Feature amount evaluation method, device and system therefor, feature amount evaluation program, and storage medium storing feature amount evaluation program

Abstract

This invention evaluates the non-uniformity of a feature amount.　In the present disclosure: local histogram information indicating a relationship between a feature amount and a frequency is obtained for each local sample range; n pieces of local histogram information are accumulated, starting from the first piece, to generate a plurality of cumulative histogram information pieces; the accumulated and generated cumulative histogram information pieces, with N being the maximum number, are used as reference cumulative histogram information; the degree of similarity to the reference cumulative histogram information of each piece of cumulative histogram information is detected; and a minimum n at which the detected degree of similarity is at least a predetermined threshold is used as part of an index indicating the non-uniformity of the feature amount.

Description

Feature evaluation method, device and system thereof, feature evaluation program, and storage medium storing the feature evaluation program

[0001] The present disclosure relates to a feature amount evaluation method for evaluating non-uniformity of a feature amount, a feature amount evaluation device, a feature amount evaluation system, a feature amount evaluation program, and a storage medium storing the feature amount evaluation program.

[0002] In evaluating a production line for an article, it is common to extract characteristic data from the manufactured products, analyze and evaluate the characteristic data, and thereby grasp the state of the production line. For example, Patent Document 1 describes a defect inspection device that calculates feature quantities of defect candidate areas from an image of an inspection object such as a semiconductor mask pattern or a printed circuit board, and inspects whether the inspection object contains a defect based on the calculated feature quantities.

[0003] Such analysis and evaluation of characteristic data, i.e., feature quantities (hereinafter simply referred to as feature quantity evaluation), is used not only at manufacturing sites but also in various data processing situations, including the processing of statistical data, such as new product development, commodity trading sites, and information analysis of social conditions, corporate performance, population trends, etc. Furthermore, such feature quantity evaluation tools are actually used by applying them to dedicated equipment (manufacturing equipment, inspection equipment, analytical equipment, diagnostic equipment) used for each evaluation target, or by being installed as computer programs in general-purpose computers, control devices, etc.

[0004] However, while conventional feature evaluation methods can measure feature amounts at multiple locations and calculate an average feature amount, they do not provide an index for objectively determining the non-uniformity (or uniformity) of feature amounts.

[0005] Patent Document 1: Japanese Patent Application Laid-Open No. 2024-11895

[0006] An object of the present disclosure is to provide a feature evaluation method, a feature evaluation device, a feature evaluation system, a feature evaluation program, and a storage medium storing a feature evaluation program, which are capable of objectively quantifying and evaluating the non-uniformity of any feature.

[0007] A feature evaluation method according to the present disclosure is a feature evaluation method for evaluating heterogeneity of a feature in an evaluation target range, which is a set of elements having a predetermined feature, the feature evaluation method comprising the steps of: obtaining, for each of a plurality of local sample ranges set in the evaluation target range, local histogram information indicating a relationship between the feature and a frequency within the sample range; accumulating the first to Nth local histogram information for each of the sample ranges, sequentially from the first to the Nth local histogram information, to generate a plurality of n (n = 1 to N) pieces of cumulative histogram information; using the cumulative histogram information generated by accumulating n = N pieces as reference cumulative histogram information, and detecting a similarity to the reference cumulative histogram information for each of the plurality of pieces of cumulative histogram information generated by sequentially accumulating the local histogram information from n = 1 to N; and sequentially detecting, from n = 1 to n = N, whether the detected similarity satisfies a predetermined criterion, and using a number n representing the nth cumulative histogram information that first satisfies the criterion, as part of an index indicating heterogeneity of the feature.

[0008] By using this feature evaluation method, it is possible to objectively quantify and evaluate the nonuniformity of features for an evaluation target. As a result, for example, in product development and manufacturing sites or commodity trading sites, products with uniform features can be appropriately supplied and traded. Furthermore, in information analysis and data processing, for example, the variation of features within the evaluation target range can be appropriately detected or evaluated as quantified features, enabling more effective information analysis and data processing. Furthermore, when the nonuniformity of features indicates functional degradation of equipment or incompatibility of a data processing system, these undesirable conditions can be appropriately detected.

[0009] a feature evaluation device according to the present disclosure that evaluates non-uniformity of a feature within an evaluation target range, which is a set of elements having a predetermined feature, comprising: means for obtaining, for each of a plurality of local sample ranges set in the evaluation target range, local histogram information indicating a relationship between the feature and a frequency within the sample range; means for accumulating the first to Nth local histogram information for each of the sample ranges, sequentially from the first to the Nth local histogram information, to generate a plurality of n (n = 1 to N) pieces of cumulative histogram information; means for using the cumulative histogram information generated by accumulating n = N pieces as reference cumulative histogram information, and for detecting a similarity to the reference cumulative histogram information for each of the plurality of pieces of cumulative histogram information generated by sequentially accumulating the local histogram information from n = 1 to N; and means for sequentially detecting, from n = 1 to n = N, whether the detected similarity satisfies a predetermined criterion, and using a number n representing the nth cumulative histogram information that first satisfies the criterion, as part of an index indicating non-uniformity of the feature.

[0010] A system according to the present disclosure is a system including a plurality of devices, and executes a program in the plurality of devices to perform a process of evaluating non-uniformity of a feature amount for an evaluation target range, which is a set of elements having a predetermined feature amount, comprising: a step of obtaining, for each of a plurality of local sample ranges set in the evaluation target range, local histogram information indicating a relationship between the feature amount and a frequency within the sample range; a step of accumulating the first to Nth local histogram information for each of the sample ranges sequentially from the first to the Nth local histogram information to generate a plurality of n (n=1 to N) pieces of cumulative histogram information; and a step of using the cumulative histogram information generated by accumulating n=N pieces as reference cumulative histogram information, and detecting a similarity to the reference cumulative histogram information for each of the plurality of cumulative histogram information generated by accumulating the local histogram information sequentially from n=1 to N. detecting whether the detected similarities conform to a predetermined criterion in order from n=1 to n=N, and using the number n representing the nth piece of cumulative histogram information that first conforms to the criterion as part of an index showing the non-uniformity of the features.

[0011] A feature evaluation program according to the present disclosure is a feature evaluation program for evaluating heterogeneity of a feature in an evaluation target range, which is a set of elements having a predetermined feature, and includes the steps of: obtaining, for each of a plurality of local sample ranges set in the evaluation target range, local histogram information indicating a relationship between the feature and a frequency within the sample range; accumulating the first to Nth local histogram information for each of the sample ranges, sequentially from the first to the Nth local histogram information, to generate a plurality of n (n=1 to N) pieces of cumulative histogram information; using the cumulative histogram information generated by accumulating n=N pieces as reference cumulative histogram information, and detecting a similarity to the reference cumulative histogram information for each of the plurality of pieces of cumulative histogram information generated by sequentially accumulating the local histogram information from n=1 to n=N; and sequentially detecting, from n=1 to n=N, whether the detected similarity satisfies a predetermined criterion, and using a number n representing the nth cumulative histogram information that first satisfies the criterion, as part of an index indicating heterogeneity of the feature. This is a feature evaluation program for causing a computer to execute the above.

