparatus according to Embodiment 1;
- FIG. 1 is a diagram showing an example of a functional block diagram of an image processing apparatus according to Embodiment 1;
- FIG. It is a figure which shows an example of a feature-value extraction map.
- It is a figure which shows an example of a pixel unit identification process.
- 4 is a diagram illustrating an example of a processing flow of an image processing method according to Example 1;
- FIG. FIG. 11 is a diagram illustrating an example of a hardware configuration of an image processing system according to Example 3;
- FIG. 4 is a diagram showing an example of input and output of an image processing device in object placement determination;
- FIG. 11 is a diagram showing a display example of an object placement verification result;
- FIG. 11 is a diagram showing a display example of an object placement verification result;
- FIG. 11 is a diagram showing a display example of an object placement verification result;
- FIG. 11 is a diagram illustrating an example of a hardware configuration of an image processing system according to Example 4;
- FIG. 11 is a diagram showing an example of an input image to the image processing apparatus 100 according to the third embodiment;
- FIG. 11 is a diagram showing an example of an input image to the image processing apparatus 100 according to the third embodiment;
- FIG. FIG. 11 is a diagram showing a display example of an object placement verification result using certainty.
- FIG. 10 is a diagram illustrating an example of a hardware configuration of an image processing system according to a second embodiment;
- FIG. FIG. 11 is a diagram showing an example of a region division result according to Example 4;
- FIG. 11 is a diagram showing an example of a region division result according to Example 4. It is a figure which shows an example of the correspondence table
- the image processing apparatus 100 includes an interface section 110 , a calculation section 111 , a memory 112 and a bus 113 , and the interface section 110 , calculation section 111 and memory 112 transmit and receive information via the bus 113 .
- the interface unit 110 is a communication device that transmits and receives signals to and from devices outside the image processing apparatus 100 .
- Devices that communicate with the interface unit 110 include an imaging device 120 such as a camera and a microscope, and a display device 121 such as a monitor and a printer.
- the calculation unit 111 is a device that executes various processes in the image processing device 100, and is, for example, a CPU (Central Processing Unit) or an FPGA (Field-Programmable Gate Array). Functions executed by the calculation unit 111 will be described later with reference to FIG.
- the memory 112 is a device for storing programs executed by the calculation unit 111, parameters, coefficients, processing results, and the like, and may be HDD, SSD, RAM, ROM, flash memory, or the like.
- FIG. 2 is an example of a functional block diagram according to the first embodiment of the image processing apparatus 100. As shown in FIG. Each of these functional units may be implemented by software that operates on the computing unit 111, or may be implemented by dedicated hardware.
- the image processing apparatus 100 includes an image input unit 201, an image generation unit 202, a feature amount extraction unit 203, a feature amount map generation unit 204, an identification unit 205, and an output unit 206 as functional units. Each functional unit will be described below.
- the image input unit 201 receives input images such as visible light images and microscope images input from the interface unit 110 .
- the image generation unit 202 uses the input image to generate an image for feature amount extraction.
- the image generation unit 202 receives a generated image signal from the feature map generation unit 204, which will be described later, the image generation unit 202 determines an image to be generated based on the generated image signal.
- an image is extracted with predetermined settings.
- a feature quantity extraction unit 203 extracts a feature quantity from the image generated by the image generation unit 202 .
- the feature quantity extraction unit 203 when receiving a feature quantity extraction map from the feature quantity map generation unit 204, which will be described later, selects a region in the feature quantity extraction map or a feature quantity type that exceeds a preset threshold value. Features are extracted only for regions and features that have values. If no feature extraction map is input, all types of features are extracted from the entire image, or only a predetermined region and feature are extracted.
- the feature quantity map generation unit 204 uses the feature quantity output by the feature quantity extraction unit 203 to generate a feature quantity extraction map.
- the feature amount extraction map designates the image to be generated next by the image generation unit 202 and the region and the type of feature amount from which the feature amount is to be extracted next by the feature amount extraction unit 203 .
- the identification unit 205 performs identification processing using all or part of the feature amount output by the feature amount extraction unit 203 .
- the identification processing includes image-based identification processing for classifying an input image into predetermined classes, pixel-based identification processing (area division), and processing for detecting the position of an object in a rectangle. Through this identification processing, it is possible to detect the presence of an object and to determine the type of object.
- the output unit 206 outputs the identification result output by the identification unit 205 to the outside of the device.
- the class information and likelihood information for each image may be output in the form of numerical data, or such information may be output in a visually understandable form such as a character string.
- the identification result is pixel-by-pixel
- the class and likelihood information for each pixel may be output in the form of numerical data, or may be output as an image expressing the class in a unique color.
- the identification result is rectangle information
- the class, likelihood, and rectangle information of each rectangle may be output in the form of numerical data. You may output as an image.
- FIG. 5 shows an example of a processing flowchart according to the first embodiment. Each step corresponds to each element in the functional block diagram shown in FIG.
- An image input step 501 receives an input image such as a visible light image or a microscope image input from the interface unit 110 .
- An image generation step 502 uses the input image to generate an image for feature amount extraction. It should be noted that the image generating step 502 determines the image to be generated based on the generated image signal when the generated image signal generated in the later-described feature map generating step 504 is received. When there is no generated image signal input, an image is generated with predetermined settings.
- a feature quantity extraction step 503 extracts a feature quantity from the image generated in the image generation step 502 .
- the feature quantity extraction step 503 when a feature quantity extraction map is received from the feature quantity map generation step 504, which will be described later, determines whether the area in the feature quantity extraction map or the type of feature quantity exceeds a preset threshold value.
- Features are extracted only for regions and features that have values. If no feature extraction map is input, all types of features are extracted from the entire image, or only a predetermined region and feature are extracted.
- a feature quantity map generation step 504 uses the feature quantity calculated in the feature quantity extraction step 503 to generate a feature quantity extraction map.
- the feature quantity extraction map designates the image to be generated next in the image generation step 502 and the type and area of the feature quantity to be extracted next in the feature quantity extraction step 503 .
- identification processing is performed using all or part of the feature amounts output in the feature amount extraction step 503.
- the identification processing includes image-based identification processing for classifying an input image into predetermined classes, pixel-based identification processing (area division), detection processing for specifying the position of an object by a rectangle, and the like.
- the output step 506 outputs the identification result calculated in the identification step 505 to the outside of the device.
- the class information and likelihood information for each image may be output in the form of numerical data, or such information may be output in a visually understandable form such as a character string.
- the identification result is pixel-by-pixel
- the class and likelihood information for each pixel may be output in the form of numerical data, or may be output as an image expressing the class in a unique color.
- the identification result is rectangle information
- the class, likelihood, and rectangle information of each rectangle may be output in the form of numerical data. You may output as an image.
- the image generation unit 202 uses the input image to generate an image for feature amount extraction.
- the operation of the image generator differs depending on whether or not the generated image signal is received from the feature quantity map generator 204, which will be described later.
- an image is generated based on the generated image signal.
- the specified image is an image with a specific frequency component using a low-pass filter, high-pass filter, band-pass filter, etc., an image that has been subjected to image processing such as correction, enhancement, noise removal, etc., and a feature amount extraction map that will be described later.
- An image or the like that has undergone masking processing may be used as a candidate.
- an image obtained by combining these image processes may be used as a candidate. Note that the types of candidate images are determined in advance.
- the image generation unit 202 sets an image with a predetermined initial setting. For example, an image is generated by reducing the input image with a predetermined reduction ratio and image size.
- the initial setting of the generated image may represent one of the above candidate images.
