Anomaly detection device, anomaly detection method, and anomaly detection program

Abstract

An object detection unit (120) detects a region, as a current defect region, where a defect of an inspection object is inferred as being visible by performing object detection on the current object image obtained by imaging the inspection object, and determines the accuracy of inference for the current defect region as the current region reliability. An inspection value calculation unit (140) calculates the current inspection value using the inspection coefficient used for calculating an inspection value and the current region reliability. An abnormality detection unit (150) determines whether an abnormality has occurred in the inspection object on the basis of the current inspection value.

Description

[Technical Field]

[0001] The present disclosure relates to a technique for detecting anomalies in an object using an image. [Background technology]

[0002] In safety inspection patrols such as infrastructure maintenance, high recall (tendency to overdetect) is common to reduce oversights. However, the trade-off is that unwanted objects such as dead leaves or fallen objects may be detected (false positives).

[0003] To reduce false positives, there is a method of accumulating inference reliability (score) and making a judgment. However, this approach carries the risk of delayed detection.

[0004] On the other hand, methods have been proposed to solve the detection delay. Patent Document 1 proposes the following method: In the method of Patent Document 1, a score is added when there is a sudden change in features between successive images in a time series, and when the score reaches or exceeds a threshold, it is deemed that an object has been detected as popping out. However, with this method, the presence of an object that was not present at the time of the previous inspection is considered a sudden change, so the presence of dead leaves, etc., adds to the score, causing the score to exceed the threshold and resulting in a false positive. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Laid-Open No. 2016-12226 Summary of the Invention [Problem to be solved by the invention]

[0006] The present disclosure aims to suppress false detection of defects in an inspection object and enable detection of abnormalities in the inspection object. [Means for solving the problem]

[0007] The abnormality detection device of the present disclosure includes: an object detection unit that performs object detection on a current target image obtained by capturing an image of the inspection target, detects an area inferred to contain a defect in the inspection target as a current defect area, and calculates the accuracy of the inference for the current defect area as a current area reliability; an inspection value calculation unit that calculates a current inspection value using an inspection coefficient used in calculating the inspection value and the current region reliability; an abnormality detection unit that determines whether an abnormality has occurred in the test object based on the current test value; Equipped with. [Effects of the Invention]

[0008] According to the present disclosure, it is possible to detect abnormalities in an inspection object while suppressing false detection of defects in the inspection object. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a configuration diagram of an abnormality detection device 100 according to a first embodiment. [Figure 2] FIG. 1 shows an example of the configuration of an anomaly detection system 200 according to the first embodiment. [Figure 3] 3 is a flowchart of an abnormality detection method according to the first embodiment. [Figure 4] FIG. 2 is a diagram for explaining the ingenuity of the first embodiment. [Figure 5] 10 is a flowchart of an abnormality detection method according to the second embodiment. [Figure 6] FIG. 10 is a diagram for explaining the ingenuity of the second embodiment. [Figure 7] 11 is a flowchart of an abnormality detection method according to the third embodiment. [Figure 8] FIG. 10 is a diagram for explaining the ingenuity of the third embodiment. [Figure 9] FIG. 1 is a hardware configuration diagram of an abnormality detection device 100 according to an embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0010] In the embodiments and drawings, the same or corresponding elements are denoted by the same reference numerals. The description of elements denoted by the same reference numerals as those already described will be omitted or simplified as appropriate. Arrows in the drawings primarily indicate the flow of data or the flow of processing.

[0011] Embodiment 1 The abnormality detection device 100 will be described with reference to FIGS.

[0012] ***Configuration Description*** The configuration of the abnormality detection device 100 will be described with reference to FIG. The abnormality detection device 100 is a computer that includes hardware such as a processor 101, a memory 102, an auxiliary storage device 103, a communication device 104, and an input / output interface 105. These pieces of hardware are connected to one another via signal lines.

[0013] The processor 101 is an IC that performs arithmetic processing and controls other hardware. For example, the processor 101 is a CPU, a DSP, or a GPU. IC is an abbreviation for Integrated Circuit. CPU is an abbreviation for Central Processing Unit. DSP is an abbreviation for Digital Signal Processor. GPU is an abbreviation for Graphics Processing Unit.

