JP6484854B2 - Solar panel appearance monitoring device - Google Patents

Description

  The present invention relates to a solar panel appearance monitoring device for monitoring the appearance of a solar panel facility, and in particular, the appearance of a solar panel is analyzed by analyzing an image taken by a monitoring camera installed in the solar panel facility. The present invention relates to a solar panel appearance monitoring device to be monitored.

  In general, large-scale solar panel equipment (solar power generation equipment) is often constructed unattended in a remote place. For this reason, there are cases in which a flying object such as an agricultural bag or sheet, or a plant such as vine grass, covers the solar panel and is not noticed for a while.

  Therefore, for example, in Patent Documents 1 to 4 below, a method for estimating the state of the solar panel based on the power generation amount of the solar panel, that is, a method for detecting an abnormality of the solar panel by detecting a decrease in the power generation amount. Has proposed.

  Further, for example, the following Non-Patent Document 1 has a method of inspecting an abnormality of a solar panel using an infrared thermography. In the following Non-Patent Document 1, paying attention to heat generation when a failure occurs in the solar panel, the operator visually observes the result of photographing the solar panel with infrared thermography, and confirms the presence or absence of a part having a high temperature. So, we have proposed a method for judging solar panel abnormalities.

JP 2014-60365 A Japanese Patent No. 5,607,772 Japanese Patent No. 5413933 JP2013-93430A

Nishimura, Shimizu, Fujiyoshi, “Classification of key points using two-stage Randomized Trees” [online], July 2010, Image Recognition and Understanding Symposium MIRU2010, pp.1412-1419, [20154 Search 20th of May], Internet ≪URL: http: //at.azbil.com/mt/product/pdf/Mobile%20Solar%20pole%20APP_STORY_JP_ys.pdf≫

  In the methods of Patent Documents 1 to 4, the presence or absence of abnormality can be estimated from the power generation amount of the solar panel, but it is not known whether or not an abnormality has actually occurred. In addition, it is not known whether the cause of the abnormality is a mechanical failure of the solar panel itself or other factors such as covering with foreign matter.

  In the method of Non-Patent Document 1, it is difficult to apply to a large-scale photovoltaic power generation facility because confirmation work by an operator is necessary. Further, infrared thermography is expensive, and it is difficult to install a large amount in a large photovoltaic power generation facility in this respect.

  The present invention has been made in view of the technical situation as described above, and can be applied to a large-scale solar panel facility by automatically monitoring the appearance of the solar panel with an inexpensive configuration. An object of the present invention is to provide a solar panel appearance monitoring device.

The solar panel appearance monitoring device according to the first invention for solving the above-mentioned problems is
Identification for solar panel inspection based on a solar panel image group in which a solar panel is photographed for a certain period of time with a surveillance camera and a foreign object image group in which the surveillance camera is photographed for a certain period of time. A solar panel inspection parameter adjustment device,
At the time of monitoring, an image obtained by photographing the solar panel with the monitoring camera is input to the discriminator, and it is determined whether or not the foreign object exists on the solar panel based on the inspection result of the discriminator. And a solar panel inspection device.

A solar panel appearance monitoring device according to the second invention for solving the above-mentioned problems is as follows.
In the solar panel appearance monitoring device according to the first invention,
The solar panel inspection parameter adjustment device is:
Each of the solar panel image groups is divided into small areas, the small areas in which the solar panels are reflected are extracted as partial images, and a set of the partial images is a solar panel partial image collection,
In the image group of the foreign matter, set a small region where the portion where the foreign matter is reflected enters, take out this small region as a partial image, a set of the partial images as a foreign matter partial image collection,
The discriminator is constructed based on the solar panel partial image collection and the foreign object partial image collection.

A solar panel appearance monitoring device according to a third invention for solving the above-mentioned problems is
In the solar panel appearance monitoring device according to the first or second invention,
The solar panel inspection parameter adjustment device is:
In addition to the small area in which the solar panel is reflected, an area shifted from the small area by a predetermined range is taken out as a partial image to form the solar panel partial image collection.