[0012] A storage medium storing a feature evaluation program according to the present disclosure is a storage medium storing a feature evaluation program for evaluating non-uniformity of a feature in an evaluation target range, which is a set of elements having a predetermined feature, and includes the steps of: obtaining, for each of a plurality of local sample ranges set in the evaluation target range, local histogram information indicating a relationship between the feature and a frequency within the sample range; accumulating the first to Nth local histogram information for each of the sample ranges sequentially from the first to the Nth local histogram information to generate a plurality of n (n = 1 to N) pieces of cumulative histogram information; and determining the cumulative histogram information generated by accumulating n = N pieces as reference cumulative histogram information, and detecting a similarity to the reference cumulative histogram information for each of the plurality of cumulative histogram information generated by accumulating the local histogram information sequentially from n = 1 to N. and a step of detecting whether the detected similarities conform to a predetermined criterion sequentially from n=1 to n=N, and using the number n representing the nth piece of cumulative histogram information that first conforms to the criterion as part of an index showing the non-uniformity of the features.

[0013] When these feature evaluation devices, feature evaluation systems, feature evaluation programs, and storage media storing the feature evaluation programs are used to operate the evaluation devices, run them on multiple devices, or run the programs on a computer to perform the functions of the present disclosure, it is possible to objectively quantify and evaluate the non-uniformity of the features of an article. As a result, products with uniform features can be appropriately supplied and traded, for example, at product development and manufacturing sites or commodity trading sites.

[0014] The elements are separated so as to be arranged in a d-dimensional space having d index axes (d=1, 2, 3, or 4 or more), the evaluation target range is a predetermined spatial range in the d-dimensional space, and the sample range may be a local spatial range included in the evaluation target range in the d-dimensional space.

[0015] By targeting the feature values ​​of each element arranged along one-, two-, three-, or four- or more-dimensional index axes, it is possible to quantify and evaluate the heterogeneity of feature values ​​for any evaluation target element (i.e., evaluation target range) arranged in any space.

[0016] The plurality of sample ranges may be set such that the distance between each other in the d-dimensional space is equal to or greater than a predetermined distance.

[0017] By setting multiple sample ranges at least a predetermined distance apart, it is possible to detect features from locations (positions) with low correlation in the space where the evaluation target range is placed and verify similarity. As a result, it is possible to more appropriately detect the distribution state of the features of each element placed in the evaluation target range, and to detect index values ​​that more appropriately reflect the heterogeneity (homogeneity) of the features.

[0018] When obtaining the local histogram information, the frequencies may be converted to logarithms, and local histogram information having the logarithm-converted frequencies as frequencies for each feature may be obtained. By converting the frequencies to logarithms, it becomes possible to perform statistical processing in which the association (correlation) between histograms is clearly reflected in the similarity between the histograms. In other words, by using a histogram (local histogram information) formed using the logarithm-converted frequencies, it becomes possible to clearly detect differences in the distribution of feature amounts as differences in the similarity between histograms.

[0019] In generating the local histogram information, the local histogram information may be normalized based on the most frequent value, and in generating the cumulative histogram information, the cumulative histogram information may be normalized based on the most frequent value.

[0020] By normalizing the feature quantities of multiple sample ranges in the evaluation target range, when local histogram information or cumulative histogram information is accumulated (added) in subsequent processing, the weights of the local histogram information for each sample range or the cumulative histogram information for each combination of accumulated sample ranges can be made approximately equal. In other words, it is possible to avoid differences in the weights of each local histogram information or effects on the similarity between cumulative histogram information depending on the absolute value (actual frequency) of the frequency of the feature quantities. As a result, appropriate histogram information can be obtained in which the contributions of the feature quantity distributions (local histogram information or cumulative histogram information) for multiple sample ranges are equalized, allowing for more appropriate evaluation of non-uniformity in the evaluation target range.

[0021] In the step of using the number n representing the nth position as part of an index indicating the non-uniformity of the feature amount, if the detected similarity satisfies the formula (2), it may be determined that the similarity meets a predetermined standard. n ≧ Hm THL ... (2) However, Hm n is the similarity between the n-th cumulative histogram information and the reference cumulative histogram information, and Hm THL is a predetermined similarity threshold.

[0022] By making this judgment, an index is generated using a value n that indicates that the nth cumulative histogram information is similar to the distribution state of the feature of the measurement target with a sufficiently high degree of similarity, and an index can be obtained that appropriately represents the heterogeneity / homogeneity of the feature of the evaluation target range.

[0023] FIG. 1 is a flowchart showing the processing steps of a feature non-uniformity detection method according to the present disclosure. FIG. 2 is a block diagram showing the configuration of a feature non-uniformity detection device according to the present disclosure. FIG. 3A is a diagram showing an example of an evaluation target for explaining the processing steps of a feature non-uniformity detection method according to the present disclosure. FIG. 3B is a diagram showing a local histogram information calculation process. FIG. 3C is a diagram showing a process of calculating the similarity between cumulative histogram information and reference cumulative histogram information and comparing it with a predetermined threshold. FIG. 4A is a diagram showing histograms of feature amounts of all elements included in an evaluation target range. FIG. 4B is a diagram showing a convergence state of the similarity between cumulative histogram information and reference cumulative histogram information for a group of capacitors in which capacitance non-uniformity was observed and a group of capacitors in which capacitance non-uniformity was observed to be low. FIG. 5 is a diagram showing the similarity of histograms. FIG. 6A is a diagram showing a histogram of particle area of ​​dielectric particles of an MLCC according to an embodiment of the present disclosure. FIG. 6B is a diagram illustrating a convergence state of the similarity between the cumulative histogram information and the reference cumulative histogram information with respect to the particle area of ​​the dielectric particles of the MLCC according to an embodiment of the present disclosure.

[0024] A method, apparatus, system, program, etc. relating to feature non-uniformity detection (evaluation) according to the present disclosure will be described with reference to FIGS. 1 to 5. FIG.

[0025] The embodiments of the present disclosure described below are examples for explaining the present disclosure. Various components according to the embodiments of the present disclosure, such as numerical values, shapes, materials, manufacturing processes, data processing methods, etc., can be modified or changed within the scope of technical problems. Furthermore, the shapes, data processing methods, etc. shown in the drawings of the present disclosure do not necessarily correspond to the actual shapes, methods, etc. This is because the shapes, etc. may be modified for the purpose of explanation.