[0014] The memory 102 is a volatile or non-volatile storage device. The memory 102 is also called a primary storage device or a main memory. For example, the memory 102 is a RAM. Data stored in the memory 102 is saved in the secondary storage device 103 as needed. RAM is an abbreviation for Random Access Memory.

[0015] The auxiliary storage device 103 is a non-volatile storage device. For example, the auxiliary storage device 103 is a ROM, a HDD, a flash memory, or a combination thereof. Data stored in the auxiliary storage device 103 is loaded into the memory 102 as needed. ROM is an abbreviation for Read Only Memory. HDD is an abbreviation for Hard Disk Drive.

[0016] The communication device 104 is a receiver and a transmitter. For example, the communication device 104 is a communication chip or a NIC. Communication of the anomaly detection device 100 is performed using the communication device 104. NIC is an abbreviation for Network Interface Card.

[0017] The input / output interface 105 is a port to which an input device and an output device are connected. For example, the input / output interface 105 is a USB terminal, the input devices are a keyboard and a mouse, and the output device is a display. Input and output of the abnormality detection device 100 is performed using the input / output interface 105. USB is an abbreviation for Universal Serial Bus.

[0018] Anomaly detection device 100 includes elements such as an image acquisition unit 110, an object detection unit 120, a coefficient determination unit 130, an inspection value calculation unit 140, and anomaly detection unit 150. These elements are realized by software.

[0019] Auxiliary storage device 103 stores an anomaly detection program for causing a computer to function as image acquisition unit 110, object detection unit 120, coefficient determination unit 130, inspection value calculation unit 140, and anomaly detection unit 150. The anomaly detection program is loaded into memory 102 and executed by processor 101. The auxiliary storage device 103 also stores an OS. At least a part of the OS is loaded into the memory 102 and executed by the processor 101. The processor 101 executes an abnormality detection program while running the OS. OS is an abbreviation for Operating System.

[0020] Input and output data of the abnormality detection program is stored in the storage unit 190. The memory 102 functions as the storage unit 190. However, a storage device such as the auxiliary storage device 103, a register in the processor 101, or a cache memory in the processor 101 may function as the storage unit 190 instead of or together with the memory 102.

[0021] The anomaly detection program can be recorded (stored) in a computer-readable manner on a non-volatile recording medium such as an optical disk or flash memory.

[0022] FIG. 2 shows an example of the configuration of an anomaly detection system 200. The anomaly detection system 200 is a system that uses the anomaly detection device 100 . The anomaly detection system 200 is implemented in, for example, a mobile object 201 . An example of the moving object 201 is a car.

[0023] The moving object 201 includes an abnormality detection device 100 , an imaging device 210 , a storage device 220 , and a positioning device 230 . The anomaly detection device 100 communicates with each of the imaging device 210 and the storage device 220 . The imaging device 210 is a device for capturing an image of an object to be inspected. The imaging device 210 captures an image at each date and time and outputs the image. An image obtained by capturing an image of an inspection target is called a target image. The inspection object is an object that is inspected using the object image, for example, the inspection object may be a road, a wall in a tunnel, or a bridge. The storage device 220 is a storage device provided outside the abnormality detection device 100. For example, the storage device 220 stores driving history data and the like. The travel history data is data indicating the position (travel position) of the mobile object 201 at each date and time. The positioning device 230 is a device that measures the position of the mobile object 201. The positioning device 230 performs positioning at each date and time, and registers the travel position at each date and time in the travel history data in the storage device 220.

[0024] ***Explanation of Operation*** The operation procedure of the anomaly detection device 100 corresponds to an anomaly detection method, and also corresponds to a processing procedure by an anomaly detection program.

[0025] The abnormality detection method will be described with reference to FIG. Steps S110 to S150 are repeatedly executed.

[0026] In step S110, the image acquisition unit 110 acquires the current target image. The current target image is the target image used in the current examination. The current target image can be acquired by any method. For example, the image acquisition unit 110 receives the latest image output from the image capture device 210 as the current target image.

[0027] The inspection target shown in the current target image (the inspection target imaged this time) is referred to as the current inspection target.