- the feature amount extraction unit 203 extracts feature amounts from the generated image output by the image generation unit 202 .
- the features to be extracted may be handcrafted features such as HOG or Haar-Like features, or may be automatically obtained by machine learning such as deep learning or random forest. Also, the feature amount extracted based on the generated image signal and the parameter of the feature amount may be changed.
- the feature amount extraction unit 203 When the feature amount extraction unit 203 receives a feature amount extraction map from the feature amount map generation unit 204, which will be described later, the feature amount extraction unit 203 limits the area from which the feature amount is extracted and the type of the feature amount based on the feature amount extraction map.
- the feature quantity extraction map is an array corresponding to the type of feature quantity and region, and when the value in the feature quantity extraction map is higher than a preset threshold value, feature extraction is performed for the corresponding feature quantity and region.
- An image 300 is a feature extraction target image input from the image generation unit 202, and feature quantity extraction maps 301, 302, and 303 are feature quantity extraction maps for feature quantities A, B, and C, respectively.
- White areas in the feature amount extraction maps 301, 302, and 303 represent areas where the values in the feature amount extraction maps are equal to or greater than a preset threshold, and black represent areas where the values are less than the threshold.
- a target object 304 is an object to be detected, and a bottle is exemplified here. Assume that the target object 304 is composed of a bottle body, a cap, and a label. Regions 305 and 306 are regions having values greater than or equal to the threshold within the feature extraction map.
- the feature quantity extraction map 301 only the region 305 is equal to or greater than the threshold, so the feature quantity A is extracted only from the region corresponding to the region 305 in the generated image. That is, the feature amount A is applied only to the area around the bottle, and is not applied to the background or floor area. Also, since only the area 306 in the feature amount extraction map 302 is equal to or greater than the threshold, the feature amount B is extracted only from the area corresponding to the area 306 in the generated image. That is, the feature quantity B is applied only to the labeled area and not applied to other areas. Further, since all areas of the feature quantity extraction map 303 are less than the threshold value, feature extraction processing is not executed. Since this feature extraction map is determined based on the feature amount extracted from the previous generated image, the area validated by the feature amount changes.
- the feature quantity extraction map may specify a common feature quantity map for all feature quantities. That is, it is a map showing only the area for feature extraction. For example, when the feature quantity extraction map 301 in FIG. 3 is input as a common feature quantity map for all feature quantities, the feature quantities A, B, and C are extracted for the region 305, and for the other regions Do not perform feature value extraction.
- the feature quantity extraction map may specify only the feature quantity to be extracted without specifying the region to be extracted.
- the feature extraction map has three arrays corresponding to the feature amounts A, B, and C, and as a result of the feature amount map calculation, only the feature amounts exceeding the threshold value are extracted, and the other feature amounts are not extracted. But it's okay.
- the feature quantity extraction unit 203 reduces the feature extraction processing by limiting the region from which the feature quantity is extracted and the type of the feature quantity based on the feature quantity extraction map, thereby enabling the identification unit 205 to perform high-speed identification processing. come true.
- the feature amount map generation unit 204 generates a generated image signal that specifies the type of image generated by the image generation unit 202, and a feature amount used by the feature amount extraction unit 203 to limit the area from which the feature amount is extracted and the type of the feature amount. Generate volume extraction maps.
- the generated image signal determines an image to be generated next by the image generation unit 202 based on the feature amount output by the feature amount extraction unit 203 .
- the feature amount extraction unit 203 calculates the amount of frequency components contained in the generated image as the feature amount, and determines the image to be generated based on the distribution of the frequency components. Alternatively, it may be determined by machine learning or the like. For example, N candidates for the image to be generated next are prepared, the feature amount output from the feature amount extraction unit 203 is input, and the number of output units is N, to design a deep learning network.
- a selector that selects the next generated image can be generated by learning so that a generated image that extracts a feature amount effective for discrimination is selected based on the input feature amount.
- the feature quantity map generation unit 204 outputs an extraction end signal indicating the end of the feature quantity extraction process to the identification unit 205 .
- the end condition of the feature amount extraction process is determined in advance by the user, and is determined based on, for example, the certainty of the provisional identification, the number of generated images, the number of extracted feature amounts, the processing time, and the like.
- the end condition based on certainty may be a case where the certainty for the entire area is above a certain level, or a case where the average of the certainty is above a certain level.
- image-by-image identification there are cases such as when the certainty of identification output for one image exceeds a certain level, or when the identification score exceeds a certain level.
- a method may be used in which upper limits are set for the number of images to be generated, the number of extracted feature values, the processing time, and the like, and the process ends when any or all of the upper limits are reached. Further, the end may be determined comprehensively based on the number of generated images, the number of extracted feature values, the processing time, and the certainty of identification.
- the identification unit 205 receives the extraction end signal output from the feature amount map generation unit 204, performs final identification using all or part of the feature amounts extracted by the feature amount extraction unit 203, Output the identification result.
- FIG. 4 shows an example of using a plurality of reduced images as generated images.
- Generated images 400, 401 and 402 are a 1/4 reduced image, a 1/2 reduced image and a non-reduced image, respectively.
- Feature amounts 410, 411, and 412 are feature amounts extracted by the feature amount extraction unit 203 from the generated images 400, 401, and 402, respectively.
- the original image unit feature amounts 420 and 421 are the feature amounts obtained by enlarging the feature amounts 410 and 411 four times and two times, respectively.
- Generated image signals 430 and 431 are generated image signals output by the feature map generation unit 204 .
- Feature quantity extraction maps 440 and 441 are feature quantity extraction maps output by the feature quantity map generation unit 204 .
- a combined feature amount 450 is a feature amount obtained by combining the feature amount 412 and the original image unit feature amounts 420 and 421 .
- a discrimination result 460 is an example of a pixel-by-pixel discrimination result obtained by inputting the combined feature amount 450 to the discrimination unit 205 .
- the generated images 400, 401, and 402 have vertical y, horizontal x, and channel ch dimensions.
- the feature amounts 410, 411, 412, the original image unit feature amounts 420, 421, the feature amount extraction maps 440, 441, and the combined feature amount 450 have dimensions of vertical direction y, horizontal direction x, and feature amount type ch.
- the identification result 460 has dimensions of vertical direction y, horizontal direction x, and class ch.
- the feature quantity extraction unit 203 first receives the generated image 400 and outputs the feature quantity 410 .
- the generated image 400 is an image generated from an input image when the image generation unit 202 does not receive an input of a generated image signal, and in the example of FIG. 4, a 1/4 reduced image is generated.
- the feature quantity map generation unit 204 receives the feature quantity 410 and outputs a generated image signal 430 and a feature quantity extraction map 440 .
- the generated image signal 430 is a signal for determining the image to be generated next, which the feature map generation unit 204 outputs to the image generation unit 202 .
- the image reduction ratio is selected as the generated image signal.
- Circles in the generated image signals 430 and 431 represent selectable image reduction ratios, and black circles represent image reduction ratios selected by the feature map generation unit 204 . It is assumed that a reduction ratio of 1/2 is selected for the generated image signal 430 , and the image generation unit 202 outputs the generated image 401 (1/2 reduced image) based on the generated image signal 430 .
- the feature amount extraction map 440 indicates a region from which the feature amount extraction unit 203 should extract the feature amount from the generated image 401 . In the example of FIG. 4, the feature amount is extracted only from the white area in the feature amount extraction map 440 .
- the feature amount extraction unit 203 receives the generated image 401 and the feature amount extraction map 440 and outputs the feature amount 411 .
- the feature quantity 411 is extracted only from the region where the feature quantity extraction map 440 is white, and is not extracted from other regions.