[0026] Here, the "non-uniformity of feature quantities" according to the present disclosure is an index indicating a state in which the state of feature quantities in the entire evaluation range differs depending on the location (position) of sample ranges (sometimes referred to as specimen ranges or sampling ranges) within the evaluation range, and feature non-uniformity indicates a state in which feature quantities "variegate" depending on the location of the sample range. Therefore, even if a feature quantity is low from the perspective of superiority, if the "lowness of feature quantities" is approximately the same throughout the evaluation range, the non-uniformity can be evaluated as low (high uniformity). Furthermore, even if a feature quantity in a certain sample range has an excellent value from the perspective of superiority, if the degree of superiority varies depending on the location of the sample range within the evaluation range, the non-uniformity is evaluated as high (low uniformity).

[0027] In this application, the concept of the feature quantities "non-uniformity (degree of non-uniformity)" and "uniformity (degree of uniformity)" is simply referred to as "non-uniformity (degree of non-uniformity)," and in particular, in contexts focusing on one of them, it may also be simply referred to as "non-uniformity."

[0028] In the method, apparatus, etc. disclosed herein, the target and feature quantity for detecting feature nonuniformity may be any target and any feature quantity. However, it is preferable to use a target for evaluation that can conceive of feature quantities for individual elements arranged in a d-dimensional space having any d index axes, and that can obtain a frequency distribution (histogram) of the feature quantities for each local range (sample range) within the range in which the elements are arranged (evaluation target range).

[0029] Specifically, the evaluation targets can include features such as the age-specific population ratio for each region (a histogram of the number of people by age for each region, arranged along a one-dimensional indicator called "by region"); sales or income by item for each sales office (a histogram of sales / income by item for each sales office, arranged along a one-dimensional indicator called "by branch"); the distribution of scores by class for each test (a histogram of the score distribution by class for each test, arranged along a one-dimensional indicator called "by test"); the distribution of the characteristics (capacitor capacitance, thickness, etc.) of each product (capacitor) for each lot, arranged along a one-dimensional indicator called "group" (e.g., production lot), which indicates the order in which the products were manufactured on a capacitor production line; or the distribution of the weight of a specific fruit, arranged along a one-dimensional indicator called "unit of arrival."

[0030] Furthermore, for evaluation target images containing images of crystal grains (images in which crystal grains are arranged in a space defined by two-dimensional indices), histogram information of feature quantities such as the diameter and circularity of crystal grains within a specific local region (sample area) can be detected to evaluate their nonuniformity. More generally, for evaluation target images containing arbitrary shapes, histogram information can be detected using the diameter, circularity, and arbitrary shape characteristics (length, number, etc.) of voids within a specific local region (sample area) as feature quantities to evaluate the nonuniformity of the shapes present in the image. Alternatively, grayscale images resulting from lattice diffraction can be used as evaluation target images to detect the nonuniformity of the distribution of brightness values ​​(features) for each specific local region (sample area), and this can be used to determine the composition of the analyzed material.

[0031] Furthermore, by detecting the non-uniformity of the particle size distribution of the powder in each local space within the deposition space (within a three-dimensional space) for a large amount of deposited (accumulated) powder material (for example, handled in units of several tons), it is possible to understand the overall characteristics and deposition (accumulation) state of the deposited (accumulated) material.

[0032] Alternatively, the elements and features exemplified above, i.e., the evaluation targets, can be arranged in two- or three-dimensional geometric space and, taking into account time changes, can be regarded as features arranged along three- or four-dimensional index axes, and heterogeneity can be detected using these as evaluation targets. Similarly, heterogeneity in the distribution of feature values ​​can be detected for elements arranged along any five or more index axes, i.e., for evaluation targets arranged by five or more dimensional indexes (conditions).

[0033] In this way, the target and feature for detecting non-uniformity of feature by the method, device, etc. disclosed herein may be arbitrarily set and defined.

[0034] 1 is a flowchart showing an example of a processing procedure of a feature non-uniformity detection method according to an embodiment of the present disclosure. As shown in FIG. 1, the non-uniformity detection method according to this embodiment includes the steps of collecting feature original data (step S1), preprocessing such as flattening (step S2), calculating local histogram information (step S3), calculating cumulative histogram information (step S4), calculating the similarity between the cumulative histogram information and reference cumulative histogram information (step S5), and detecting the degree of non-uniformity (step S6).

[0035] The feature non-uniformity detection (evaluation) method according to the present disclosure can also be achieved, for example, by the device shown in Fig. 2. Fig. 2 is a block diagram showing an example of the configuration of the feature non-uniformity detection device according to an embodiment of the present disclosure. The configuration of the feature non-uniformity detection device shown in Fig. 2 will be described later.

[0036] 3A to 4B are diagrams illustrating a method for detecting non-uniformity in feature quantities using the method of the present disclosure, etc. In this embodiment, the capacitance of each capacitor manufactured on a capacitor production line is evaluated, and a process for detecting specific variations (non-uniformity) in capacitance of capacitors for each predetermined unit over time from the start of production is described.

[0037] 3A is a diagram showing a capacitor manufacturing line in which 20,000 capacitors are manufactured in sequence. The process of each step in the flowchart shown in FIG. 1 will be described below with reference to the example shown in FIG. 3A.

[0038] First, in the process of detecting non-uniformity of feature quantities, data on feature quantities (capacitance) of each element (capacitor) within the evaluation target range is collected, and the feature quantity original data Do n (n=1 to N) is generated (step S1). The feature data is obtained for each element within a sample range by setting multiple local ranges (sample ranges) within the non-uniformity detection target range (evaluation target range).

[0039] In the example shown in FIG. 3A, for a group of 20,000 capacitors, which is the entire range of the detection target (evaluation target), N sample ranges S are arranged uniformly over the entire range A of the manufacturing time. n (n=1 to N) is set.

[0040] More specifically, the 1,400 capacitors from the a1th manufactured capacitor to the a1+1,399th manufactured capacitor are defined as a first sample range (group) S1. Furthermore, the 1,400 capacitors from the a2th manufactured capacitor to the a2+1,399th manufactured capacitor, spaced apart by a predetermined time T (or a period T for manufacturing a predetermined number of capacitors) from the manufacture of the a1+1,399th manufactured capacitor, are defined as a second sample range (group) S2. In this way, for the 20,000 capacitors manufactured sequentially, N (N=10 in this embodiment) sample ranges S1...S2 are defined. N Each sample range S n The size (number of elements included) of each capacitor (n=1 to N) is the same, and each capacitor contains 1,400 capacitors manufactured in series.