A solar panel appearance monitoring device according to a fourth invention for solving the above-mentioned problems is as follows.
In the solar panel appearance monitoring device according to any one of the first to third inventions,
The classifier is a multi-class classifier;
The solar panel inspection parameter adjustment device constructs the discriminator based on the solar panel image group and a plurality of types of the foreign object image groups obtained by photographing a plurality of types of the foreign objects for a certain period with the monitoring camera. It is characterized by that.

A solar panel appearance monitoring device according to a fifth invention for solving the above-mentioned problems is
In the solar panel appearance monitoring device according to any one of the first to fourth inventions,
The solar panel inspection apparatus is
A rectangular portion in which the solar panel is reflected is sequentially extracted as a partial image from the image obtained by photographing the solar panel with the monitoring camera, and the partial image is input to the discriminator. Based on the inspection result of the discriminator. And determining whether or not the foreign matter exists on the solar panel.

  According to the solar panel appearance monitoring apparatus according to the present invention, the appearance of the solar panel can be automatically monitored with an inexpensive configuration by analyzing the image taken by the monitoring camera.

  Moreover, according to the solar panel external appearance monitoring device according to the present invention, since the monitoring camera image at the time of occurrence of abnormality is stored, the cause at the time of occurrence of abnormality can be confirmed by an image.

  Furthermore, since the monitoring apparatus can be configured by installing a large number of inexpensive small cameras or surveillance cameras with pan / tilt mechanisms, it can be applied to large-scale photovoltaic power generation facilities.

It is the schematic explaining the method to install the small camera which image | photographs each sunlight panel in Example 1 of this invention. It is the schematic explaining the method to install the camera with a pan-tilt mechanism for every division of photovoltaic power generation equipment in Example 1 of this invention. It is a schematic diagram explaining the outline | summary of operation | movement of the solar panel external appearance monitoring apparatus which concerns on Example 1 of this invention. It is a schematic diagram explaining the construction | assembly outline | summary of the discriminator for a solar panel test | inspection in Example 1 of this invention. It is a schematic diagram which shows the example which takes out a partial image from the image of a solar panel in Example 1 of this invention. It is a schematic diagram which shows the example which takes out the partial image from the image of a foreign material in Example 1 of this invention. It is a schematic diagram explaining the outline | summary of the test | inspection process with respect to the solar panel in Example 1 of this invention. It is a flowchart of the test | inspection process with respect to the solar panel in Example 1 of this invention. It is a block diagram which shows the structure of the solar panel external appearance monitoring apparatus which concerns on Example 1 of this invention. It is a block diagram which shows the structure of the parameter adjustment apparatus for solar panel inspection in Example 1 of this invention. It is a block diagram which shows the structure of the solar panel inspection apparatus in Example 1 of this invention. It is an example of the output in the integrated interface of the test result of the solar panel inspection apparatus in Example 1 of this invention. It is a schematic diagram which shows the example which takes out a partial image from the image of a sunlight panel in Example 2 of this invention.

  The solar panel appearance monitoring device according to the present invention automatically monitors the presence or absence of a foreign object covering a solar panel by analyzing a picture taken by a monitoring camera for a large-scale photovoltaic power generation facility installed on the ground. It detects automatically and monitors the appearance of the solar panel.

  Hereinafter, a solar panel appearance monitoring apparatus according to the present invention will be described with reference to the drawings in the embodiments.

[Example 1]
Large-scale solar power generation facilities have a large number of solar panels. There are two types of surveillance camera installation methods for reducing equipment costs. One is a method of installing a plurality of inexpensive small cameras 12 as monitoring cameras on each solar panel 11 as shown in FIG.

  Secondly, as shown in FIG. 2, a plurality of solar panels 11 are sequentially photographed by a monitoring camera 13 with a pan / tilt mechanism installed in each section of the photovoltaic power generation facility. The pan / tilt mechanism 13a is a mechanism that can rotate the surveillance camera in the vertical direction and the horizontal direction.

