Sequence data analysis device, sequence data analysis method, and sequence data analysis program
阅读说明：本技术 序列数据分析装置、序列数据分析方法以及序列数据分析程序 (Sequence data analysis device, sequence data analysis method, and sequence data analysis program ) 是由 清水尚吾 草野胜大 奥村诚司 于 2018-02-06 设计创作，主要内容包括：在序列数据分析装置(10)中,计算部(21)比较作为观测某事件而得到的时间序列数据的输入序列数据(32)与作为关于该事件的标准时间序列数据的参照序列数据(31),提取相互对应的输入序列数据(32)的数据元素与参照序列数据(31)的数据元素的组合。计算部(21)按照提取出的每个组合,计算输入序列数据(32)中的事件的进展相对于参照序列数据(31)中的事件的进展的相对速度即相对进展速度。判定部(22)根据由计算部(21)计算出的相对进展速度,判定输入序列数据(32)中的事件的进展是否存在异常。(In a sequence data analysis device (10), a calculation unit (21) compares input sequence data (32) that is time-series data obtained by observing an event with reference sequence data (31) that is standard time-series data regarding the event, and extracts a combination of data elements of the input sequence data (32) and data elements of the reference sequence data (31) that correspond to each other. A calculation unit (21) calculates a relative progress rate, which is a relative speed of the progress of the event in the input sequence data (32) with respect to the progress of the event in the reference sequence data (31), for each of the extracted combinations. A determination unit (22) determines whether or not there is an abnormality in the progress of an event in the input sequence data (32) on the basis of the relative progress rate calculated by the calculation unit (21).)
1. A sequence data analysis device is provided with:
a calculation unit that compares input sequence data that is time-series data obtained by observing an event with reference sequence data that is standard time-series data regarding the event, extracts combinations of data elements of the input sequence data and data elements of the reference sequence data that correspond to each other, and calculates a relative progress rate that is a relative speed of a progress of the event in the input sequence data with respect to a progress of the event in the reference sequence data for each of the extracted combinations; and
and a determination unit that determines whether or not there is an abnormality in the progress of the event in the input sequence data, based on the relative progress rate calculated by the calculation unit.
2. The sequence data analysis device of claim 1, wherein,
the calculation unit calculates a normalized path between the input sequence data and the reference sequence data, and extracts an element of the calculated normalized path as a combination of a data element of the input sequence data and a data element of the reference sequence data.
3. The sequence data analysis device according to claim 1 or 2, wherein,
the input sequence data is data of a video, and each frame of the video is processed as a data element of the input sequence data.
4. The sequence data analysis device according to any one of claims 1 to 3, wherein,
the determination unit determines whether or not the abnormality is present, based on the relative progress speed and whether or not consecutive data elements of the reference sequence data correspond to the same data element of the input sequence data in 2 or more combinations extracted by the calculation unit.
5. The sequence data analysis device according to any one of claims 1 to 4, wherein,
the judgment unit judges whether or not there is an abnormality in which data in a section corresponding to a certain section is missing from the input sequence data, based on a relative speed with respect to the progress of the event in data in the section included in the reference sequence data.
6. The sequence data analysis device according to any one of claims 1 to 5, wherein,
the determination unit determines whether or not the abnormality is present, based on the relative progress speed and whether or not consecutive data elements of the input sequence data correspond to the same data element of the reference sequence data in 2 or more combinations extracted by the calculation unit.
7. The sequence data analysis device according to any one of claims 1 to 6, wherein,
the judgment unit judges whether or not there is an abnormality that the data of a certain section is data other than the standard data, based on the relative speed of the progress of the event in the data of the section included in the input sequence data.
8. The sequence data analysis device according to any one of claims 1 to 7, wherein,
the judgment unit corrects the input sequence data by deleting data of a 2 nd section from the input sequence data and inserting the deleted data as data of a section corresponding to the 1 st section into the input sequence data when judging that there is an abnormality that data of a section corresponding to the 1 st section included in the reference sequence data is missing from the input sequence data and an abnormality that data of a 2 nd section included in the input sequence data is data other than standard.
9. The sequence data analysis device of claim 8, wherein,
the calculation unit calculates a 1 st similarity that is a similarity between the input sequence data and the reference sequence data before the correction by the determination unit and a 2 nd similarity that is a similarity between the input sequence data and the reference sequence data after the correction by the determination unit,
the judgment unit compares the 1 st similarity with the 2 nd similarity, and judges that there is an abnormality such as a sequence change in the event in the input sequence data when the 2 nd similarity is higher than the 1 st similarity.
10. A method for sequence data analysis, wherein,
the calculation unit compares input sequence data that is time-series data obtained by observing an event with reference sequence data that is standard time-series data regarding the event, extracts combinations of data elements of the input sequence data and data elements of the reference sequence data that correspond to each other, and calculates a relative progress rate that is a relative speed of a progress of the event in the input sequence data with respect to a progress of the event in the reference sequence data for each of the extracted combinations,
the judgment unit judges whether or not there is an abnormality in the progress of the event in the input sequence data, based on the relative progress speed calculated by the calculation unit.