[0028] In step S120, the object detection unit 120 detects the current defective area by executing object detection on the current target image. The defect area is the area in the target image that is inferred to contain a defect in the inspection target. The current defect area is the area that is inferred to contain a defect that occurred on the surface of the inspection target at the time the current target image was obtained (current imaging time). Defects are generated continuously over time. Their location remains constant over time, but they grow larger as time passes. Defects are also generated over a long period of time (e.g., one or several years). Examples of defects are cracks and holes. Object detection can be performed by any method. For example, the object detection unit 120 performs object detection using an object detection model with the current target image as input. The object detection model is a trained model for object detection. An example of an object detection algorithm is YOLO, which stands for You Only Look Once.

[0029] Furthermore, the object detection unit 120 calculates the area reliability for the current defective area. The region confidence is a value that indicates the accuracy of the inference for the defect region.

[0030] The region reliability is calculated, for example, as follows. First, the object detection unit 120 acquires an inference score for the defect area from the object detection model. The inference score calculated by the object detection model for the object area (bounding box) detected by the object detection model is used as the inference score for the defect area. When multiple object areas are detected and multiple inference scores for the multiple object areas are obtained, for example, a statistical value of the multiple inference scores is used as the inference score for the defect area. Examples of the statistical value are the average value, maximum value, or minimum value. The object detection unit 120 also determines a visibility score for the target image. The visibility score is a value that represents the visibility of the image. The more visible the edges are, the higher the visibility score. For example, the higher the average brightness and contrast, the higher the visibility score. Then, the object detection unit 120 calculates the region reliability using the inference score for the defect region and the visibility score for the target image. The region reliability is calculated using an arbitrary method. The region confidence may be calculated using other scores, such as an inference score for the defect region, a visual score for the target image, or other scores.

[0031] The area reliability for the current defective area is referred to as the current area reliability. The current region reliability is saved and used as the previous region reliability when the next inspection is performed on the same inspection object as the current inspection object. For example, the current area reliability is stored as follows: The target image is accompanied by information indicating the image capture date and time. The object detection unit 120 selects the date and time closest to the image capture date and time of the current target image from the dates and times indicated in the driving history data in the storage device 220, and registers the current area reliability in the driving history data in association with the selected date and time.

[0032] In step S130, the coefficient determination unit 130 determines the check coefficients. The check coefficient is a coefficient used to calculate the check value. The inspection value is a value that is referenced to determine whether an abnormality has occurred in the inspection target.

[0033] Specifically, the coefficient determination unit 130 determines the check coefficient based on the length of the unchecked period. The check coefficient is determined by an arbitrary method based on the length of the unchecked period. The check coefficient is proportional to the length of the unchecked period, i.e., the longer the unchecked period, the larger the check coefficient. The unexamined period corresponds to the period from the previous imaging of the same examination subject to the current imaging.

[0034] The non-inspection period is determined, for example, as follows. The target image is provided with information indicating the date and time of capture. First, the coefficient determination unit 130 refers to the driving history data in the storage device 220 to determine the current date and time corresponding to the current image capture. The current date and time is the date and time indicated in the driving history data that is closest to the date and time when the current target image was captured. The driving position corresponding to the current date and time is referred to as the current position. Next, the coefficient determination unit 130 finds a traveling position (corresponding position) that corresponds to the current position by referring to the traveling history data in the storage device 220. The corresponding position is a traveling position whose distance from the current position is within an allowable range. Then, the coefficient determination unit 130 determines the previous date and time corresponding to the previous image capture time by referring to the travel history data in the storage device 220. The previous date and time is the date and time closest to the current date and time among the dates and times associated with each corresponding position. The period from the previous date and time to the current date and time is the uninspected period.

[0035] In step S140, the inspection value calculation unit 140 calculates the current inspection value using the inspection coefficient and the current region reliability.

[0036] The current test value is calculated, for example, as follows. First, the inspection value calculation section 140 calculates the region reliability variation dS using the previous region reliability and the current region reliability. The region reliability change amount dS is the difference between the previous region reliability and the current region reliability. Specifically, the test value calculation unit 140 calculates the region reliability change amount dS by subtracting the previous region reliability from the current region reliability.

[0037] The previous region reliability is the region reliability calculated as the current region reliability during the previous inspection of the same inspection object as the current inspection object. For example, the previous area reliability is obtained as follows: The driving history data in the storage device 220 indicates the area reliability together with the driving location in association with each date and time. The previous date and time is as described in step S130. The inspection value calculation unit 140 obtains the area reliability associated with the previous date and time from the driving history data. The obtained area reliability is the previous area reliability.