- a predetermined value such as 0, for example, is stored in an area outside the feature amount extraction target in the feature amount 411 .
- the feature quantity map generation unit 204 receives the feature quantity 411 and outputs a generated image signal 431 and a feature quantity extraction map 441 .
- the image generator 202 outputs a generated image 402 (non-reduced image) based on the generated image signal 431 .
- the feature amount extraction map 441 indicates an area from which the feature amount extraction unit 203 should extract the feature amount from the generated image 402 .
- the feature amount extraction unit 203 receives the generated image 402 and the feature amount extraction map 441 and outputs the feature amount 412 .
- the feature amount 412 is extracted only from the area shown in white in the feature amount extraction map 441, and is not extracted from other areas.
- the feature amounts 410 and 411 are each reduced by the same magnification as the generated image from which the feature is extracted, they cannot be assigned as the feature amount to each pixel as they are. Therefore, the feature amounts 410 and 411 are converted into a form that can be assigned to each pixel of the original image, such as the original image unit feature amounts 420 and 421 . For example, in the example of FIG. 4, the feature quantities 410 and 411 are enlarged four times and two times. After that, by combining all the feature quantities like the post-combination feature quantity 450, a comprehensive feature quantity corresponding to each pixel can be obtained. Finally, the identification unit 205 receives the combined feature amount 450 and outputs the identification result 460 .
- the identification result 460 represents class information for each pixel.
- the method of creating the original image unit feature amounts 420 and 421 described above is an example, and other methods may be used as long as each feature amount can be assigned to each pixel in the original image. For example, a feature amount obtained by enlarging the feature amount 410 twice and a feature amount 411 are combined, a new feature amount is generated based on the combined feature amount, and then the feature amount is further doubled. A method of assigning is also acceptable.
- the step of assigning feature values to pixels may be omitted, and identification may be performed using all obtained feature values. Quantity may be used for identification.
- the output unit 206 outputs the identification result output by the identification unit 205 to the outside of the device.
- the array and numerical values of the identification results may be output as they are, or in the case of area division, an image in which each pixel is colored according to the identification results may be output.
- the identification result may be output in a form that is easy for the user to understand, such as a character string.
- Embodiment 2 is an image processing system that detects an object at high speed based on the region segmentation result or identification result output by the image processing apparatus according to Embodiment 1 and a detection-related designation signal that designates settings and information regarding detection. be.
- An image processing system 1200 includes an imaging device 1201 , an image processing device 100 , an object detection device 1202 , a display device 1203 and a storage device 1204 .
- the imaging device 1201 is a device for imaging a target object, such as a camera.
- the image processing apparatus 100 is the image processing apparatus described in the first embodiment, and calculates area segmentation and/or identification results of a specific object area from the image captured by the image capturing apparatus 1201 .
- the object detection device 1202 is based on the region segmentation result or identification result output by the image processing device 100 and the detection target designation signal, which is a signal designating the detection target among the detection-related designation signals output by the storage device 1204. Outputs object detection information.
- a display device 1203 presents the result of object detection by the object detection device 1202 to the user.
- the storage device 1204 holds detection-related designation signals preset by the user.
- the detection-related designation signal is a signal for designating various settings and information related to detection.
- a detection-related designation signal in the second embodiment includes a detection target designation signal that designates a detection target in the object detection device 1202 .
- Storage device 1204 outputs a detection target designation signal to object detection device 1202 .
- the image processing system 1200 may include a user interface device such as a keyboard, and may be configured so that the user can rewrite detection-related designation signals recorded in the storage device 1204 during system operation. good.
- FIG. 7 shows an example of an image captured by the imaging device 1201 and an example of the area segmentation result output by the image processing device 100 .
- An image 701 is an example of an image obtained by imaging a plurality of medicine bottles
- an area segmentation result 702 is an example of a result of segmentation of the image 701 by the image processing apparatus 100 .
- Bottles 703, 704 and 705 are bottles of drugs A, B and C, respectively
- bottle regions 706, 707 and 708 are bottle regions of drugs A, B and C obtained by region division.
- the bottle areas 706, 707, 708 are divided into different classes, and the difference in texture indicates the difference in class.
- the region division result 702 is effective for the user to visually confirm the situation. However, if the user needs per-object information (e.g. whether there is a particular type of bottle, how many bottles are in the image, and in what order the bottles are arranged) , requires additional processing. Therefore, the object detection device 1202 receives the detection target designation signal from the storage device 1204, extracts object information based on the detection target designation signal, and outputs it as a detection result.
- per-object information e.g. whether there is a particular type of bottle, how many bottles are in the image, and in what order the bottles are arranged
- the object detection device 1202 can output information determined by the presence or absence of objects belonging to each class, the number of objects, coordinate information, location information, or a combination thereof as a detection result. Designated by the detection target designation signal.
- a method for extracting each piece of information will be described.
- the presence or absence of an object belonging to each class can be detected by extracting the combined regions within the segmentation result and determining whether or not there is a combined region belonging to each class.
- the result of the segmentation may include misidentified regions
- the merged regions may be extracted after excluding the misidentified regions based on the state of the merged regions.
- a threshold for the size of the combined area is set in advance, and if the number of pixels in the combined area is less than the threshold, it is determined as an erroneously identified area and excluded.
- it may be excluded using geometric transformation (contraction/expansion processing, etc.) of the image, noise removal processing such as median filter, etc. It may be excluded by a method of performing threshold processing on the quantitative value.
- the object detection device 1202 can calculate the number of objects belonging to each class by extracting the joined regions belonging to each class in the result of segmentation by the method described above and counting the number of joined regions of each class. be.
- the object detection device 1202 extracts the joint area of each class by the above-described method, and calculates the representative points of the joint areas (for example, the upper end, the lower end, the left end, the right end, the center of gravity, etc.), and the coordinate positions of the representative points. can be output as the detection result.
- the representative points of the joint areas for example, the upper end, the lower end, the left end, the right end, the center of gravity, etc.
- the object detection device 1202 can output the arrangement position of each combined area in the area segmentation result as a detection result.
- the coordinate information of the representative point is extracted from each connected area in the result of area division by the method described above, and is recorded in association with the class information of each connected area.
- the images are sorted based on the horizontal (or vertical) coordinate information, and the class information is extracted from the sorted result to obtain the horizontal (or vertical) arrangement order of the objects in the image as the arrangement information.
- the image may be divided into grids, and grid-shaped object placement information may be obtained by determining the presence or absence of an object in each class in the grid based on the occupied area of the combined area in each grid. .
- the overall state of one or a plurality of objects may be detected by combining the presence/absence of objects belonging to each class, the number of objects, coordinate information, and location information. As described above, it is possible to provide an image processing system that detects a target object in an image at high speed.
- Example 3 is an image processing system that uses the image processing system described in Example 2 to determine whether or not the arrangement of target objects has a predetermined positional relationship.
- FIG. 6 shows a hardware configuration diagram according to the third embodiment.
- An image processing system 600 according to the second embodiment includes an imaging device 1201 , an image processing device 100 , an object detection device 1202 , an object placement verification device 602 , a display device 1203 and a storage device 1204 .
- the imaging device 1201, the image processing device 100, the object detection device 1202, and the display device 1203 are the same as the devices described in the second embodiment, so descriptions thereof are omitted.
- the object detection device 1202 receives from the storage device 1204 a detection target designation signal designating a detection target.
- the detection target designating signal designates arrangement information of each combined region in the result of segmentation.
- the object detection device 1202 extracts object location information by the method described in the second embodiment.