  As a feature of large-scale solar power generation facilities, the number of solar panels is large. Therefore, regardless of which of the two types of installation methods shown in FIGS. It is difficult to transmit frequently. In view of this feature, in this embodiment, an updated image is obtained from the surveillance camera at a frequency of a long update cycle such as about one per minute for one solar panel.

  In this embodiment, a Random Forests classifier is constructed using an image obtained by photographing a solar panel, and the presence or absence of foreign matter on the solar panel is inspected by the judgment of the Random Forests classifier. The Random Forests classifier is one of multiclass classifiers that can classify a plurality of objects at once (multiple types of classification in one process).

  If typical foreign objects covering solar panels, such as “agricultural bags”, “sheets”, and “plants (such as vines)” are known, “solar panels” and “agricultural bags” should be classified. ”,“ Sheet ”,“ Plant (such as vine) ”and“ Others ”are prepared to construct a classifier. In this way, typical foreign objects such as “agricultural bags”, “sheets”, and “plants (such as vines)” can be classified with high detection accuracy. Therefore, the detection accuracy of the presence or absence of foreign matter is improved.

  In addition to Random Forests classifiers, there are generally classifiers such as Adaboost and Support Hector Machine. These are two-class classifiers that perform two types of classification, “solar panel” and “others”. Even such a two-class classifier can detect the presence or absence of foreign matter. However, for example, when the surface of the “sheet” is relatively similar to the solar panel, the two-class classifier may erroneously detect. In contrast, multi-class classifiers such as Random Forests classifiers can detect with high accuracy even in this case.

  Although a multi-class classifier can be constructed by configuring the 2-class classifier in multiple stages, it is necessary to construct a plurality of classifiers individually for each monitoring target. On the other hand, a multi-class classifier such as a Random Forests classifier has only to construct one classifier and can be a simple procedure.

  FIG. 3 is a schematic diagram for explaining the outline of the operation of the solar panel appearance monitoring apparatus according to the present embodiment. In the following, for simplification, only the first to third solar panels (not shown) of the plurality of solar panels 11 will be described, but actually, a larger number of solar panels are targeted for monitoring.

  As shown in FIG. 3, in the solar panel appearance monitoring device according to the present embodiment, the first to third sunlights constructed in advance as step S <b> 1 for the monitoring camera images 11 a to 11 c of the first to third solar panels. By using the discriminators 14a to 14c for panel inspection, the inspection processes of the first to third solar panels are individually performed, and the inspection result output is performed as step S2. Then, the presence or absence of foreign matter is detected (determined) based on the inspection result.

  In addition, the apparatus which performs each test | inspection process of step S1 may prepare separately for every monitoring camera image 11a-11c of the 1st-3rd solar panel, or performs many test | inspection processes within one apparatus. You may do it.

  Hereinafter, an outline of the construction of the classifiers (classifiers 14a to 14c) for solar panel inspection will be described with reference to FIG.

The procedure for constructing a discriminator for solar panel inspection is as follows.
Step S11: An image group of each solar panel is prepared.
Step S12: A plurality of types of foreign matter image groups are prepared.
Step S13: A partial image collection is created from the image group of the solar panel.
Step S14: A partial image collection is created from the foreign matter image group.
Step S15: A corresponding solar panel inspection classifier is constructed.
Hereinafter, steps S11 to S15 will be described in detail.

  In step S11, the first to third solar panels are previously photographed by the monitoring camera 12 (13) for a certain period of time without any foreign matter, and the monitoring camera images 21a-1, 21a-2,. The second solar panel monitoring camera images 21b-1, 21b-2,... As the first solar panel image group 21A, and the third solar panel monitoring as the second solar panel image group 21B. The camera images 21c-1, 21c-2, ... are stored as an image group 21C of the third solar panel, respectively.