[0038] Then, the inspection value calculation unit 140 multiplies the region reliability variation dS by the inspection coefficient K to calculate a value (inspection score) and calculates the value (inspection score) by the previous inspection value P N-1 This adds up to the current test value P N is calculated. This test value P N corresponds to the value (integrated value) obtained by integrating the test scores.

[0039] This test value P N is expressed by the following formula: P N = P N-1 +K·dS

[0040] This test value P N is saved, and the previous test value P N-1 Used as. For example, the current test value P N is stored as follows: Information indicating the date and time of image capture is added to the target image. The inspection value calculation unit 140 selects the date and time closest to the date and time of image capture of the current target image from the dates and times indicated in the driving history data in the storage device 220, and associates the selected date and time with the current inspection value P N is registered in the driving history data.

[0041] Previous test value P N-1 is the current test value P N The test value calculated as follows: For example, the previous test value P N-1 is acquired as follows. The driving history data in the storage device 220 indicates the test value together with the driving location in association with each date and time. The previous date and time is as explained in step S130. The test value calculation unit 140 acquires the test value associated with the previous date and time from the driving history data. The acquired test value is the previous test value P N-1 is.

[0042] In step S150, the abnormality detection unit 150 determines whether an abnormality has occurred in the test object based on the current test value.

[0043] The occurrence of an abnormality is determined as follows: The abnormality detection unit 150 compares the current test value with a threshold value, which is determined in advance. If the current test value is greater than the threshold value, the abnormality detection unit 150 determines that an abnormality has occurred in the test object.

[0044] Then, the abnormality detection unit 150 outputs the determination result. The determination result is information indicating whether or not an abnormality has occurred in the inspection object. For example, the abnormality detection unit 150 registers the image capture date and time of the current target image, the current date and time, the current position, the determination result, etc. in the determination history data. The determination history data is stored in the storage unit 190, for example.

[0045] ***Features of the first embodiment*** The devised features of the first embodiment will be described with reference to FIG. The anomaly detection device 100 accumulates scores (inference scores, visual recognition scores, etc.) during inspection, and considers that an abnormality has been detected when the accumulated value P (inspection value) exceeds a threshold value α. If the uninspected period is long, it is possible that a defect has formed on the surface of the object to be inspected (for example, the road surface) during that period. However, it is strange to accumulate a high score when the uninspected period is short, causing the accumulated value P to rise sharply, and to conclude that a defect has formed on the surface of the object to be inspected. Therefore, the anomaly detection device 100 sets a coefficient K proportional to the time elapsed since the previous inspection. Then, the anomaly detection device 100 adds the product of the coefficient K and the amount of change dS in the inference reliability (area reliability) to the integrated value P. (1) If the period without inspection is long, the coefficient K is large and the increase in the integrated value P is large. (2) When the period without inspection is short, the coefficient K is small and the increase in the integrated value P is small.

[0046] ***Effects of the First Embodiment*** With conventional technology, the score rises sharply when a previously non-existent object (such as a dead leaf) is detected, making it prone to false positives due to discontinuous conditions such as dead leaves. In the first embodiment, a coefficient is determined according to the length of the untested period. The longer the untested period, the larger the coefficient, and the shorter the untested period, the smaller the coefficient. Then, a judgment is made using a score (test value) based on this coefficient. As a result, the first embodiment can detect abnormalities in the inspection object while suppressing false detection of defects.

[0047] Embodiment 2 The embodiment in which the check coefficients are changed in response to changes in the region reliability will be described below with reference to FIGS. 5 and 6, focusing on the differences from the first embodiment.

[0048] ***Configuration Description*** The configuration of the abnormality detection device 100 is the same as that in the first embodiment.

[0049] ***Explanation of Operation*** The abnormality detection method will be described with reference to FIG. Steps S210 to S250 correspond to steps S110 to S150 in the first embodiment.

[0050] In step S210, the image acquisition unit 110 acquires the current target image. Step S210 is the same as step S110 in the first embodiment.

[0051] In step S220, the object detection unit 120 detects the current defective area by executing object detection on the current target image. Furthermore, the object detection unit 120 calculates the area reliability for the current defective area.