11. A sequence data analysis program that causes a computer to execute:
a calculation process of comparing input sequence data that is time-series data obtained by observing an event with reference sequence data that is standard time-series data regarding the event, extracting combinations of data elements of the input sequence data and data elements of the reference sequence data that correspond to each other, and calculating a relative progress rate that is a relative speed of a progress of the event in the input sequence data with respect to a progress of the event in the reference sequence data for each of the extracted combinations; and
a determination process of determining whether there is an abnormality in the progress of the event in the input sequence data, based on the relative progress speed calculated by the calculation process.
Technical Field
The present invention relates to a sequence data analysis device, a sequence data analysis method, and a sequence data analysis program.
Background
Patent document 2 describes the following method: a normal model of the waveform is learned based on an extreme value generated in a waveform of the time series of the sensor data and a generation timing of the extreme value, the normal model is compared with the sensor data of the diagnostic object, and it is determined whether the sensor data of the diagnostic object is abnormal or normal.
Patent document 3 describes the following method: a reference score is obtained from the reference data and a feature coefficient extracted from the time series of the reference data, a target score is obtained from the target data and a feature coefficient extracted from the time series of the target data, and the reference score and the target score are compared to determine whether the target data is abnormal or normal.
Disclosure of Invention
Problems to be solved by the invention
In a production site of a factory, there is a demand for finding an abnormal operation of an operator such as "use of a dropped member", "forget to tighten a screw", and "different operation procedure" for quality control. However, the detection of abnormal operation is currently performed by visually checking a monitoring image, and a load on a monitor is high. Therefore, a technique for automatically detecting an abnormal operation scene from an operation image by an image analysis technique is required.
In a work image of a factory, a series of work operation patterns are repeated. Therefore, by comparing the reference work image and the monitor image and finding out the difference, it is possible to automatically detect the abnormal operation of the operator. Here, if the analysis of the work image is interpreted as the analysis of the series data, "use of a dropped component", "forget to tighten a screw", and "different work order" given as the detection target scene may be interpreted as "insertion of series data other than the standard", "absence of series data", and "replacement of series data", respectively. Therefore, a sequence data analysis technique capable of detecting an abnormality of these sequence data in units of frames is required. However, in the conventional technique, it is impossible to detect an abnormality of the sequence data in units of frames.
The purpose of the present invention is to detect an abnormality in sequence data in units of data elements.
Means for solving the problems
A sequence data analysis device according to an aspect of the present invention includes:
a calculation unit that compares input sequence data that is time-series data obtained by observing an event with reference sequence data that is standard time-series data regarding the event, extracts combinations of data elements of the input sequence data and data elements of the reference sequence data that correspond to each other, and calculates a relative progress rate that is a relative speed of a progress of the event in the input sequence data with respect to a progress of the event in the reference sequence data for each of the extracted combinations; and
and a determination unit that determines whether or not there is an abnormality in the progress of the event in the input sequence data, based on the relative progress rate calculated by the calculation unit.
Effects of the invention
In the present invention, a relative progress velocity, which is a relative velocity of progress of an event in input sequence data with respect to progress of an event in reference sequence data, is calculated for each combination of data elements of the input sequence data and data elements of the reference sequence data that correspond to each other. Then, it is determined whether there is an abnormality in the progress of the event in the input sequence data, based on the relative progress speed. Therefore, according to the present invention, it is possible to detect an abnormality in input sequence data in units of data elements.
Drawings
Fig. 1 is a block diagram showing the configuration of a sequence data analyzer according to
Fig. 2 is a flowchart showing the operation of the sequence data analyzer according to
Fig. 3 is a flowchart showing the procedure of the regular path calculation processing according to
Fig. 4 is a diagram showing an example of a regular path calculated by the regular path calculation process according to
Fig. 5 is a flowchart showing the procedure of the relative progress speed calculation process according to
Fig. 6 is a diagram showing an example of the slope of the warping path calculated by the relative progress rate calculation process according to
Fig. 7 is a diagram showing an example of the relative progress rate calculated by the relative progress rate calculation process according to
Fig. 8 is a flowchart showing the procedure of the sequence data missing section detection process according to
Fig. 9 is a diagram showing an example of a sequence data missing section detected by the sequence data missing section detection process according to
Fig. 10 is a flowchart showing the procedure of the out-of-standard sequence data insertion section detection process according to
Fig. 11 is a diagram showing an example of an out-of-standard sequence data insertion section detected by the out-of-standard sequence data insertion section detection process in
Fig. 12 is a flowchart showing the sequence data replacement section detection processing procedure in
Fig. 13 is a diagram showing an example of a sequence data replacement section detected by the sequence data replacement section detection process according to
Fig. 14 is a diagram showing an example of the abnormality information recorded by the abnormality information recording process according to
Fig. 15 is a diagram showing an example of the abnormality information recorded by the abnormality information recording process according to
Fig. 16 is a block diagram showing the configuration of a sequence data analyzer according to a modification of
Detailed Description
Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the drawings, the same or corresponding portions are denoted by the same reference numerals. In the description of the embodiments, the same or corresponding portions will be omitted or simplified as appropriate. The present invention is not limited to the embodiments described below, and various modifications may be made as necessary. For example, the embodiments described below may be partially implemented.
- 上一篇：一种医用注射器针头装配设备
- 下一篇：工厂设备监视控制系统以及工厂设备监视控制方法