  Note that the “monitoring camera image of the solar panel” here is different from the monitoring camera images 11a to 11c of the first to third solar panels in FIG. 3 described above, for the construction of the discriminators 14a to 14c. Previously taken.

  In step S12, a plurality of images obtained by separately capturing foreign objects such as agricultural bags, sheets, plants (such as vine grass), etc. (images captured by a camera different from the surveillance camera 12 (13) may be used) The first foreign object images 22a-1, 22a-2,... As the first foreign object image group 22A and the second foreign object images 22b-1, 22b. -2,... Are stored as the second foreign object image group 22B.

  Here, for simplification, only two foreign matter images are described, but in practice, images of various types of foreign matter may be stored as an image group as described above.

  In step S13, monitoring camera images 21a-1, 21a-2,... In the first solar panel image group 21A are displayed as shown in FIG. 5 (in FIG. 5, the first solar panel is monitored as an example. (Only the camera image 21a-1 is represented), each of which is divided into a grid pattern, and the rectangular portions in which the solar panel 11 is reflected are represented by partial images 23a-1, 23a-2 (partial images 23a-1 in FIG. 5). Are taken out and stored, and the partial images 23A of the first solar panel are made by collecting these partial images.

  Further, the same operation is performed for the image groups 21B and 21C of the second and third solar panels to create partial image collections 23B and 23C of the second and third solar panels.

  In step S14, as shown in FIG. 6 on the foreign matter images 22a-1, 22a-2,... In the first foreign matter image group 22A (in FIG. 6, as an example, the monitoring camera image 22a of the first foreign matter). -1 only), a rectangular range in which a portion where the foreign object 10 is reflected is set, and the rectangular ranges are set as partial images 24a-1, 24a-2 (in FIG. 6, the partial image 24a- 1 is shown) and is stored and a partial image collection 24A of the first foreign matter is created.

  Further, the same operation is performed on the second foreign matter image group 22B to create a partial image collection 24B of the second foreign matter.

In step S15, the first solar panel inspection discriminator 14a is learned and constructed using the first solar panel partial image collection 23A and the first and second foreign substance partial image collections 24A and 24B. Moreover, the same operation | work is performed using the partial image collections 23B and 23C of the 2nd and 3rd solar panel, and the discriminators 14b and 14c for the 2nd and 3rd solar panel inspection are constructed.
This completes the description of the construction of the discriminator for solar panel inspection.

  Next, the outline | summary of the test | inspection process (step S1) with respect to the 1st-3rd solar panel in a present Example is demonstrated using FIG.

  A monitoring camera image 11a obtained by photographing the first solar panel with the monitoring camera 12 (13) (the monitoring camera image 21a-1 and the like are used for constructing the discriminator 14a and the like, but the monitoring camera image 11a is A partial image 15a obtained by dividing the actual monitoring image (identifiers 14b and 14c as well) into a grid and taking out a rectangular portion in which the solar panel is reflected was previously constructed in steps S11 to S15. An inspection process is performed by inputting to the discriminator 14a for solar panel inspection and judging the inspection result.

  The same processing is performed on the monitoring camera images 11b and 11c of the second and third solar panels using the discriminators 14b and 14c, respectively. In addition, after judging an inspection result, an inspection result is output (step S2).

  As a procedure for dividing the monitoring camera images 11a to 11c into a grid pattern, the solar camera monitoring camera images 21a-1 and the like are partitioned into a grid pattern when constructing the above-described identifiers 14a to 14c for solar panel inspection. The same method as in the procedure (Step S13) is used.

  Moreover, about the rectangular part in which the solar panel is reflected, it shall set beforehand with respect to the monitoring camera image of the monitoring camera 12 (13) which image | photographs each solar panel before implementing an inspection process.