[0052] Specifically, the object detection unit 120 detects one or more current defect areas and obtains one or more current area reliabilities corresponding to the one or more current defect areas.

[0053] The object detection method is the same as that in the first embodiment. The method for calculating the region reliability is the same as the method in the first embodiment.

[0054] In step S230, the coefficient determination unit 130 determines the current check coefficient.

[0055] The current inspection coefficients are stored in the same manner as the current region reliability, and are used as the previous inspection coefficients in the next inspection of the same inspection object as the current inspection object.

[0056] The check coefficient is determined as follows: First, the coefficient determination unit 130 calculates the average of one or more current region reliabilities. The calculated average is referred to as the current average reliability. The current average reliability is stored in the same manner as the current region reliability, and is used as the previous average reliability at the time of the next inspection of the same inspection object as the current inspection object.

[0057] Next, the coefficient determination unit 130 calculates the amount of change in the average reliability using the previous average reliability and the current average reliability. The previous average reliability is the average reliability calculated as the current average reliability during the previous inspection of the same inspection object as the current inspection object. For example, the previous average reliability is obtained as follows: The driving history data in the storage device 220 indicates the average reliability together with the driving location in association with each date and time. The previous date and time is as described in step S130 of the first embodiment. The coefficient determination unit 130 obtains the average reliability associated with the previous date and time from the driving history data. The obtained average reliability is the previous average reliability. The average reliability change amount is the difference between the previous average reliability and the current average reliability. Specifically, the coefficient determination unit 130 calculates the average reliability change amount by subtracting the previous average reliability from the current average reliability.

[0058] Then, the coefficient determination unit 130 determines the current check coefficient based on the calculated average reliability variation and the previous check coefficient. For example, the previous check coefficient is obtained as follows: The driving history data in the storage device 220 indicates the check coefficient together with the driving location in association with each date and time. The previous date and time is as explained in step S130 of the first embodiment. The coefficient determination unit 130 obtains the check coefficient associated with the previous date and time from the driving history data. The obtained check coefficient is the previous check coefficient.

[0059] The current inspection coefficient is determined by an arbitrary method based on the average reliability change amount and the previous inspection coefficient. If the current average reliability is greater than the previous average reliability and the average reliability change amount is a positive value, the current inspection coefficient is greater than the previous inspection coefficient. If the current average reliability is smaller than the previous average reliability and the average reliability change amount is a negative value, the current inspection coefficient is smaller than the previous inspection coefficient. The difference between the previous test coefficient and the current test coefficient (the magnitude of change in the test coefficient) is larger the greater the average reliability change.

[0060] In step S240, the test value calculation section 140 calculates the current test value using the current test coefficient and the current region reliability. Step S240 is the same as step S140 in the first embodiment.

[0061] In step S250, abnormality detection section 150 determines whether an abnormality has occurred in the test object based on the current test value. Step S250 is the same as step S150 in the first embodiment.

[0062] ***Features of the second embodiment*** The improvements of the second embodiment will be described with reference to FIG. If the reliability of objects in a time series of images of the same location improves overall, the latter image can be considered to be more reliable (e.g., more visible) than the former image. Therefore, the abnormality detection device 100 changes the coefficient K according to the amount of change in the average value of the inference reliability (area reliability) of the surrounding objects before and after in a time series. (1) If the average confidence in the inference of the detection results of the object detected in the image is relatively low, the coefficient K will be small. (2) The average confidence of the inferences about the detected objects in the image is relatively high. In this case, the coefficient K is large.

[0063] ***Effects of the Second Embodiment*** In the second embodiment, the coefficient K is changed according to the amount of change in the average value of the inference reliability (area reliability) of all objects (peripheral objects) detected from the image over time. Then, in the second embodiment, a score (inspection value) based on this coefficient is used to make a judgment. As a result, the second embodiment can detect abnormalities in the inspection object while suppressing false detection of defects.

[0064] Embodiment 3 The embodiment in which an inspection coefficient is determined based on the aspect ratio of a defect area will be described below, mainly with reference to FIGS. 7 and 8, with respect to differences from the first embodiment.

[0065] ***Configuration Description*** The configuration of the abnormality detection device 100 is the same as that in the first embodiment.

[0066] ***Explanation of Operation*** The abnormality detection method will be described with reference to FIG. Steps S310 to S350 correspond to steps S110 to S150 in the first embodiment.