- the detection-related designation signals held in the storage device 1204 include two detection target designation signals that designate detection targets and correct placement state designation signals that represent the correct placement state of the target object.
- the storage device 1204 outputs a detection target designation signal to the object detection device 1202 and outputs a correct placement state designation signal to the object placement verification device 602 . It is assumed that the detection target designation signal and the normal arrangement state designation signal are set in advance by the user.
- the object placement verification device 602 receives the object placement information output by the object detection device 1202 and the correct placement state designation signal output by the recording device 1204, and based on whether or not the two match, the imaging device 1201 Verify that the objects in the captured image are positioned correctly.
- the object detection device 1202 acquires the horizontal arrangement information of each object area from the area segmentation result 702 .
- the label arrangement information is calculated as "ABC"
- the object placement verification device 602 creates an image for presenting the user with the correct/wrong judgment result of the placement.
- 8A, 8B, and 8C show display examples of object placement verification results.
- a window 801 represents a result display window presented to the user
- a comprehensive verification result 802 represents a comprehensive placement verification result
- individual verification results 803, 804, and 805 represent placement verification results of each target object.
- the comprehensive verification result 802 is displayed as "OK" when all the target objects are correctly arranged, and is displayed as "Error" otherwise.
- each bottle is correctly placed, so the individual verification results 803, 804, 805 and the comprehensive verification result 802 are all displayed as "OK”.
- the placement of the second and third bottles from the left is incorrect, so the individual verification results 804 and 805 are displayed as "Error", and the overall verification result 802 is also "Error”.
- the example of FIG. 8C is an example in which the bottle on the right end was different from the expected bottle and could not be detected correctly. Since the placement of the left and middle bottles is correct, the individual verification results 803 and 804 are displayed as "OK”. However, the calculation result of the arrangement information is "AB”, which does not match the registered correct bottle arrangement "ABC”. Therefore, "Error” is displayed in the comprehensive verification result 802 . Instead of displaying "Error”, it is also possible to display a message indicating that the number of target objects to be detected was not detected, or display the number of detected target objects.
- the display of "OK” and "Error” indicates whether the placement of each object is correct or incorrect.
- the correctness or incorrectness of the arrangement of the area, the type of the target object, etc. may be indicated not only by characters, but also by the color of the detection frame or area, the line type, the icon, etc., as long as the display can convey the arrangement information of the target object to the user. Any display is acceptable.
- the arrangement from the left has been described above as an example, up, down, left, right, nested structures, etc. may be determined as long as the relative arrangement of the target objects can be defined.
- the object placement verification device 602 may also use the reliability of the segmentation result output by the image processing device 100 to determine the individual verification result of each bottle. Confidence is a value that indicates how confident the image processing apparatus 100 has segmented each pixel. For example, when obtaining identification scores for the background, bottle A, bottle B, and bottle C for each pixel by region division, there is a method of using the maximum value of the identification scores as the degree of certainty. At this time, it is desirable to normalize so that the value range of each discrimination score is from 0.0 to 1.0 and the sum of all classes of discrimination scores is 1.0. For example, the object placement verification device 602 determines that bottle regions that have not been detected with sufficient confidence are determined to be insufficient confidence or false detection, and displayed as individual verification results. Also, the degree of certainty may be presented to the user by superimposing it on the object placement verification result screen.
- FIG. 11 shows an example of object placement verification using certainty.
- a result display window 1101 is a result display window presented to the user
- an area division result 1102 is an area division result by the image processing apparatus 100
- a certainty superimposition result 1113 is the certainty of the area division result output by the image processing apparatus 100 for each pixel.
- 11 shows an example in which degrees are superimposed on a result display window 1101.
- FIG. A result display window 1101 includes bottle A 1103, bottle B 1104, and bottle C 1105, and displays individual verification results 1106, 1107, 1108 and overall verification result 1109 of each bottle.
- the bottle C1105 does not have a label with the name of the drug written on it, and the name of the drug cannot be visually recognized.
- Region division result 1102 shows region division result A1110, region division result B1111, and region division result C1112 as region division results for bottle A1103, bottle B1104, and bottle C1105. It is assumed that bottle A 1103, bottle B 1104, and bottle C 1105 are correctly identified and their types are correctly identified.
- the certainty factor superimposition result 1113 displays the certainty factors A1113, B1114, and C1115 as certainty factors of the segmentation result A1110, the segmentation result B1111, and the segmentation result C1112 in a form superimposed on the result display window 1101. .
- the certainty factor for each pixel is displayed as a luminance value, and the closer to white the higher the certainty factor, and the closer to black the lower the certainty factor.
- the image processing apparatus 100 extracts features from a plurality of generated images based on the feature extraction map, so this feature extraction method is reflected in the degree of certainty.
- the confidence factor superimposition result 1113 in FIG. 11 shows an example of feature extraction from an image reduced by a plurality of magnifications, and the resolution of the confidence factor is locally different.
- high-resolution confidence values can be obtained for character areas in labels and bottle contours because features are extracted at high resolution. ing.
- the image processing apparatus 100 correctly identifies the bottle types of bottle A1103, bottle B1104, and bottle C1105 as described above, there is a difference in the degree of certainty.
- a threshold for the certainty is determined in advance, and individual A method for determining the determination result will be described. Confidence levels A1113 and B1114 for bottles A1103 and B1104 whose labels are visually recognizable exceed the above thresholds, especially around the labels. Assume that the certainty factor C1115 does not exceed the threshold value. In this case, bottle A 1103 and bottle B 1104 are determined to have sufficient certainty, and the bottle type obtained by region division is adopted as the detection result as it is.
- the user can determine based on which area the image processing apparatus 100 has identified the bottle type, or use the information for identification. It is possible to obtain information about the lack of characteristics of
- Example 4 is an image processing system for automatic blood analysis that uses the image processing system described in Example 2 and includes a function for quickly and highly accurately determining the type of container storing a sample to be analyzed.
- the blood analysis includes biochemical analysis and immunoassay.
- a specimen state judgment function for example, detection of the presence or absence of air bubbles or foreign matter that affects analysis
- FIG. 9 shows a hardware configuration diagram according to the fourth embodiment.
- An image processing system 900 includes an imaging device 1201, an image processing device 100, an object detection device 1202, a specimen state determination device 902, a sampling device 903, an analysis device 904, and a display device 1203. , and a storage device 1204 . Since the imaging device 1201 and the display device 1203 are the same as the devices described in the second embodiment, their description is omitted. However, it is assumed that the imaging device 1201 captures an image of a container in which a sample is stored (hereinafter referred to as a sample) from the side.
- the image processing apparatus 100 performs area division for identifying the container type, and the object detection apparatus 1202 determines and outputs the container type based on the area division result. Details will be described later.
- the detection-related designation signals held in the storage device 1204 include a detection target designation signal that designates a detection target, a sample and/or container state determination target that designates a state determination target in the specimen state determination device 902, and a detection target designation signal that specifies a detection target. Contains two of the designated signals.
- the storage device 1204 outputs a detection target designation signal to the object detection device 1202 and outputs a sample and/or container state determination target designation signal to the specimen state determination device 902 .
- the detection target designation signal the type information of the container determined based on the presence or absence of the object belonging to each class and its coordinate position is designated.
- the user designates the state of the sample and/or container that the user wants to determine, such as the presence or absence of air bubbles or foreign matter in the sample, or the tilt of the container. It is assumed that the detection target designation signal and the sample state and/or container state determination target designation signal are set in advance by the user.
- FIGS. 10A and 10B Examples of input images to the image processing apparatus 100 according to the fourth embodiment are shown in FIGS. 10A and 10B.