FIG. 8 is a flowchart for explaining the above-described step S1, that is, the inspection process for the solar panel. As shown in FIG. 8, the above step S1 specifically includes the following steps S21 to S27.
Step S21: Input a monitoring camera image of the solar panel.
Step S22: Extract a partial image to be inspected from the monitoring camera image.
Step S23: The partial image is input to the discriminator for solar panel inspection.
Step S24: An inspection by a discriminator for solar panel inspection is performed.
Step S25: The partial image inspection result is temporarily stored.
Step S26: It is determined whether or not all partial images have been inspected.
Step S27: Determine the inspection result.

  Hereinafter, steps S21 to S27 will be described in detail. In the following, for simplification, only the case where the first solar panel inspection camera image 11a is processed using the first solar panel inspection classifier 14a will be described, but another classifier was used. The same processing is also performed for the monitoring camera image processing of other solar panels.

  In step S21, the monitoring camera image 11a of the first solar panel is input.

  In step S22, the monitoring camera image 11a of the first solar panel input in step S21 is divided into a grid pattern, and rectangular portions in which the solar panel is reflected are sequentially extracted as partial images (for example, partial image 15a in FIG. 7). .

  In step S23, the partial image 15a taken out in step S22 and the like are input to the first solar panel inspection classifier 14a already constructed in steps S11 to S15.

  In step S24, the partial image 15a input in step S23 is inspected by the first solar panel inspection discriminator 14a, and the input partial image 15a is the first solar panel or foreign matter. Is output as the inspection result.

  In step S25, the inspection result of the partial image output from the discriminator 14a for first solar panel inspection is temporarily stored.

  In step S26, it is determined whether or not all partial images have been inspected by the first solar panel inspection discriminator 14a. If it is determined that all the partial images have been inspected, the process proceeds to step S27, and if not, the process proceeds to step S22.

  In step S27, the output of the discriminator 14a for the first solar panel inspection is integrated to determine the inspection result. That is, a determination ratio of probability values is provided, and the probability value that is a foreign object is higher than the probability value that is the first solar panel, and the probability value of a foreign object having the highest probability value is multiplied by the determination ratio. When the probability value of one solar panel is low, it is determined that a foreign object exists in the input partial image. This determination is performed for all partial images on the monitoring camera image 11a of the first solar panel, and the positions of the partial images determined to contain foreign matter are recorded.

  After step S27, step S2, that is, the output of the inspection result is performed. This inspection result indicates the presence / absence of a foreign substance and the corresponding position on the monitoring camera image 11a (the position of the partial image determined to have a foreign substance) when the foreign substance is present.

  As described above, the solar panel appearance monitoring device according to the present embodiment is roughly divided into two functions. One is a function for constructing a discriminator for solar panel inspection, and the other is an inspection function for solar panels.

  As shown in the schematic diagram of the solar panel appearance monitoring device (solar panel appearance monitoring device 1) according to the present embodiment in FIG. 9, the above two functions are separate devices, that is, parameter adjustment for solar panel inspection. Provided by the device 3 and the first to third solar panel inspection devices 4a to 4c (however, as already explained, a larger number of solar panels 11 are actually arranged, and each solar panel inspection device Are provided corresponding to a large number of solar panels 11. The same applies to discriminators 14a to 14c described later).

  Data exchange between the solar panel inspection parameter adjustment device 3 and the solar panel inspection devices 4 a to 4 c is performed via the storage 5. The storage 5 also has a function of storing image data input from the monitoring camera 12 (13) and outputting it to first and second storage units 34 and 44 described later. An integrated interface 2 for operating these devices is also provided. The solar panel appearance monitoring device 1 has a configuration including these.

  In the solar panel inspection parameter adjusting device 3, as in steps S11 to S15 described above, the monitoring camera image 21a obtained by photographing the first to third solar panels for a certain period of time with the monitoring camera 12 (13) in the absence of foreign matter in advance. -21, 21a-2,..., And the like, and images 22a-1, 22a-2 of the foreign matter obtained by photographing the foreign matter for a certain period with the monitoring camera 12 (13). Based on the image groups 22A to 22B (foreign object image group) composed of etc., the classifiers 14a to 14c are constructed, and the parameters of the classifiers 14a to 14c are calculated.