[0067] In step S310, the image acquisition unit 110 acquires the current target image. Step S310 is the same as step S110 in the first embodiment.

[0068] In step S320, the object detection unit 120 detects the current defective area by executing object detection on the current target image. Furthermore, the object detection unit 120 calculates the area reliability for the current defective area.

[0069] Specifically, the object detection unit 120 detects one or more current defect areas and obtains one or more current area reliabilities corresponding to the one or more current defect areas.

[0070] The object detection method is the same as that in the first embodiment. The method for calculating the region reliability is the same as the method in the first embodiment.

[0071] In step S330, the coefficient determination unit 130 determines the check coefficients.

[0072] The check coefficient is determined as follows: First, the coefficient determining unit 130 calculates the aspect ratio of each of the one or more current defect areas to obtain one or more aspect ratios corresponding to the one or more current defect areas. Next, the coefficient determination unit 130 calculates the similarity between one or more of the obtained aspect ratios. The calculated similarity is referred to as aspect ratio similarity. Then, the coefficient determination unit 130 determines a check coefficient based on the calculated aspect ratio similarity. The check coefficient is determined by an arbitrary method based on the aspect ratio similarity. The check coefficient is inversely proportional to the aspect ratio similarity, i.e., the higher the aspect ratio similarity, the smaller the check coefficient.

[0073] In step S340, the inspection value calculation section 140 calculates the current inspection value using the inspection coefficient and the current region reliability. Step S340 is the same as step S140 in the first embodiment.

[0074] In step S350, abnormality detection section 150 determines whether an abnormality has occurred in the test object based on the current test value. Step S350 is the same as step S150 in the first embodiment.

[0075] ***Features of the third embodiment*** The improvements of the third embodiment will be described with reference to FIG. If the aspect ratios are similar, there is a high probability that there are many objects of the same type (e.g., fallen leaves). Therefore, the coefficient K needs to be small. Therefore, when the similarity of the aspect ratios of the surrounding objects is high (for example, when the variance of the aspect ratios is small), the abnormality detection device 100 reduces the coefficient K. (1) When the aspect ratios of the bounding boxes of the objects detected in the image are not similar to each other (when the similarity is low), the coefficient K becomes large. (2) When the aspect ratios of the bounding boxes of the objects detected in the image are similar to each other (when the similarity is high), the coefficient K becomes small.

[0076] ***Effects of the Third Embodiment*** When the similarity of the aspect ratios of the peripheral objects is high (for example, when the number of peripheral objects with high similarity of aspect ratios is equal to or greater than a threshold), the coefficient K is reduced in the third embodiment. Then, the third embodiment makes a judgment using a score (inspection value) based on this coefficient. As a result, the third embodiment can detect abnormalities in the inspection object while suppressing false detection of defects.

[0077] ***Supplementary explanation of implementation form*** The hardware configuration of the abnormality detection device 100 will be described with reference to FIG. The abnormality detection device 100 includes a processing circuit 109 . The processing circuit 109 is hardware that realizes the image acquisition unit 110 , the object detection unit 120 , the coefficient determination unit 130 , the test value calculation unit 140 , and the abnormality detection unit 150 . The processing circuitry 109 may be dedicated hardware, or may be a processor 101 that executes a program stored in memory 102 .

[0078] When processing circuitry 109 is dedicated hardware, processing circuitry 109 may be, for example, a single circuit, a multiple circuit, a programmed processor, parallel programmed processors, an ASIC, an FPGA, or a combination thereof. ASIC is an abbreviation for Application Specific Integrated Circuit. FPGA is an abbreviation for Field Programmable Gate Array.

[0079] In the processing circuit 109, some functions may be realized by dedicated hardware, and the remaining functions may be realized by software or firmware.

[0080] In this way, the functions of the abnormality detection device 100 can be realized by hardware, software, firmware, or a combination of these.