[0041] Sample range S n Each of the elements (n=1 to N) is set to be spaced apart from each other by a predetermined distance or more in the evaluation space. In the example shown in FIG. 3A, the elements (capacitors) are arranged along one index axis, the time axis, so the sample range S nEach of the (n=1 to N) is set (placed) at a predetermined distance on the time axis, that is, at a predetermined time T (≧Tmin) apart.

[0042] However, the sample range S n The arrangement of the sample range S (n=1 to N) is not limited to the example shown in FIG. n If each of the (n=1 to N) is separated from each other by a predetermined time T (≧Tmin) or more, the sample range S n (n=1 to N) may be placed at any position in the evaluation target range A. For example, each sample range S n (n=1 to N) may be randomly arranged within the evaluation range A. Also, for example, the sample range S n (n=1 to N) may not be uniformly arranged within the evaluation range A, but may be arranged in a large number in a range (manufacturing time period) where non-uniformity is predicted to be large. However, when detecting non-uniformity in the evaluation range A where no tendency of non-uniformity is predicted, multiple sample ranges S n (n=1 to N) are preferably arranged uniformly over substantially the entire surface of the evaluation range A.

[0043] Sample range S n The number N (n=1 to N) is 10 in the example of FIG. 3A, but is not limited to 10 and may be any number. For example, when it is predicted that there is little heterogeneity in the feature amount, the sample range S n The set number N (n=1 to N) may be relatively small. If it is predicted that the feature amount is highly heterogeneous, the sample range S n The set number N (n=1 to N) may be set to a relatively large value (for example, 1000).

[0044] Each sample range S set in this way n For each of the sample ranges S (n=1 to N), the feature quantity (capacitance) of each element (capacitor) within the sample range is collected. In the example shown in FIG. 3A, n Since 1400 capacitors are included in the sample range S (n = 1 to N), n The capacitance of each of the 1,400 capacitors (n = 1 to N) is measured. nFor each (n=1 to N), the sample range S n The original feature data Do is the capacitance data of each capacitor included in n (n=1 to N) is obtained.

[0045] Next, the feature amount original data Do obtained in step S1 n (n=1 to N), preprocessing such as flattening is performed, and the preprocessed feature data D n (n=1 to N) is generated (step S2).

[0046] The flattening is performed for each sample range S n The feature quantity original data Do for each element obtained for (n = 1 to N) n (n=1 to N), components such as noise acting on the entire evaluation target range A are removed (offset), and each sample range S n Feature data D in which information on the feature quantities (n=1 to N) is clearly expressed n (n=1 to N).

[0047] In the pre-processing step (step S2), other processing may be performed in addition to flattening. For example, the feature original data Do may be affected by the mechanical characteristics of the feature measurement device used, the measurement environment, etc. n If such an influence appears (is superimposed) on the image signal (n=1 to N), it is desirable to perform a preprocessing step to eliminate such an influence.

[0048] Next, the feature data D obtained in step S2 n (n=1 to N), the sample range S n For each (n = 1 to N), local histogram information Rhg n (n=1 to N) is calculated (generated) (step S3).

[0049] Specifically, the sample range S n For each (n=1 to N), the sample range S n Feature data D n Based on this, the histogram hg is created by using the capacitance as a feature (class value) and the number of capacitors having that feature as a frequency.n Next, the frequency is normalized based on the most frequent value of the frequency (the maximum number of capacitors for each feature) to generate local histogram information Rhg n (n=1 to N) is calculated.

[0050] Furthermore, by performing normalization, in the next step, different sample ranges S n Local histogram information Rhg (n=1 to N) n When accumulating (n = 1 to N) local histogram information for each sample range, the weights of the local histogram information for each sample range can be made approximately equal. If normalization is not performed, the weights of the local histogram information in the accumulated histogram may differ depending on the magnitude of the absolute value (actual frequency) of the frequency of the feature. However, by performing normalization, the contributions of the distributions (local histogram information) of feature amounts in multiple sample ranges can be made equal and accumulated, allowing for appropriate evaluation of the non-uniformity of feature amounts in the evaluation target range.

[0051] Next, the local histogram information Rhg obtained in step S3 n (n=1 to N), a plurality of local histogram information Rhg1, . . . , Rhg n Accumulated histogram information Rrp (n=1 to N) n (n=1 to N) is calculated (step S4).

[0052] Specifically, as shown in FIG. 3B, N pieces of local histogram information Rhg1 to Rhg2 (N=100 in this embodiment) are N For each number (n) from 1 to n, the first to nth local histogram information Rhg1, ..., Rhg n The accumulated results are normalized by the mode, and the accumulated histogram information Rrp corresponding to the accumulated number n is calculated. n (n=1 to N) are calculated. As a result, N pieces of cumulative histogram information Rrp1, ..., Rrp N is obtained.

[0053] Accumulative histogram information Rrp nWhen calculating (n=1 to N), the cumulative result of the local histogram information is normalized by the mode, so that the cumulative histogram information Rrp is calculated in the next step. n When calculating the similarity between (n = 1 to N), the influence of the combination of sample ranges that are the subject of accumulation can be eliminated. In other words, if normalization is not performed, differences in the combination of sample ranges will cause differences in the absolute frequency values ​​of the cumulative histogram information, and these differences in the absolute frequency values ​​may adversely affect the calculated similarity. However, by performing normalization, it is possible to appropriately compare the distribution states of the feature quantities in the cumulative histogram information regardless of the combination of sample ranges, and to appropriately evaluate the non-uniformity of the feature quantities in the sample ranges.

[0054] The cumulative histogram information Rrp calculated in this way n (n=1 to N), all (1st to Nth) local histogram information Rhg1, ..., Rhg N Accumulated histogram information Rrp N is the reference cumulative histogram information that indicates the characteristics of the overall feature amount of the evaluation target range, and as will be described later, each cumulative histogram information Rrp n (n=1 to N-1) is used as a basis for calculating the similarity.

[0055] The cumulative histogram information Rrp1 with a cumulative number of 1 is the same as the initially selected piece of local histogram information Rhg1.

[0056] In FIG. 3B, the sample ranges S1, . . . , S1 described above with reference to FIG. 3A are n , ..., S N Local histogram information Rhg1, ..., Rhg n , ..., Rhg N the cumulative histogram information Rrp n Although the order of the local histogram information accumulated to calculate (n=1 to N) is described, these do not need to be the same. That is, N pieces of accumulated histogram information Rrp n The order of the local histogram information accumulated to calculate (n=1 to N) is N pieces of local histogram information Rhg n(n=1 to N), and a predetermined number (n) of local histogram information Rhg1, ..., Rhg n may be sequentially accumulated.