  FIG. 10 is a block diagram illustrating a configuration example of the solar panel inspection parameter adjustment device 3. The solar panel inspection parameter adjustment device 3 includes a first processing setting unit 31, a first partial image creation unit 32, a first classifier construction unit 33, and a first storage unit 34 that stores various data.

  In the first process setting unit 31, partial images on the monitoring camera images 21a-1, 21a-2,..., Etc. (for example, partial images 23a-1 in FIGS. 4 and 5) to be used for constructing the classifiers 14a to 14c. ), The amount of types of objects (solar panels and a plurality of foreign objects) to be classified by the classifiers 14a to 14c, the scale of the classifiers 14a to 14c, the number of trials when the classifiers 14a to 14c are constructed, First processing parameters configured by the amount of partial images used when constructing the classifiers 14a to 14c are set. In other words, the first processing parameter is a parameter and image range for constructing the Random Forests classifier, and “number of branch trees” “depth of branch tree” “number of trials for selecting branch nodes”. It consists of “the position and size of the rectangular part where the solar panel is reflected”. The set first processing parameter is stored in the first storage unit 34.

  In the first partial image creation unit 32, the first processing parameter, the image data of the image groups 21A, 21B, 21C such as the monitoring camera images 21a-1, 21a-2,... Stored in the first storage unit 34, and Based on the image data of the image groups 22A, 22B such as the foreign object images 22a-1, 22a-2,..., The partial image of the solar panel (for example, the part in FIGS. 4 and 5). An image 23a-1) and a partial image of a foreign object (for example, the partial image 24a-1 in FIG. 6) are created. The created partial image data is stored in the first storage unit 34.

  In the first classifier construction unit 33, based on the first processing parameter and the partial image data of each of the partial image collections 23A to 23C, 24A, and 24B stored in the first storage unit 34, as in step S15 described above. , Construct classifiers 14a-14c and calculate classifier parameters. The discriminator parameter is a parameter of the Random Forests discriminator itself, and is composed of “the number of branch trees”, “the depth of branch trees”, “condition judgment value of branch nodes”, and “probability distribution of end nodes”.

  The storage 5 stores image data of the monitoring camera images 11a to 11c of the first to third solar panels taken by the monitoring camera 12 (13), and the solar panel inspection parameter adjusting device 3 (particularly, the first memory). And the discriminator parameters are input from the solar panel inspection parameter adjustment device 3.

  In the solar panel inspection apparatuses 4a to 4c, as in steps S21 to S27 described above, the monitoring camera images 11a to 11c are input to the discriminators 14a to 14c, and the first to first screens are based on the inspection results of the discriminators 14a to 14c. 3 The appearance of the solar panel is inspected by determining whether or not there is a foreign object on the solar panel.

  In addition, as already demonstrated, each solar panel inspection apparatus 4a-4c is responsible for the inspection of one solar panel. That is, solar panel inspection apparatuses 4a to 4c are prepared for the first to third solar panels, respectively, and the inspection process is performed individually.

  FIG. 11 is a block diagram illustrating a configuration example of the solar panel inspection apparatus 4a (4b, 4c). The solar panel inspection apparatus 4a (4b, 4c) includes an image input unit 40, a second process setting unit 41, a second partial image creation unit 42, a second discriminator inspection unit 43, an inspection result determination unit 45, and an inspection result output. Unit 46 and a second storage unit 44 for storing various data.

  The image input unit 40 inputs the image data of the monitoring camera images 11a to 11c from the monitoring camera 12 (13) as in step S21 described above. The input image data is then saved and stored in the storage 5 and the second storage unit 44.