[0081] Each embodiment is an example of a preferred embodiment and is not intended to limit the technical scope of the present disclosure. Each embodiment may be implemented in part or in combination with other embodiments. Procedures described using flowcharts, etc. may be modified as appropriate. The "part" of each element of the abnormality detection device 100 may be read as a "process," a "step," a "circuit," or a "circuitry." [Explanation of symbols]

[0082] 100 Anomaly detection device, 101 Processor, 102 Memory, 103 Auxiliary storage device, 104 Communication device, 105 Input / output interface, 109 Processing circuit, 110 Image acquisition unit, 120 Object detection unit, 130 Coefficient determination unit, 140 Inspection value calculation unit, 150 Anomaly detection unit, 190 Storage unit, 200 Anomaly detection system, 201 Mobile object, 210 Imaging device, 220 Storage device, 230 Positioning device.

Claims

1. an object detection unit that performs object detection on a current target image obtained by capturing an image of the inspection target, detects an area inferred to contain a defect in the inspection target as a current defect area, and calculates the accuracy of the inference for the current defect area as a current area reliability; an inspection value calculation unit that calculates an area reliability change amount using the previous area reliability and the current area reliability, multiplies the calculated area reliability change amount by an inspection coefficient used to calculate the inspection value, and adds the calculated value to the previous inspection value to calculate the current inspection value; an abnormality detection unit that determines whether an abnormality has occurred in the test object based on the current test value; An abnormality detection device comprising:

2. an object detection unit that performs object detection on a current target image obtained by capturing an image of the inspection target, detects an area inferred to contain a defect in the inspection target as a current defect area, and calculates the accuracy of the inference for the current defect area as a current area reliability; a coefficient determination unit that determines an inspection coefficient to be used in calculating an inspection value based on the length of an uninspected period corresponding to the period from the previous imaging to the current imaging; an inspection value calculation unit that calculates a current inspection value using the inspection coefficient and the current region reliability; an abnormality detection unit that determines whether an abnormality has occurred in the test object based on the current test value; An abnormality detection device comprising:

3. an object detection unit that performs object detection on a current target image obtained by capturing an image of the inspection target, thereby detecting one or more areas inferred to contain defects in the inspection target as one or more current defect areas, and calculates the accuracy of the inference for the one or more current defect areas as one or more current area reliabilities; a coefficient determination unit that calculates an average of the one or more current region reliabilities as a current average reliability, calculates an average reliability change amount using the previous average reliability and the current average reliability, and determines a current inspection coefficient based on the calculated average reliability change amount and the previous inspection coefficient; an inspection value calculation unit that calculates a current inspection value using the current inspection coefficient and the one or more current region reliabilities; an abnormality detection unit that determines whether an abnormality has occurred in the test object based on the current test value; An abnormality detection device comprising:

4. an object detection unit that performs object detection on a current target image obtained by capturing an image of the inspection target, thereby detecting one or more areas inferred to contain defects in the inspection target as one or more current defect areas, and calculates the accuracy of the inference for the one or more current defect areas as one or more current area reliabilities; a coefficient determination unit that calculates an aspect ratio of each of the one or more current defect areas to obtain one or more aspect ratios corresponding to the one or more current defect areas, calculates a similarity between the one or more obtained aspect ratios as an aspect ratio similarity, and determines an inspection coefficient to be used in calculating an inspection value based on the calculated aspect ratio similarity; an inspection value calculation unit that calculates a current inspection value using the inspection coefficient and the one or more current region reliabilities; an abnormality detection unit that determines whether an abnormality has occurred in the test object based on the current test value; An abnormality detection device comprising:

5. An abnormality detection device, performing object detection on a current target image obtained by capturing an image of the inspection target, and detecting an area in which a defect of the inspection target is inferred to be reflected as a current defect area; and calculating the accuracy of the inference for the current defect area as a current area reliability; calculating a region reliability change amount using the previous region reliability and the current region reliability, multiplying the calculated region reliability change amount by an inspection coefficient used to calculate the inspection value, and adding the calculated value to the previous inspection value to calculate the current inspection value; Based on the current test value, it is determined whether an abnormality has occurred in the test object. Anomaly detection methods.

6. An abnormality detection device, performing object detection on a current target image obtained by capturing an image of the inspection target, and detecting an area in which a defect of the inspection target is inferred to be reflected as a current defect area; and calculating the accuracy of the inference for the current defect area as a current area reliability; determining a test coefficient to be used in calculating the test value based on the length of the non-test period corresponding to the period from the previous imaging time to the current imaging time; Calculating a current inspection value using the inspection coefficient and the current region reliability; Based on the current test value, it is determined whether an abnormality has occurred in the test object. Anomaly detection methods.