[0057] Alternatively, the accumulated local histogram information Rhg n The order of (n=1 to N) is not determined, and when accumulating, the local histogram information Rhg n (n=1 to N) is selected appropriately (for example, randomly), and the cumulative histogram information Rrp n (n=1 to N) may be calculated.

[0058] Accumulative histogram information Rrp n (n=1 to N), the accumulated local histogram information Rhg n The number of (n=1 to N) is important, and different local histogram information Rhg n If selecting from (n=1 to N), the local histogram information Rhg is appropriately determined according to the number of accumulated data. n In other words, the arrangement and order of the sample ranges S1 (n=1 to N) described above with reference to FIG. 3A are not important, and any sample range S1 can be selected. n Local histogram information Rhg corresponding to (n=1 to N) n (n=1 to N) may be accumulated sequentially.

[0059] On the other hand, the sample range S that you want to prioritize as the accumulation target n If there is a sample range S (n=1 to N), n Local histogram information Rhg (n=1 to N) n (n=1 to N) are accumulated early so that the accumulated local histogram information Rhg n The order of (n=1 to N) may be set intentionally. n (n = 1 to N) means that the characteristics related to the non-uniformity of the evaluation range are predicted (for example, when a range where non-uniformity is likely to occur is predicted), or when the cumulative number is small, the sample range S is evenly distributed from the entire evaluation range as much as possible. nFor example, if you want to select

[0060] Next, as shown in FIG. 3C, the cumulative histogram information Rrp obtained in step S4 n (n=1 to N) (among these, the cumulative histogram information Rrp generated by accumulating n=N items N is the reference cumulative histogram information) n For each of (n=1 to N-1), the reference cumulative histogram information Rrp N Similarity Hm n (n=1 to N-1) is found (step S5).

[0061] Accumulative histogram information Rrp n (n=1 to N-1) reference cumulative histogram information Rrp N Similarity Hm n (n=1 to N-1) is calculated using formula (1).

[0062]

[0063] The similarity between histograms will now be described with reference to Fig. 5. When the histogram shown in Fig. 5(A) is used as a reference, it is preferable to determine whether the histogram shown in Fig. 5(B) or the histogram shown in Fig. 5(C) is similar to the reference histogram by the area of ​​the overlap when they are superimposed on the reference histogram.

[0064] That is, when the histogram shown in Fig. 5(B) is superimposed on the reference histogram shown in Fig. 5(A), the overlapping portion is shown by the hatched area in Fig. 5(D).On the other hand, when the histogram shown in Fig. 5(C) is superimposed on the reference histogram shown in Fig. 5(A), the overlapping portion is shown by the hatched area in Fig. 5(E).

[0065] 5(D) and 5(E), it is clear that the area of ​​the shaded portion is clearly larger in Fig. 5(D), and the histogram shown in Fig. 5(B) is closer to, i.e., more similar to, the reference histogram shown in Fig. 5(A) than the histogram shown in Fig. 5(C). From this perspective, the above formula (1) is a formula for calculating the proportion of overlap with the reference histogram as the similarity between the histograms.

[0066] Next, the cumulative histogram information Rrp obtained in step S5 n (n=1 to N-1) reference cumulative histogram information Rrp N Similarity Hm n (n=1 to N-1) are sequentially judged from n=1 to n=N to see whether they conform to a predetermined standard, and a number n representing the nth piece of cumulative histogram information that is first detected as conforming to the standard is detected. Then, the detected number n and the reference cumulative histogram information Rrp N and the number N of local histogram information accumulated to generate the index f(n, N), an index f(n, N) indicating the non-uniformity of the feature quantity in the evaluation target range is generated and output (step S6).

[0067] Accumulative histogram information Rrp n (n=1 to N-1) reference cumulative histogram information Rrp N Similarity Hm n (n=1 to N-1) conforms to a predetermined standard. n If satisfies the formula (2), it is determined that the similarity satisfies a predetermined standard.

[0068] Hm n ≧ Hm THL ... (2) However, Hm THL is a predetermined threshold value.

[0069] Therefore, in step S6, for example, as shown in the flowchart of FIG. 3C, starting from the n=1th cumulative histogram information Rrp1, the nth cumulative histogram information Rrp n Reference cumulative histogram information Rrp NSimilarity Hm n is the reference similarity Hm THL It is determined whether the reference similarity Hm THL If it is not equal to or greater than n, the similarity Hm n is the reference similarity Hm THL It is detected whether or not it is equal to or greater than this (step S61).

[0070] This process is repeated to obtain the similarity Hm n is the reference similarity Hm THL If a case where the similarity is equal to or greater than the first detected similarity Hm n In other words, the number "n" representing the nth number of the reference cumulative histogram information Rrp N Similarity Hm n is the reference similarity Hm THL The cumulative histogram information Rrp is n The local histogram information Rhg1, ..., Rhg n Then, the number "n" and the reference cumulative histogram information Rrp N and the number "N" of local histogram information accumulated to generate the index f(n, N), an index f(n, N) indicating the non-uniformity of the feature amount in the evaluation target range is generated and output (step S63).

[0071] This number "n" is the number of N pieces of local histogram information Rhg1, ..., Rhg N The cumulative histogram information (reference cumulative histogram information) Rrp is calculated by accumulating the above values ​​to show the characteristics of the entire feature quantity in the evaluation range. N For n pieces of local histogram information Rhg1, ..., Rhg n The cumulative histogram information Rrp calculated by accumulating n is a number that indicates sufficient similarity.

[0072] Therefore, if n is small, a small number of local histogram information Rhg n The characteristics of the overall feature quantity of the evaluation target range can be expressed by (n=1 to N), and it can be determined that the non-uniformity of the feature quantity is small. nIf n=1 to N is not used, it is not possible to express the characteristics of the feature quantities of the entire range to be evaluated, and therefore it can be determined that the feature quantities are highly non-uniform.

[0073] Note that this number "n" is a numerical value that indicates how much "n" can express the characteristics of the feature to be evaluated relative to N, and therefore the index that indicates the non-uniformity of the feature in the evaluation target range generated based on n is a function f(n, N) based on n and N. The index that indicates the non-uniformity of the feature in the evaluation target range expressed by f(n, N) can be, for example, n / N.

[0074] The significance of the above n and N as indices showing the non-uniformity of the feature amount will be explained with reference to FIGS. 4A and 4B.