- An input image 1000 in FIGS. 10A and 10B shows an example in which a cup A 1003 and a cup B 1004 are placed on bottom-raising containers 1001 and 1002, respectively.
- a sample 1005 is stored in the cup A1003 and the cup B1004.
- the bottom-raised containers 1001 and 1002 and the cup A 1003 and cup B 1004 are separated and may be combined in any way.
- the container type is determined based on the combination of the cup type and the height of the bottom-up container will be described.
- a feature amount extraction map is generated from a reduced image, and necessary for identifying the type of cup and detecting the position of the cup only from around the cup A 1003 and the cup B 1004. Since the feature quantity is extracted, the speed can be increased.
- FIGS. 13A and 13B are examples of segmentation results when FIGS. 10A and 10B are used as input images.
- the region segmentation result 1300 includes a cup region A 1303 and a cup region B 1304, each of which is expressed as a texture of a different type of cup.
- the region segmentation result by the image processing device 100 is input to the object detection device 1202 .
- Object detection device 1202 also receives a detection target designation signal from storage device 1204 .
- the detection target designation signal designates the presence or absence of an object belonging to each class (that is, the type of cup present in the image and the presence or absence thereof) and the type information of the container determined based on the coordinate position thereof.
- the object detection device 1202 first extracts a combined area from the area segmentation result 1300 by the method described in the second embodiment, and identifies the type of cup in the image. After that, the object detection device 1202 estimates the length of the bottom-up container from the position of the extracted joint area.
- d2 is the vertical coordinate position of the lower end of the raised bottom container
- 1 is the vertical coordinate position of the lower end of the cup region estimated by the image processing apparatus 100
- d1 is the distance from the lower end of the cup to the bonding point with the raised floor container.
- is the distance of FIG. 15 illustrates the relationship of each parameter in the example of FIG. 10A.
- the distance d1 from the lower end of the cup to the bonding point with the raised container is measured in advance for each type of container and recorded in the storage device 1204. Read from storage device 1204 . It is assumed that the coordinate position d2 in the vertical direction of the lower end of the raised-bottom container is the same position regardless of the raised-bottom container, and is measured and recorded in the storage device 1204 in advance. However, if the lower end of the raised bottom container fluctuates significantly, the image processing apparatus 100 may estimate the position of the lower end of the raised bottom container separately from the cup region.
- the container type is determined from the specified cup type and the length of the bottom-up container.
- a correspondence table 1400 of the cup type and the length of the bottom-raised container as shown in FIG. 14 is stored in the storage device 1204 in advance.
- the horizontal axis of the correspondence table 1400 indicates the type of cup, and the vertical axis indicates the height of the bottom-up container. It is possible to determine Regarding the length of the bottom-raising container, the element having the closest value from the candidates in the correspondence table 1400 to the estimated value obtained by the above formula is selected.
- the sample state determination device 902 is a device that determines the state of the sample and/or container, and uses any existing technology to determine whether there are bubbles in the sample, whether the container is tilted, or the like. At this time, by switching the determination algorithm based on the container type output by the object detection device 1202, it becomes possible to determine the specimen state with higher accuracy.
- the sample collection device 903 controls sample collection processing based on the sample state determination result output by the sample state determination device 902 .
- the analysis device 904 performs blood analysis by mixing the sample obtained by the sampling device 903 with the drug.
- the output device 905 is a device for displaying the blood analysis results, such as a monitor or a printer.
- the image processing device 100 and the object detection device 1202 can identify the sample container type at high speed and with high accuracy.
- the sample collection device 903 controls sample collection based on the sample state determined by the sample state determination device 902 . For example, if the sample contains air bubbles that affect the analysis, the user is notified of the reason for non-collection without collecting the sample, or the sample is sent to another sample transport route.
- the generated image signal may be calculated so as to generate a plurality of images according to the score of the generated image signal.
- the image generation unit 202 that has received the generated image signal generates a plurality of images based on the generated image signal and outputs the images to the feature amount extraction unit 203 .
- the feature quantity map generating unit 204 generates a feature quantity extraction map for each generated image.
- the feature amount extraction unit 203 uses the generated image group received from the image generation unit 202 and the feature amount extraction map group received from the feature amount map generation unit 204 to extract from each pair (a combination of the generated image and the feature amount extraction map) Extract features.
- the types of feature amounts selected by the feature amount extraction unit 203 of the first embodiment include not only differences in characteristics of feature amounts such as differences in weighting coefficients of feature amount extraction filters, but also strides and dirates in convolution processing. It may be a parameter related to the application method.
- a certainty degree display button is arranged on the object placement verification result screen, and the user switches between ON/OFF of superimposing the certainty degree and whether to display the original image or the certainty degree. method is fine.
- the degree of certainty is expressed as a brightness value
- a heat map, a contour line, or a color-coded display for each class may be used.
- only pixels having certainty or more may be superimposed and displayed.
- the condition of the bottle or the photograph tends to be abnormal. may be displayed.
- the method of displaying each individual determination result in Example 3 may display not only the finally determined bottle type, but also the degree of certainty for a plurality of bottle types.
- the certainty factor to be displayed for example, the maximum value, average value, or sum of the certainty factors in the combined area may be used, and may be combined with other image processing such as threshold processing or normalization processing.
- the image processing apparatus for identifying an object in an image disclosed in the first embodiment includes an image input unit 201 that receives an input image, and an image generator that generates an image for feature extraction from the input image.
- a feature amount extraction unit 203 for extracting a feature amount from the generated image generated by the image generation unit; and an identification unit for identifying an object in the image using the feature amount output by the feature amount extraction unit.
- 205 an output unit 206 that outputs the identification result output by the identification unit, and an instruction to the image generation unit to generate a new generated image as necessary based on the feature amount output by the feature amount extraction unit 203.
- a feature quantity map generation unit 204 for generating a feature quantity map indicating extraction conditions for the feature quantity for the newly generated image and outputting the feature quantity map to the feature quantity extraction unit 203 .
- the image generation unit 202 reduces the input image to generate the first generated image, and the feature amount map generation unit 204 determines that the already extracted feature amount is insufficient for the identification. , to instruct generation of a new generated image with a reduced reduction ratio.
- the feature amount map generation unit 204 identifies an area where the already extracted feature amount is insufficient for the identification, and generates a feature amount map that designates the area as a feature amount extraction range. In this way, by excluding regions where sufficient feature values have been obtained for identification, limiting the range of feature value extraction, and re-executing feature value extraction, the target object in the image can be detected with high accuracy and speed. can be identified.
- the feature amount map generation unit 204 determines conditions for generating a new generated image and the type of feature amount to be extracted from the new generated image based on the extraction result of the existing feature amount. Therefore, the type of generated image, the reduction ratio, and the type of feature amount to be extracted can be appropriately changed, and the target object in the image can be identified with high accuracy and high speed.
- the feature amount map generation unit 204 outputs an extraction end signal indicating the end of the feature amount extraction process to the identification unit 205, and the identification unit 205 outputs the extraction end signal.
- the features associated with one or more generated images generated so far are collectively used to identify the object in the image. Therefore, it is possible to eliminate unnecessary processing loads, efficiently generate necessary feature amounts, and identify a target object in an image with high accuracy and high speed.
- the feature quantity map generation unit 204 performs provisional identification using part or all of the feature quantity extracted by the feature quantity extraction unit 203 so far, and extracts the feature quantity based on the certainty of the identification. Generate a map. Therefore, it is possible to accurately determine the necessity of new feature quantity extraction.