7. An abnormality detection device, performing object detection on a current target image obtained by capturing an image of the inspection target, thereby detecting one or more areas inferred to contain defects in the inspection target as one or more current defect areas, and calculating the accuracy of the inference for the one or more current defect areas as one or more current area reliabilities; calculating an average of the one or more current region reliabilities as a current average reliability, calculating an average reliability change amount using the previous average reliability and the current average reliability, and determining a current inspection coefficient based on the calculated average reliability change amount and the previous inspection coefficient; Calculating a current inspection value using the current inspection coefficient and the one or more current area reliabilities; Based on the current test value, it is determined whether an abnormality has occurred in the test object. Anomaly detection methods.

8. An abnormality detection device, performing object detection on a current target image obtained by capturing an image of the inspection target, thereby detecting one or more areas inferred to contain defects in the inspection target as one or more current defect areas, and calculating the accuracy of the inference for the one or more current defect areas as one or more current area reliabilities; calculating an aspect ratio of each of the one or more current defect areas to obtain one or more aspect ratios corresponding to the one or more current defect areas; calculating a similarity between the obtained one or more aspect ratios as an aspect ratio similarity; and determining an inspection coefficient to be used in calculating an inspection value based on the calculated aspect ratio similarity; Calculating a current inspection value using the inspection coefficient and the one or more current region reliabilities; Based on the current test value, it is determined whether an abnormality has occurred in the test object. Anomaly detection methods.

9. An object detection process in which object detection is performed on a current target image obtained by capturing an image of the inspection target, thereby detecting an area inferred to contain a defect in the inspection target as the current defect area, and determining the accuracy of the inference for the current defect area as the current area reliability; an inspection value calculation process for calculating an area reliability change amount using the previous area reliability and the current area reliability, and multiplying the calculated area reliability change amount by an inspection coefficient used in calculating the inspection value, and adding the calculated value to the previous inspection value to calculate a current inspection value; an abnormality detection process for determining whether an abnormality has occurred in the test object based on the current test value; An anomaly detection program that causes a computer to run the following.

10. An object detection process in which object detection is performed on a current target image obtained by capturing an image of the inspection target, an area inferred to contain a defect in the inspection target is detected as the current defect area, and the accuracy of the inference for the current defect area is calculated as the current area reliability; a coefficient determination process for determining an inspection coefficient to be used in calculating the inspection value based on the length of an uninspected period corresponding to the period from the previous imaging to the current imaging; an inspection value calculation process for calculating a current inspection value using the inspection coefficient and the current region reliability; an abnormality detection process for determining whether an abnormality has occurred in the test object based on the current test value; An anomaly detection program that causes a computer to run the following.

11. An object detection process in which object detection is performed on a current target image obtained by imaging the object to be inspected, thereby detecting one or more areas inferred to contain defects in the object to be inspected as one or more current defect areas, and calculating the accuracy of the inference for the one or more current defect areas as one or more current area reliabilities; a coefficient determination process of calculating an average of the one or more current region reliabilities as a current average reliability, calculating an average reliability change amount using the previous average reliability and the current average reliability, and determining a current inspection coefficient based on the calculated average reliability change amount and the previous inspection coefficient; an inspection value calculation process for calculating a current inspection value using the current inspection coefficient and the one or more current region reliabilities; an abnormality detection process for determining whether an abnormality has occurred in the test object based on the current test value; An anomaly detection program that causes a computer to run the following.

12. An object detection process in which object detection is performed on a current target image obtained by imaging the object to be inspected, thereby detecting one or more areas inferred to contain defects in the object to be inspected as one or more current defect areas, and calculating the accuracy of the inference for the one or more current defect areas as one or more current area reliabilities; a coefficient determination process for calculating an aspect ratio of each of the one or more current defect areas to obtain one or more aspect ratios corresponding to the one or more current defect areas, calculating a similarity between the obtained one or more aspect ratios as an aspect ratio similarity, and determining an inspection coefficient to be used in calculating an inspection value based on the calculated aspect ratio similarity; an inspection value calculation process for calculating a current inspection value using the inspection coefficient and the one or more current region reliabilities; an abnormality detection process for determining whether an abnormality has occurred in the test object based on the current test value; An anomaly detection program that causes a computer to run the following.

Images