[0075] Consider capacitor group α, which was manufactured under the same conditions as evaluation range A shown in FIG. 3A, in which non-uniformity in capacitance was observed, and capacitor group β, which was observed to have low non-uniformity in capacitance. If a histogram of the capacitance (feature value) of the 20,000 capacitors included in these capacitor groups α and β is generated, a histogram similar to that of capacitor groups α and β, such as that shown in FIG. 4A, is obtained. For both capacitor groups α and β, the capacitance values ​​(electrostatic capacitance) of the 20,000 capacitors are distributed between 36 μF and 80 μF, with a mode value of approximately 46 μF to 52 μF. Therefore, even if a histogram of the feature value of capacitance is generated for the entire evaluation range, it is difficult to identify differences between these capacitor groups α and β from the feature value.

[0076] For each of the capacitor groups α and β, the capacitance distribution in the production unit (1,400 capacitors) for each time period is detected, and the process of detecting the non-uniformity described above is performed. n (n=1 to N) n The number of (n=1 to N), i.e., the numerical value "n", and the cumulative histogram information Rrp n (n=1 to N) and reference cumulative histogram information Rrp N Similarity Hm nThe relationship with (n=1 to N) is, for example, as shown in FIG. 4B.

[0077] In the process of detecting the non-uniformity described above, the reference similarity Hm THL = 0.95, that is, the similarity Hm n When the cumulative histogram information Rrp is 0.95 or more, n (n=1 to N) and reference cumulative histogram information Rrp N are defined as being similar, the non-uniformity of the capacitor group α is expressed as n=5 for N=10, while the non-uniformity of the capacitor group β is expressed as n=2 for N=10.

[0078] 4B , for capacitor group α, in which non-uniformity in capacitance was observed, a cumulative count n close to the maximum value of 5 was required for the similarity to converge to close to 1, whereas for capacitor group β, in which low non-uniformity in capacitance was observed, the similarity converged to approximately 1 before the cumulative count n reached 2. That is, for capacitor group α, it was necessary to accumulate feature quantities from five sample ranges (each containing 1,400 capacitors) to represent the distribution of feature quantities over the entire evaluation range (20,000 capacitors), whereas for capacitor group β, accumulating feature quantities from two sample ranges was sufficient to represent the distribution of feature quantities over the entire evaluation range, and the numerical values ​​clearly show that capacitor group β has low non-uniformity in feature quantities.

[0079] In this way, by using the indices n and N calculated by the feature non-uniformity detection method, detection device, etc. according to the present disclosure, non-uniformity in the distribution of feature quantities that cannot be detected simply by using raw feature quantity data can be quantified numerically and objectively and clearly detected and evaluated.

[0080] The method for detecting (evaluating) non-uniformity in feature quantities according to the present disclosure can also be implemented by a program configured to cause a computer to execute each process (step).

[0081] Furthermore, as described above, the feature non-uniformity detection (evaluation) method according to the present disclosure can also be achieved by, for example, the device shown in Fig. 2. Fig. 2 is a block diagram showing an example of the configuration of a feature non-uniformity detection device according to an embodiment of the present disclosure. As shown in Fig. 2, the feature non-uniformity detection device 1 includes a calculation processing unit 10, a storage unit 20, an input I / F unit 31, and an output I / F unit 32. The calculation processing unit 10 also includes a preprocessing unit 11, a local histogram information calculation unit 12, a cumulative histogram information calculation unit 13, a similarity calculation unit 14, and a non-uniformity detection unit 15.

[0082] The preprocessing unit 11, local histogram information calculation unit 12, cumulative histogram information calculation unit 13, similarity calculation unit 14, and non-uniformity detection unit 15 of the arithmetic processing unit 10 perform processing corresponding to the preprocessing step (step S2), local histogram information calculation step (step S3), cumulative histogram information calculation step (step S4), similarity calculation step (step S5), and non-uniformity detection step (step S6) of the method according to the present disclosure described above, respectively.

[0083] The functions of each processing unit are executed by driving the arithmetic processing unit 10 based on a program stored in the storage unit 20. Data calculated / generated by the execution of processing in each processing unit is stored in the storage unit 20 as appropriate, and is referenced by other processing units, and is also output via the output I / F unit 32.

[0084] The processing of the feature original data collection step (step S1) of the method according to the present disclosure described above is performed by a feature collection device 50, which is a device separate from the feature nonuniformity detection device 1. Specifically, the feature collection device 50 is, for example, a feature measurement device that actually measures the capacitance of manufactured capacitors.

[0085] However, the feature collection device 50 does not have to be a device that actually measures features. For example, the feature collection device 50 may be configured as an input device that simply inputs the pre-measured capacitance of each capacitor. Alternatively, the feature collection device 50 may be configured as a storage device that simply stores the pre-measured capacitance of each capacitor. The I / F for data transfer between the feature collection device 50 and the feature non-uniformity detection device 1 may be configured, for example, so that data is transferred from the feature collection device 50 to the feature non-uniformity detection device 1 in response to a trigger (e.g., a ready signal) from the feature collection device 50, or so that the feature non-uniformity detection device 1 reads data from the feature collection device 50 as an I / O device. Any suitable method / configuration may be used.

[0086] The configuration of the feature non-uniformity detection device 1 itself is not limited to the configuration shown in Fig. 2. For example, a configuration may be adopted in which part of the processing executed by the feature non-uniformity detection device 1 is performed by the feature collection device 50. Specifically, for example, processing in the pre-processing unit 11, such as flattening of measured data, or further processing in the local histogram calculation unit 12, may be performed on the feature collection device 50 side by performing such processing collectively when storing measured features.

[0087] As described above, the feature non-uniformity detection method and feature non-uniformity detection apparatus of this embodiment can quantify the non-uniformity of the capacitance of capacitors numerically and objectively detect and evaluate it. As a result, it is possible to reduce the variation in quality (non-uniformity) when mass-producing capacitors, and to appropriately supply capacitors with uniform quality.

[0088] The method and the like disclosed herein are not limited to the case of detecting non-uniformity in the capacitance of a capacitor, but can be applied to any feature, any element, and any evaluation target (distribution, dimension).

[0089] For example, the methods disclosed herein can be applied to distribution information (histogram information) of any feature, such as physical quantities that indicate the characteristics of any product, including industrial and mining products including electrical / electronic products and mechanical products, natural products, and foods; feature quantities such as the roughness and precision between samples to be inspected; numerical values ​​(i.e., feature quantities) that are arbitrarily defined as indicating the characteristics, actions, or effects of these; and feature quantities in various statistical / analytical / evaluation data used for information analysis of social conditions, corporate performance, population trends, etc.; and the nonuniformity can be quantified numerically and objectively detected and evaluated.