  The second process setting unit 41 sets second process parameters including the positions of the partial images 15a and the like on the monitoring camera images 11a to 11c and the determination ratio of the probability value of being a foreign object. In other words, the second processing parameter is a threshold value and an image range for making a final determination, and “inspection result determination threshold value” “position of a rectangular portion in which a solar panel is reflected, and Consists of “size”. The set second processing parameter 41 is stored in the second storage unit 44.

  In the 2nd partial image creation part 42, as shown in step S22 mentioned above based on the 2nd process parameter and the image data of the monitoring camera images 11a-11c preserve | saved at the 2nd memory | storage part 44, the part of a solar panel An image 15a and the like are created. The created partial image data is stored in the second storage unit 44.

  In the second discriminator inspection unit 43, the discriminator parameters of the discriminators 14a to 14c stored in the second storage unit 44 (input from the storage 5) and the second part as in steps S23 and S24 described above. Based on the partial image data such as the partial image 15a created by the image creating unit 42 and stored in the second storage unit 44, a probability value is calculated as to whether the partial image is a solar panel or a foreign object. . This probability value is stored in the second storage unit 44 as step S25 as test result data.

  The inspection result determination unit 45 determines the presence of the foreign matter based on the second processing parameter and the inspection result data for the partial image for one monitoring camera image stored in the second storage unit 44. Foreign matter data including the presence / absence and the position of the foreign matter on the image is created. The inspection result output unit 46 outputs foreign matter data.

  The inspection result of the solar panel inspection apparatus can be viewed through the integrated interface 2. For example, as shown in FIG. 12, when the foreign matter data of the solar panel inspection apparatus corresponding to a certain monitoring camera 12 (13) has foreign matter data, it is determined that there is a foreign matter on the image of the monitoring camera 12 (13). The position of the partial image is indicated by a rectangle as a foreign object detection location. In FIG. 12, the flying object 10a has adhered to the upper part of the solar panel 11, and the state which the grass 10b which hangs the lower part has illustrated is illustrated.

  By using the solar panel appearance monitoring apparatus according to the present embodiment and analyzing the image taken by the monitoring camera, the appearance of the solar panel can be automatically monitored.

[Example 2]
The solar panel appearance monitoring apparatus according to the present embodiment will be described with a focus on differences from the first embodiment.

  When the surveillance camera 12 (13) is installed outdoors, the surveillance camera shakes due to the influence of wind or the like, and the position of the surveillance target on the surveillance camera image (for example, the surveillance camera image 11a in FIG. 7) is several pixels vertically and horizontally. May move. In this case, the position of the partial image (for example, the partial image 23a-1 in FIGS. 4 and 5) taken out when the classifier (for example, the classifier 14a in FIG. 7) is constructed, and the corresponding solar panel The position of the partial image (for example, the partial image 15a in FIG. 7) when the inspection process is performed is different, and erroneous detection may occur.

  FIG. 13 is a schematic diagram illustrating creation of a partial image of the monitoring camera image in the solar panel appearance monitoring apparatus according to the present embodiment. Although only the monitoring camera image 21a-1 is described in FIG. 13, partial images are created by the same method for other monitoring camera images.

  Considering the occurrence of shaking in the monitoring camera image 11a, etc., when constructing the discriminator 14a, etc., as shown in FIG. 13, when the monitoring camera image 21a-1 is divided into a grid, the solar panel 11 In addition to the rectangular portion in which is shown, a range in which the position is shifted up and down and left and right by a predetermined number of pixels from the rectangular portion is used as a partial image (for example, partial image 33a-1) when constructing the classifier 14a and the like. A partial image collection is created based on the partial image 33a-1 and the like.

  That is, the solar panel inspection parameter adjustment device according to the present embodiment adds the shift amount when extracting the partial image to the first processing parameter set by the first processing setting unit 31 (see FIG. 10), In the partial image creation unit 32, a function of creating a partial image 33a-1 based on the shift amount is added.

  In the solar panel appearance monitoring apparatus according to the present embodiment, in addition to the effects of the first embodiment, the occurrence of false detection is suppressed even when the surveillance camera is shaken, and the appearance of the solar panel is automatically monitored. Can do.