- the image processing system disclosed in the second embodiment for detecting an object in an image at high speed includes an imaging device 1201 that captures an image of a target object to obtain the input image, the image processing device 100 disclosed in the first embodiment, An object detection device 1202 that detects a target object based on the identification result output by the image processing device 100 and detection target information output from a storage device 1204, and a detection target that specifies the target to be detected by the object detection device 1202 A storage device 1204 for outputting information and a display device 1203 for presenting the object detection result output by the object detection device 1202 to the user are provided. According to the configuration shown in the second embodiment, it is possible to detect the detection target object in the image at high speed.
- the image processing system disclosed in Example 3 uses object identification results and feature quantity extraction maps to suggest the possibility of ⁇ the placement of objects has been replaced'' and ⁇ the orientation of labels is incorrect''. It is also possible to display
- the present invention is not limited to the above examples, and includes various modifications.
- the above-described embodiments have been described in detail in order to explain the present invention in an easy-to-understand manner, and are not necessarily limited to those having all the described configurations.
- not only deletion of each component but also replacement and addition of the component are possible.
- Image processing device 110: Interface unit, 111: Calculation unit, 112: Memory, 113: Bus, 201: Image input unit, 202: Image generation unit, 203: Feature amount extraction unit, 204: Feature amount map generation unit , 205: identification unit, 206: output unit, 500: image processing method, 501: image input step, 502: image generation step, 503: feature amount extraction step, 504: feature amount map generation step, 505: identification step, 506 : output step 1200: image processing system 1201: imaging device 1202: object detection device 1203: display device 1204: storage device 1400: correspondence table
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
Description
また、本発明で開示される画像処理方法は、画像内の物体を識別する画像処理方法であって、入力画像を受付ける画像入力ステップと、前記入力画像から特徴量抽出のための画像を生成する画像生成ステップと、前記画像生成ステップにより生成された生成画像から特徴量を抽出する特徴量抽出ステップと、前記特徴量抽出ステップにて算出した特徴量を用いて、画像内の物体を検出あるいは識別する識別ステップと、前記識別ステップにて算出した識別結果を出力する出力ステップと、を含み、前記特徴量抽出ステップが出力する特徴量に基づき、必要に応じて新たな生成画像の生成を前記画像生成ステップにより行わせるとともに、当該新たな生成画像に対する特徴量の抽出条件を示す特徴量マップを生成して前記特徴量抽出ステップによる処理に反映させる特徴量マップ生成ステップをさらに含む。
図1を用いて実施例1に係る画像処理装置のハードウェア構成について説明する。画像処理装置100は、インターフェース部110と、演算部111と、メモリ112と、バス113とを備え、インターフェース部110、演算部111、メモリ112はバス113を介して情報の送受信を行う。
インターフェース部110は、画像処理装置100の外部にある装置と信号の送受信を行う通信装置である。インターフェース部110と通信を行う装置としてはカメラや顕微鏡などの撮像装置120、モニタやプリンタ等の表示装置121がある。
メモリ112は、演算部111が実行するプログラムや、パラメタ、係数、処理結果等を保存する装置であり、HDD、SSD、RAM、ROM、フラッシュメモリ等である。
図2は画像処理装置100の第一の実施例に係る機能ブロック図の一例である。これらの各機能部は、演算部111上で動作するソフトウェアで実現しても良いし、専用のハードウェアで実現しても良い。
以下、機能部の内、画像生成部202、特徴量抽出部203、特徴量マップ生成部204、識別部205の動作について詳しく説明する。
画像処理装置100は、実施例1に記載の画像処理装置であり、前記撮像装置1201で撮像した画像から特定の物体領域の領域分割または識別結果、あるいはその両方を算出する。
物体検出装置1202は、前記画像処理装置100が出力する領域分割結果または識別結果と、記憶装置1204が出力する検出関連指定信号の内、検出対象を指定する信号である検出対象指定信号に基づいて物体の検出情報を出力する。
表示装置1203は、前記物体検出装置1202による物体検出結果をユーザーに提示する。
撮像装置1201、表示装置1203、および記憶装置1204の動作は上述の通りである。また、画像処理装置100の動作は実施例1にて説明した通りである。そのためここでは物体検出装置1202について説明する。
以上により、画像内の対象物体を高速に検出する画像処理システムを提供可能となる。
実施例3に係るハードウェア構成図を図6に示す。実施例2に係る画像処理システム600は、撮像装置1201と、画像処理装置100と、物体検出装置1202と、物体配置検証装置602と、表示装置1203と、記憶装置1204と、から構成される。
記憶装置1204の動作について説明する。本実施例において、記憶装置1204に保持されている検出関連指定信号は、検出対象を指定する検出対象指定信号と、対象物体の正しい配置状態を表す正配置状態指定信号の二つを含む。記憶装置1204は検出対象指定信号を物体検出装置1202に出力し、正配置状態指定信号を物体配置検証装置602に出力する。尚、検出対象指定信号および正配置状態指定信号は予めユーザーにより設定されているものとする。
物体配置検証装置602は、物体検出装置1202が出力する物体の配置情報と、記録装置1204が出力する正配置状態指定信号を受付け、両者が一致しているか否かに基づいて、撮像装置1201が撮像した画像内の物体が正しく配置されているかを検証する。
図8Bの例では左から二番目と三番目のボトルの配置が正しくないため、個別検証結果804および805は“Error”と表示されており、総合検証結果802も“Error”となっている。
図8Cの例は、右端のボトルが想定と異なるボトルであるため正しく検出できなかった例である。左と真ん中のボトルの配置は正しいため、個別検証結果803および804は“OK”と表示されている。しかし、配置情報の算出結果は“AB”となり登録されたボトルの正しい配置“ABC”とは一致しない。そのため総合検証結果802には“Error”と表示されている。また“Error”という表示ではなく、検出されるべき数の対象物体が検出されなかった旨の表示、あるいは検出した対象物体の数を表示する等しても良い。
撮像装置1201、表示装置1203は実施例2にて説明した各装置と同様であるため説明を省略する。ただし、撮像装置1201は、試料を格納した容器(以下、検体)を横から撮像するものとする。画像処理装置100は容器種別を識別するための領域分割を行い、物体検出装置1202は領域分割結果に基づいて容器種別を判定し、出力する。詳細は後述する。
以下では画像処理装置100、物体検出装置1202、検体状態判定装置902、試料採取装置903、分析装置904の動作について詳細を説明する。
L=d2-l+d1