[0090] More specifically, the method etc. disclosed herein can quantify and objectively detect and evaluate the nonuniformity of the electrical and magnetic properties of various electronic components such as resistors, inductors, and capacitors; the sensitivity of various sensors; the resistance values ​​and transmission characteristics of connectors, switches, cables, etc.; the film thickness of film products such as PET film and MLCC green sheet; characteristic values ​​such as the efficiency, power factor, and fuel efficiency of power sources or power modules of motors, generators, engines, etc.; the amount of ingredients contained in food etc.; the age-based population ratio for each region as described above; sales or income by item for each sales office; the distribution of scores by class for each test; the distribution of weights of agricultural products etc. for each unit of arrival; the distribution of diameters and circularity of crystal grains exposed on a two-dimensional surface; the distribution of shapes in an image; the distribution of brightness values ​​in a grayscale image; the distribution of materials in a large-scale material storage space; and the like.

[0091] It should be noted that the method etc. of the present disclosure may exclude evaluation of the non-uniformity of the surface roughness of an object.

[0092] By applying the method according to the present disclosure to such evaluation targets, for example, at an article manufacturing site, it becomes possible to compare the feature quantities and quality non-uniformity between products, articles, components, and materials. As a result, it is possible to reduce the quality variation (non-uniformity) when mass-producing products, and appropriately supply products of uniform quality. For example, at an electronic component or electronic equipment manufacturing site, it becomes possible to mass-supply products with uniform characteristics.

[0093] Furthermore, if the non-uniformity of the feature quantities indicates deterioration in the function or effect or inappropriateness / incompatibility of each of the above-mentioned evaluation targets (devices, machines, instruments, and various systems (including various social systems)), detecting the non-uniformity of the feature quantities using the method disclosed herein makes it possible to appropriately know the performance and status (deterioration state, durability, compatibility, etc.) of the device or system, or changes in the processing target, and enables subsequent appropriate and prompt response.

[0094] The method and the like disclosed herein can be implemented in various forms. For example, they may be realized as a standalone device, i.e., an independent "feature non-uniformity detection device." Such a device may be realized as a dedicated device using an electrical circuit composed of basic electronic components and a programmable semiconductor device or so-called microcomputer, or may be realized by downloading a program to a general-purpose computer and executing it.

[0095] Furthermore, such a "detection device" may be incorporated into, for example, a processing device (such as a manufacturing device or an inspection device) that handles elements to be evaluated, and the method of the present disclosure may be executed. Furthermore, the method of the present disclosure may be realized in an environment (system) in which multiple devices (including dedicated devices and computers) are connected via a network, in other words, in a server-client system, a network system using a cloud environment, or a distributed processing environment including these. In such a case, the method of the present disclosure may be realized by distributing the program of the present disclosure to each node.

[0096] Examples

[0039] Other specific examples of evaluation targets to which the methods and the like of the present disclosure can be applied, which are different from the non-uniformity of capacitance of capacitors described above, will be described below as examples of the present disclosure. In this example, detection of non-uniformity in the grain area (crystal grain area) of dielectric grains in a dielectric layer of a specific type of multilayer ceramic capacitor (MLCC) product will be described.

[0097] To evaluate the non-uniformity of the dielectric particle area of ​​an MLCC, first, the MLCC is cut in a direction perpendicular to the lamination direction of the dielectric layers and electrode layers so that one of the dielectric layers is exposed, or the MLCC is polished so that the lamination surface of one of the dielectric layers is exposed. Next, the cross section of the obtained dielectric layer (the cut surface or the polished exposed surface) is scanned with an SEM (scanning electron microscope) to measure the particle area distribution of the dielectric particles (feature amount original data Do n (n=1 to N, N is the number of MLCC samples, 18 in this embodiment) is obtained. At this time, a process of emphasizing the crystal grain boundaries of the dielectric material by image analysis may be performed to detect the dielectric particles and the crystal area.

[0098] Next, preprocessing (flattening) such as noise removal (removal of influences based on the characteristics of the SEM, etc., the measurement environment, etc.) is performed on the obtained particle area distribution, and a histogram of the particle area of ​​the dielectric particles, for example, as shown in FIG. 6A is obtained (local histogram information Rhg n (n=1 to N)). Fig. 6A is a histogram of the particle area of ​​the dielectric particles of one MLCC. Such a particle area histogram (local histogram information Rhg n (n=1 to N)) is calculated for each of the N sample MLCCs.

[0099] Next, the obtained N pieces of local histogram information Rhg n (n=1 to N), a plurality of local histogram information Rhg1, . . . , Rhg n Accumulated histogram information Rrp (n=1 to N) n (n=1 to N) and calculate the cumulative histogram information Rrp n For each of (n=1 to N-1), the reference cumulative histogram information Rrp N Similarity Hm n (n=1 to N-1).

[0100] As a result of the detection process of the non-uniformity of the particle area of ​​the dielectric particles, the cumulative histogram information Rrp as shown in FIG. n (n=1 to N)n The cumulative degree is the number of (n=1 to N) (i.e., the number "n") and the cumulative histogram information Rrp n (n=1 to N) and reference cumulative histogram information Rrp N Similarity Hm n (n=1 to N).

[0101] The cumulative degree n (n = 1 to N) and the similarity Hm n In the relationship (n=1 to N), if the reference similarity Hm THL = 0.95, that is, the similarity Hm n When the cumulative histogram information Rrp is 0.95 or more, n (n=1 to N) and reference cumulative histogram information Rrp N are defined as being similar, the non-uniformity of the particle area of ​​the dielectric particles can be expressed as n=8 for N=18.

[0102] As described above, according to this embodiment, it is possible to numerically express the non-uniformity of the particle area of ​​the dielectric particles in the dielectric layer of a group of MLCC products. As a result, it is possible to compare the non-uniformity of the particle area for each group of MLCC products numerically, and it is possible to appropriately evaluate product manufacturing conditions such as design, manufacturing process, and lot size by numerically evaluating the characteristics of the MLCC.

[0103] REFERENCE SIGNS LIST 1... feature non-uniformity detection device 10... arithmetic processing unit 11... pre-processing unit 12... local histogram information calculation unit 13... cumulative histogram information calculation unit 14... similarity calculation unit 15... non-uniformity detection unit 20... storage unit 31... input I / F unit 32... output I / F unit 50... feature collection device

Claims

1. A feature evaluation method for evaluating heterogeneity of a feature in an evaluation target range which is a set of elements having a predetermined feature, comprising the steps of: obtaining local histogram information indicating a relationship between the feature and a frequency within each of a plurality of local sample ranges set in the evaluation target range; accumulating the 1st to Nth local histogram information for each of the sample ranges, successively from the 1st to the Nth local histogram information, to generate a plurality of cumulative histogram information accumulated over n (n=1 to N) pieces; setting the cumulative histogram information generated by accumulating over n=N pieces as reference cumulative histogram information, and detecting a similarity to the reference cumulative histogram information for each of the plurality of cumulative histogram information generated by successively accumulating the local histogram information from n=1 to N; and successively detecting whether the detected similarity satisfies a predetermined criterion, from n=1 to n=N, and using a number n representing the nth cumulative histogram information that first satisfies the criterion, as part of an index indicating heterogeneity of the feature.