  The present invention is suitable as a solar panel appearance monitoring device.

DESCRIPTION OF SYMBOLS 1 Solar panel external appearance monitoring apparatus 2 Integrated interface 3 Solar panel inspection parameter adjustment apparatus 4a 1st solar panel inspection apparatus 4b 2nd solar panel inspection apparatus 4c 3rd solar panel inspection apparatus 5 Storage 10 Foreign object 10a Projectile 10b vine 11 solar panel 11a first solar panel monitoring camera image 11b second solar panel monitoring camera image 11c third solar panel monitoring camera image 12 small camera 13 pan-tilt mechanism monitoring camera 13a pan-tilt mechanism 14a Discriminator 14b for first solar panel inspection Discriminator 14c for second solar panel inspection Discriminator 15a for third solar panel inspection Partial images 21a-1, 21a-2 Surveillance cameras for the first solar panel Images 21b-1, 21b-2 Monitoring camera images 21c-1, 2 of the second solar panel c-2 Monitoring camera image 21A of the third solar panel Image group 21B of the first solar panel Image group 21C of the second solar panel Image group 22a-1, 22a-2 of the third solar panel Image 22b-1, 22b-2 Second foreign object image 22A First foreign object image group 22B Second foreign object image group 23a-1, 23a-2 Partial image 23A First solar panel partial image collection 23B Second sun Light Panel Partial Image Collection 23C Third Solar Panel Partial Image Collection 24a-1, 24a-2
24A First foreign matter partial image collection 24B Second foreign matter partial image collection 31 First processing setting section 32 First partial image creation section 33 First classifier construction section 33a-1 Partial image 34 First storage section 40 Image input section 41 2nd process setting part 42 2nd partial image creation part 43 2nd discriminator test | inspection part 44 2nd memory | storage part 45 Test result judgment part 46 Test result output part

Claims (5)

Identification for solar panel inspection based on a solar panel image group in which a solar panel is photographed for a certain period of time with a surveillance camera and a foreign object image group in which the surveillance camera is photographed for a certain period of time. A solar panel inspection parameter adjustment device,
At the time of monitoring, an image obtained by photographing the solar panel with the monitoring camera is input to the discriminator, and it is determined whether or not the foreign object exists on the solar panel based on the inspection result of the discriminator. A solar panel appearance monitoring device comprising: a solar panel inspection device. The solar panel inspection parameter adjustment device is:
Each of the solar panel image groups is divided into small areas, the small areas in which the solar panels are reflected are extracted as partial images, and a set of the partial images is a solar panel partial image collection,
In the image group of the foreign matter, set a small region where the portion where the foreign matter is reflected enters, take out this small region as a partial image, a set of the partial images as a foreign matter partial image collection,
The solar panel appearance monitoring device according to claim 1, wherein the discriminator is constructed based on the solar panel partial image collection and the foreign object partial image collection. The solar panel inspection parameter adjustment device is:
In addition to the small area in which the solar panel is reflected, an area shifted from the small area by a predetermined range is taken out as a partial image to form the solar panel partial image collection. The solar panel appearance monitoring device described. The classifier is a multi-class classifier;
The solar panel inspection parameter adjustment device constructs the discriminator based on the solar panel image group and a plurality of types of the foreign object image groups obtained by photographing a plurality of types of the foreign objects for a certain period with the monitoring camera. The solar panel appearance monitoring apparatus according to any one of claims 1 to 3, wherein The solar panel inspection apparatus is
A rectangular portion in which the solar panel is reflected is sequentially extracted as a partial image from the image obtained by photographing the solar panel with the monitoring camera, and the partial image is input to the discriminator. Based on the inspection result of the discriminator. The solar panel appearance monitoring device according to any one of claims 1 to 4, wherein it is determined whether or not the foreign substance exists on the solar panel.

Images