ここで、d2は底上げ容器下端の垂直方向の座標位置であり、1は画像処理装置100が推定したカップ領域下端の垂直方向の座標位置であり、d1はカップ下端から床上げ容器との接着点までの距離である。図15に図10Aの例における各パラメタの関係を図示する。尚、カップ下端から床上げ容器との接着点までの距離d1は、容器の種類毎に予め計測して記憶装置1204に記録しておくものとし、物体検出装置1202が特定したカップの種別に応じて記憶装置1204から読み出す。底上げ容器下端の垂直方向の座標位置d2は、どの底上げ容器であっても同じ位置であると仮定し、予め計測して記憶装置1204に記録しておくものとする。ただし、もし、底上げ容器下端の変動が大きい場合は、画像処理装置100により、カップ領域とは別に、底上げ容器下端の位置を推定するようにしてもよい。
分析装置904は、試料採取装置903が取得した試料と薬品を混合することで血液分析を行う。
出力装置905は、前記血液分析結果を表示する装置であり、モニタやプリンタ等である。
実施例1の特徴量マップ生成部204における生成画像信号算出の際、生成画像信号のスコアに応じて複数の画像を生成するように生成画像信号を算出しても良い。この場合、生成画像信号を受付けた画像生成部202は、生成画像信号に基づいて複数の画像を生成し、特徴量抽出部203に出力する。この時、特徴量マップ生成部204は各生成画像に対する特徴量抽出マップを生成する。特徴量抽出部203は画像生成部202から受付けた生成画像群と、特徴量マップ生成部204から受付けた特徴量抽出マップ群を用いて、各ペア(生成画像と特徴量抽出マップの組み合わせ)から特徴量を抽出する。
このように、縮小率の高い生成画像から処理を開始し、必要に応じて縮小率を下げた生成画像を用いることにより、画像内の対象物体を高精度かつ高速に識別することができる。
また、特徴量マップ生成部204は、既に抽出された特徴量では前記識別に不十分な領域を特定し、当該領域を特徴量抽出の範囲として指定する特徴量マップを生成する。
このように、識別に十分な特徴量を得られた領域を除外し、特徴量抽出の範囲を限定して再度の特徴量抽出を実行することで、画像内の対象物体を高精度かつ高速に識別することができる。
このため、生成画像の種類、縮小率、抽出する特徴量の種別を適宜変更し、画像内の対象物体を高精度かつ高速に識別することができる。
このため、不要な処理負荷を排除し、必要な特徴量を効率的に生成し、画像内の対象物体を高精度かつ高速に識別することができる。
このため、新たな特徴量抽出の必要性を精度良く判定することができる。
特徴量抽出マップを外部出力することで、どのような識別処理を行っているかを情報提供できる。
実施例2に示す構成によれば、画像内の検出対象物体を高速に検出することが可能となる。
また、実施例3に開示した画像処理システムにおいて物体の識別結果や特徴量抽出マップを利用し、「物体の配置が入れ替わっている」、「ラベルの向きが誤っている」などの可能性を示唆する表示を行うことも可能である。
Claims (15)
- 画像内の物体を識別する画像処理装置であって、
入力画像を受付ける画像入力部と、
前記入力画像から特徴量抽出のための画像を生成する画像生成部と、
前記画像生成部により生成された生成画像から特徴量を抽出する特徴量抽出部と、
前記特徴量抽出部が出力する特徴量を用いて、画像内の物体を識別する識別部と、
前記識別部が出力する識別結果を出力する出力部と、
前記特徴量抽出部が出力する特徴量に基づき、必要に応じて新たな生成画像の生成を前記画像生成部に指示するとともに、当該新たな生成画像に対する特徴量の抽出条件を示す特徴量マップを生成して前記特徴量抽出部に出力する特徴量マップ生成部と
を備えることを特徴とする画像処理装置。 - 請求項1に記載の画像処理装置であって、
前記画像生成部は、前記入力画像を縮小して最初の生成画像を生成し、
前記特徴量マップ生成部は、既に抽出された特徴量では前記識別に不十分と判定した場合に、縮小率を下げた新たな生成画像の生成を指示することを特徴とする画像処理装置。 - 請求項1に記載の画像処理装置であって、
前記特徴量マップ生成部は、既に抽出された特徴量では前記識別に不十分な領域を特定し、当該領域を特徴量抽出の範囲として指定する特徴量マップを生成することを特徴とする画像処理装置。 - 請求項1に記載の画像処理装置であって、
前記特徴量マップ生成部は、既存の特徴量の抽出結果に基づいて、新たな生成画像の生成条件と、当該新たな生成画像から抽出すべき特徴量の種別とを決定することを特徴とする画像処理装置。 - 請求項1に記載の画像処理装置であって、
前記特徴量マップ生成部は、所定の条件が成立した場合に、特徴量抽出処理の終了を示す抽出終了信号を前記識別部に出力し、
前記識別部は、前記抽出終了信号を受けた場合に、それまでに生成された1又は複数の生成画像に係る特徴量を総合的に用いて、前記画像内の物体を識別する
ことを特徴とする画像処理装置。 - 請求項1に記載の画像処理装置であって、
前記特徴量マップ生成部は、前記特徴量抽出部がそれまでに抽出した特徴量の一部または全てを用いて仮の識別を実行し、当該識別の確信度に基づいて特徴量抽出マップを生成することを特徴とする画像処理装置。 - 請求項1に記載の画像処理装置であって、
前記特徴量マップ生成部が出力する特徴量抽出マップを、前記出力部を介して外部に出力することを特徴とする画像処理装置。 - 画像内の物体を高速に検出する画像処理システムであって、
対象物体を撮像して前記入力画像を得る撮像装置と、
前記撮像装置が出力する前記入力画像に基づいて領域分割結果または識別結果を出力する請求項1に記載の画像処理装置と、
検出に関連する設定や情報を指定するための信号である検出関連指定信号として、検出対象を指定するための検出対象指定信号を出力する記憶装置と、
前記画像処理装置が出力する領域分割結果または識別結果と、前記記憶装置が出力する検出対象指定信号とに基づいて物体を検出する物体検出装置と、
前記物体検出装置が出力した物体検出結果をユーザーに提示する表示装置と、
を備えることを特徴とする画像処理システム。 - 請求項8に記載の画像処理システムであって、
前記画像処理装置は前記入力画像に基づく領域分割結果を出力し、
前記記憶装置は検出関連指定信号として、前記入力画像内に存在する複数物体の配置情報を検出するための検出対象指定信号と、予めユーザーによって設定された対象物体の正配置状態指定信号を出力し、
前記物体検出装置は前記領域分割結果と、前記検出対象指定信号を受付け、画像内に存在する複数物体の配置情報を検出し、
前記物体検出装置が出力する物体の配置情報と前記記憶装置が出力する正配置状態指定信号に基づいて、前記対象物体の配置が正しい配置であるか否かを検証する物体配置検証装置
をさらに備え、
前記表示装置は前記物体配置検証装置が出力する物体配置検証結果をユーザーに提示することを特徴とする画像処理システム。 - 請求項8に記載の画像処理システムであって、
前記撮像装置は試料を格納した容器を撮像して前記入力画像を取得し、
前記画像処理装置は前記入力画像に基づく領域分割結果または識別結果を出力し、
前記記憶装置は検出関連指定信号として、前記入力画像内に存在する容器の種別を検出するための検出対象指定信号と、前記試料及び/又は前記容器の状態を判定するための試料状態判定対象指定信号及び/又は容器状態判定対象指定信号を出力し、
前記物体検出装置は前記領域分割結果または識別結果と、前記検出対象指定信号に基づいて容器の種別を判定し、
前記物体検出装置が出力する前記容器の種別と、前記記憶装置が出力する前記試料状態判定対象指定信号及び/又は容器状態判定対象指定信号に基づいて、前記試料及び/又は前記容器の状態を判定する検体状態判定装置と、
前記検体状態判定装置が判定した前記試料及び/又は前記容器の状態に基づいて、前記試料の採取制御を行う試料採取装置と、
前記試料採取装置が採取した試料と試薬を混合することで血液分析を行う分析装置と、
をさらに備え、
前記表示装置は前記分析装置による血液分析結果をユーザーに提示することを特徴とする画像処理システム。 - 画像内の物体を識別する画像処理方法であって、
入力画像を受付ける画像入力ステップと、
前記入力画像から特徴量抽出のための画像を生成する画像生成ステップと、
前記画像生成ステップにより生成された生成画像から特徴量を抽出する特徴量抽出ステップと、
前記特徴量抽出ステップにて算出した特徴量を用いて、画像内の物体を検出あるいは識別する識別ステップと、
前記識別ステップにて算出した識別結果を出力する出力ステップと、を含み、
前記特徴量抽出ステップが出力する特徴量に基づき、必要に応じて新たな生成画像の生成を前記画像生成ステップにより行わせるとともに、前記特徴量抽出ステップにおける当該新たな生成画像に対する特徴量の抽出条件を示す特徴量マップを生成する特徴量マップ生成ステップをさらに含むことを特徴とする画像処理方法。 - 請求項11に記載の画像処理方法であって、
前記画像生成ステップは、前記入力画像を縮小して最初の生成画像を生成し、
前記特徴量マップ生成ステップは、既に抽出された特徴量では前記識別に不十分と判定した場合に、縮小率を下げた新たな生成画像の生成を指示することを特徴とする画像処理方法。 - 請求項11に記載の画像処理方法であって、
前記特徴量マップ生成ステップは、既に抽出された特徴量では前記識別に不十分な領域を特定し、当該領域を特徴量抽出の範囲として指定する特徴量マップを生成することを特徴とする画像処理方法。 - 請求項11に記載の画像処理方法であって、
前記特徴量マップ生成ステップは、前記特徴量抽出ステップがそれまでに抽出した特徴量の一部または全てを用いて仮の識別を実行し、当該識別の確信度に基づいて特徴量抽出マップを生成することを特徴とする画像処理方法。 - 請求項11に記載の画像処理方法であって、
前記特徴量マップ生成ステップが出力する特徴量抽出マップを、前記出力ステップにて外部に出力することを特徴とする画像処理方法。
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202280043801.9A CN117546209A (zh) | 2021-07-02 | 2022-06-23 | 图像处理装置、方法以及图像处理系统 |
US18/572,539 US20240119698A1 (en) | 2021-07-02 | 2022-06-23 | Image processing device and method, and image processing system |