2. A feature evaluation method as described in claim 1, wherein the elements are classified so as to be arranged in a d-dimensional space having d index axes (d=1, 2, 3, or 4 or more), the evaluation target range is a predetermined spatial range in the d-dimensional space, and the sample range is a local spatial range included in the evaluation target range in the d-dimensional space.

3. A feature evaluation method according to claim 2, wherein the plurality of sample ranges are set at a distance in the d-dimensional space that is equal to or greater than a predetermined distance from each other.

4. A feature evaluation method according to any one of claims 1 to 3, wherein in the step of generating local histogram information, local histogram information normalized based on the most frequent frequency is generated.

5. A feature evaluation method according to any one of claims 1 to 3, wherein in the step of generating cumulative histogram information, cumulative histogram information normalized based on the most frequent frequency is generated.

6. The feature evaluation method according to any one of claims 1 to 3, wherein in the step of using the number n representing the nth position as a part of an index showing the non-uniformity of the feature, if the detected similarity satisfies formula (2), it is determined that the similarity complies with the predetermined criterion. Hm n ≧Hm THL ... (2) where Hm n is the similarity between the n-th cumulative histogram information and the reference cumulative histogram information, Hm THL is a predetermined similarity threshold, 7. A feature evaluation device for evaluating heterogeneity of a feature in an evaluation target range which is a set of elements having a predetermined feature, comprising: means for obtaining local histogram information indicating a relationship between the feature and frequency within each of a plurality of local sample ranges set in the evaluation target range; means for accumulating the 1st to Nth local histogram information for each of the sample ranges, successively from the 1st to the Nth local histogram information, to generate a plurality of cumulative histogram information accumulated by n (n = 1 to N) pieces of information; means for determining, as reference cumulative histogram information, the cumulative histogram information generated by accumulating the n=N pieces of local histogram information, and detecting a similarity to the reference cumulative histogram information for each of the plurality of cumulative histogram information generated by successively accumulating the local histogram information from n=1 to N; and means for successively detecting, from n=1 to n=N, whether the detected similarity satisfies a predetermined criterion, and using a number n representing the nth cumulative histogram information that first satisfies the criterion, as a part of an index indicating heterogeneity of the feature.

8. A feature evaluation device as described in claim 7, wherein the elements are classified so as to be arranged in a d-dimensional space having d index axes (d=1, 2, 3, or 4 or more), the evaluation target range is a predetermined spatial range in the d-dimensional space, and the sample range is a local spatial range included in the evaluation target range in the d-dimensional space.

9. A feature evaluation device according to claim 8, wherein the plurality of sample ranges are set such that the distance between each of the sample ranges in the d-dimensional space is equal to or greater than a predetermined distance.

10. A feature evaluation device according to any one of claims 7 to 9, wherein the means for generating local histogram information generates local histogram information normalized based on the most frequent frequency.

11. A feature evaluation device according to any one of claims 7 to 9, wherein said cumulative histogram information generating means generates cumulative histogram information normalized with respect to the most frequent frequency.

12. The feature evaluation device according to any one of 7 to 9, wherein the means for using the number n representing the nth position as a part of an index showing the non-uniformity of the feature determines that the detected similarity satisfies the predetermined criterion when the detected similarity satisfies the formula (2). Hm n ≧Hm THL ... (2) where Hm n is the similarity between the n-th cumulative histogram information and the reference cumulative histogram information, Hm THL is a predetermined similarity threshold, 13. A system having a plurality of devices, comprising: a process for evaluating non-uniformity of a feature amount for an evaluation target range, which is a set of elements having a predetermined feature amount, by executing a program in the plurality of devices, comprising: a step of obtaining local histogram information indicating a relationship between the feature amount and a frequency within each of a plurality of local sample ranges set in the evaluation target range; a step of accumulating the first to Nth local histogram information for each of the sample ranges, successively from the first to the Nth local histogram information, to generate a plurality of pieces of cumulative histogram information accumulated over n (n=1 to N); a step of determining the cumulative histogram information generated by accumulating n=N pieces as reference cumulative histogram information, and detecting a similarity to the reference cumulative histogram information for each of the plurality of pieces of cumulative histogram information generated by successively accumulating the local histogram information from n=1 to N; detecting whether the detected similarity satisfies a predetermined criterion, sequentially from n=1 to n=N, and using a number n representing the n-th piece of cumulative histogram information that first satisfies the criterion as part of an index of non-uniformity of the feature.

14. A feature evaluation program for evaluating nonuniformity of a feature for an evaluation target range which is a set of elements having a predetermined feature, comprising: a step of obtaining local histogram information indicating a relationship between the feature and a frequency within each of a plurality of local sample ranges set in the evaluation target range; a step of accumulating the first to Nth local histogram information for each of the sample ranges from the first to the Nth sample ranges to generate a plurality of cumulative histogram information accumulated over n (n=1 to N) pieces; a step of using the cumulative histogram information generated by accumulating n=N pieces as reference cumulative histogram information, and detecting a similarity to the reference cumulative histogram information for each of the plurality of cumulative histogram information generated by accumulating the local histogram information sequentially from n=1 to N; and a step of detecting whether the detected similarity satisfies a predetermined criterion, sequentially from n=1 to n=N, and using a number n representing the nth cumulative histogram information that first satisfies the criterion, as a part of an index indicating nonuniformity of the feature. A feature evaluation program for causing a computer to execute the above.

15. A storage medium storing a feature evaluation program for evaluating non-uniformity of a feature for an evaluation target range which is a set of elements having a predetermined feature, the feature evaluation program comprising: a step of obtaining local histogram information indicating a relationship between the feature and a frequency within each of a plurality of local sample ranges set in the evaluation target range; a step of accumulating the first to Nth local histogram information for each of the sample ranges from the first to the Nth sample ranges in sequence from the first to the Nth sample ranges to generate a plurality of cumulative histogram information accumulated over n (n=1 to N) pieces; a step of setting the cumulative histogram information generated by accumulating the n=N pieces as reference cumulative histogram information, and detecting a similarity to the reference cumulative histogram information for each of the plurality of cumulative histogram information generated by accumulating the local histogram information in sequence from n=1 to N; and a step of detecting whether the detected similarity satisfies a predetermined criterion from n=1 to n=N in sequence, and using a number n representing the nth cumulative histogram information that first satisfies the criterion as a part of an index indicating non-uniformity of the feature. A storage medium that is a feature evaluation program for causing a computer to execute the above.

Images