EP22833000.7A EP4365833A1 (en) | 2021-07-02 | 2022-06-23 | Image processing device, method, and image processing system |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2021-110546 | 2021-07-02 | ||
JP2021110546A JP2023007599A (ja) | 2021-07-02 | 2021-07-02 | 画像処理装置、方法及び画像処理システム |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2023276853A1 true WO2023276853A1 (ja) | 2023-01-05 |
Family
ID=84691305
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2022/025146 WO2023276853A1 (ja) | 2021-07-02 | 2022-06-23 | 画像処理装置、方法及び画像処理システム |
Country Status (5)
Country | Link |
---|---|
US (1) | US20240119698A1 (ja) |
EP (1) | EP4365833A1 (ja) |
JP (1) | JP2023007599A (ja) |
CN (1) | CN117546209A (ja) |
WO (1) | WO2023276853A1 (ja) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2012113460A (ja) * | 2010-11-24 | 2012-06-14 | Sony Corp | 情報処理装置および方法、並びにプログラム |
JP2013210967A (ja) * | 2012-03-30 | 2013-10-10 | Canon Inc | 体検出方法及び物体検出装置 |
JP2016516245A (ja) * | 2013-03-13 | 2016-06-02 | コファックス, インコーポレイテッド | モバイル装置を用いた画像内のオブジェクトの分類 |
JP2020020744A (ja) * | 2018-08-03 | 2020-02-06 | Automagi株式会社 | 錆検出プログラム、錆検出システム及び錆検出方法 |
JP2021064120A (ja) * | 2019-10-11 | 2021-04-22 | キヤノン株式会社 | 情報処理装置、情報処理方法、及びプログラム |
-
2021
- 2021-07-02 JP JP2021110546A patent/JP2023007599A/ja active Pending
-
2022
- 2022-06-23 US US18/572,539 patent/US20240119698A1/en active Pending
- 2022-06-23 WO PCT/JP2022/025146 patent/WO2023276853A1/ja active Application Filing
- 2022-06-23 EP EP22833000.7A patent/EP4365833A1/en active Pending
- 2022-06-23 CN CN202280043801.9A patent/CN117546209A/zh active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2012113460A (ja) * | 2010-11-24 | 2012-06-14 | Sony Corp | 情報処理装置および方法、並びにプログラム |
JP2013210967A (ja) * | 2012-03-30 | 2013-10-10 | Canon Inc | 体検出方法及び物体検出装置 |
JP2016516245A (ja) * | 2013-03-13 | 2016-06-02 | コファックス, インコーポレイテッド | モバイル装置を用いた画像内のオブジェクトの分類 |
JP2020020744A (ja) * | 2018-08-03 | 2020-02-06 | Automagi株式会社 | 錆検出プログラム、錆検出システム及び錆検出方法 |
JP2021064120A (ja) * | 2019-10-11 | 2021-04-22 | キヤノン株式会社 | 情報処理装置、情報処理方法、及びプログラム |
Also Published As
Publication number | Publication date |
---|---|
EP4365833A1 (en) | 2024-05-08 |
US20240119698A1 (en) | 2024-04-11 |
JP2023007599A (ja) | 2023-01-19 |
CN117546209A (zh) | 2024-02-09 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
TWI512684B (zh) | Defect observation method and defect observation device | |
US10297021B2 (en) | Defect quantification method, defect quantification device, and defect evaluation value display device | |
US5933519A (en) | Cytological slide scoring apparatus | |
US8135202B2 (en) | Automated method and system for nuclear analysis of biopsy images | |
KR101523159B1 (ko) | 패턴 매칭 장치 및 컴퓨터 프로그램을 기록한 기록 매체 | |
US20030050761A1 (en) | Inspection method and its apparatus, inspection system | |
JP5010207B2 (ja) | パターン検査装置及び半導体検査システム | |
JP2019061484A (ja) | 画像処理装置及びその制御方法及びプログラム | |
CN110974306B (zh) | 一种超声内镜下识别和定位胰腺神经内分泌肿瘤的系统 | |
JP6733983B2 (ja) | 画像解析装置 | |
KR102140385B1 (ko) | 골수 판독을 위한 셀존 레이블링 장치 및 이를 포함하는 셀존 검출 시스템 | |
TW201237363A (en) | Image processing apparatus and computer program | |
KR20190059083A (ko) | 영상 분할 기반의 해양 상황 인식 장치 및 그 방법 | |
JP3661635B2 (ja) | 画像処理方法およびその装置 | |
CN114332058A (zh) | 基于神经网络的血清质量识别方法、装置、设备及介质 | |
WO2023276853A1 (ja) | 画像処理装置、方法及び画像処理システム | |
JP4500138B2 (ja) | 細胞画像解析方法 | |
JP2013058036A (ja) | 画像処理装置、画像処理プログラム、ロボット装置及び画像処理方法 | |
EP3438987B1 (en) | Tool that analyzes image data and generates and displays a confidence indicator along with a cancer score | |
US20210166076A1 (en) | System and method for analysis of microscopic image data and for generating an annotated data set for classifier training | |
CN115830025A (zh) | 白细胞分类计数方法、系统、存储介质及计算机设备 | |
JP2001314374A (ja) | 角膜内皮細胞測定装置 | |
CN110543802A (zh) | 眼底图像中左右眼识别方法与装置 | |
CN114742803A (zh) | 一种结合深度学习与数字图像处理算法的血小板聚集检测方法 | |
CN113302472A (zh) | 生物体试样检测装置 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 22833000 Country of ref document: EP Kind code of ref document: A1 |
|
DPE1 | Request for preliminary examination filed after expiration of 19th month from priority date (pct application filed from 20040101) | ||
WWE | Wipo information: entry into national phase |
Ref document number: 202280043801.9 Country of ref document: CN Ref document number: 18572539 Country of ref document: US |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2022833000 Country of ref document: EP |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
ENP | Entry into the national phase |
Ref document number: 2022833000 Country of ref document: EP Effective date: 20240202 |