Detecting an anomaly in a solar panel

Abstract

The present invention relates to a computer-implemented method of detecting an anomaly in a solar panel. The computer-implemented method comprises: receiving a thermal image of a solar panel; generating a mask of one or more regions that are above a temperature threshold within the thermal image; applying a size filter to the mask to retain any of the one or more regions having a size less than or equal to a predetermined size and to disregard any of the one or more regions having a size larger than the predetermined size; and sending a signal indicating presence of an anomaly based on the or each retained region being retained.

Description

DETECTING AN ANOMALY IN A SOLAR PANELFIELD

[0001] The subject-matter of the present disclosure relates to detecting anomalies in solar panels, computer-implemented methods of detecting such anomalies in a solar panel, transitory, or non-transitory, computer readable media, and autonomous vehicles.BACKGROUND

[0002] A common issue with solar panels is overheating of individual cells in the panel. When a cell overheats it is called a hotspot. Anomalies such as hotspots can cause permanent damage to the panel if they are not detected early enough. Hotspots occur from micro-cracks inside the panel, e.g. faulty or broken diodes, or non-uniform distribution of lighting across the panel caused by obstructions, e.g., from bird droppings. Both types of hotspots induce performance drops and equipment degradation.

[0003] Hotspots are locations on the panel that are hotter than the surrounding area on the panel, e.g. at least 3 degrees Celsisu. Hence, it is possible to detect hotspots by measuring the temperature of the whole panel and looking at discontinuities in relative heat signatures.

[0004] It is known to implement temperature sensors on a solar panel in an attempt to detect hotspots. A comparison is made between real-time measurements and modelbased prediction results of electric power production for solar panels. The model is estimated by analysing solar irradiance and temperature of the panel measurements. This method required extensive sensor installations on all panels in the solar farm.

[0005] Another method uses thermal imaging cameras to provide visual and thermal data about solar panels, which can be processed to detect hotspots. This operation can be scaled to assess panels across an entire solar farm. A classical computer vision technique is known which combines a canny edge detector and standard thermal image processing to detect faults. However, the approach was only tested on a single solar panel and with a static camera, while in solar farms solar panels may be stacked where edges don’t have the same appearance as a single unit since the boarders of solar panels are adjacent to each other. A more complex feature extraction machine-learning technique is known which achieves good performance. However, such approaches require large amounts of data from different scenarios in the training process, which requires considerable computational capability at runtime. This is a major concern in embedded systems because of limited payload, for example.

[0006] It is an aim of the present invention to address such problems and improve on the prior art.SUMMARY

[0007] According to an aspect of the present disclosure, there is provided a computer- implemented method of detecting an anomaly in a solar panel, the computer- implemented method comprising: receiving a thermal image of a solar panel; generating a mask of one or more regions that are above a temperature threshold within the thermal image; applying a size filter to the mask to retain any of the one or more regions having a size less than or equal to a predetermined size and to disregard any of the one or more regions having a size larger than the predetermined size; and sending a signal indicating presence of an anomaly based on the or each retained region being retained.

[0008] This method provides accurate identification of anomalies without requiring copious training data that may not be available.

[0009] In an embodiment, the anomaly is a hotspot. A hotspot is when a cell overheats.

[0010] In an embodiment, the computer-implemented method further comprises: identifying any retained region as a hotspot if a temperature difference between the retained region and the solar panel outside the one or more mask regions is greater than or equal to a temperature difference threshold; identifying any retained region as a potential hotspot if the temperature difference between the retained region and the solar panel outside the one or more mask regions is less than the temperature difference threshold, wherein the sending the signal indicating the presence of the hotspot comprises sending a signal identifying any retained region as a hotspot or a potential hotspot.

[0011] In an embodiment, the temperature threshold is set according a percentage of pixels to be retained from the thermal image. It is more appropriate to use this approach than, for example, an absolute temperature threshold, since other factors such as ambient temperature and time of year may affect the overall temperature of the features in the thermal image.

[0012] In an embodiment, the percentage of pixels to be retained is between 5% and 25%.

[0013] In an embodiment, the percentage of pixels to be retained is about 10% or about

[0014] In an embodiment, the temperature difference threshold is between 3 and 7 degrees Celsius.

[0015] In an embodiment, the predetermined size is between 0.3% and 30% of a total area of the thermal image.

[0016] In an embodiment, the receiving a thermal image of the solar panel comprises: capturing, by a thermal imaging camera mounted on an autonomous vehicle, the thermal image of the solar panel.

[0017] In an embodiment, the autonomous vehicle is a land-based vehicle, and wherein the capturing the thermal image of the solar panel comprises: driving the land based autonomous vehicle past the solar panel; and capturing, using the thermal imaging camera, the thermal image while driving.

[0018] In an embodiment, the driving the land based autonomous vehicle past the solar panel comprises driving the land based autonomous vehicle past the solar panel at a 5 miles per hour or less. In this way, the hotspot will be detected more accurately since its shape will be less likely to distort than at higher temperatures.

[0019] In an embodiment, the computer-implemented method further comprises: receiving, from one or more sensors of the autonomous vehicle, positions of objects in an environment of the autonomous vehicle, and a map including positions of the objects; and localising a position of the autonomous vehicle based on the positions of the objects and the map.

[0020] In an embodiment, the computer-implemented method further comprises: determining a position of the solar panel associated with the identified anomaly based on the localised position of the autonomous vehicle and the position of the identified anomaly in the thermal image. Knowledge of the location of the solar panel can help operators easily locate it and investigate the anomaly.

[0021] In an embodiment, the computer-implemented method further comprises: capturing, using a camera of the autonomous vehicle, an image of the solar panel, wherein the sending the signal indicating the anomaly identification comprises: sending a signal to a display device to display simultaneously the image of the solar panel, the thermal image of the solar panel, and the position of the solar panel, and optionally the identification of the retained region as being a anomaly or a potential anomaly.

[0022] In an embodiment, the computer-implemented method further comprises: performing segmentation on the thermal image to identify pixels associated with the solarpanel and pixels not associated with the solar panel; comparing pixels associated with the identified anomaly to the segmentation; classifying the identified anomaly as a anomaly if the pixels of the identified hotspot correspond to a position associated with the solar panel; and disregarding the identified anomaly if the pixels of the identified hotspot correspond to a position not associated with the solar panel.

[0023] According to an aspect of the present invention, there is provided a transitory, or non-transitory, computer-readable medium, having instructions stored thereon that when executed by one or more processors, cause the one or more processors to perform the computer-implemented method of any preceding aspect or embodiment.

[0024] According to an aspect of the present disclosure, there is provided an autonomous vehicle, comprising: a thermal imaging camera; a processor; and storage including instructions stored thereon that, when executed by the processor, cause the processor to perform the computer-implemented method of any preceding aspect or embodiment.BRIEF DESCRIPTION OF DRAWINGS

[0025] The subject-matter of the present disclosure is best described with reference to the accompanying figures, in which:

[0026] Figure 1 shows a schematic view of an autonomous vehicle according to one or more embodiments;

[0027] Figure 2 shows a flow chart summarising a computer-implemented method of detecting a hotspot in a solar panel according to one or more embodiments;

[0028] Figures 3A and 3B show a red, green, blue (RGB) image and a thermal image, respectively, of a solar panel;

[0029] Figures 4A and 4B show an RGB image and a thermal image, respectively, of another solar panel;

[0030] Figure 5 shows a hotspot detecting evaluation graph plotting performance metrics against parameter variation, for hotspot detection without a size filter;

[0031] Figures 6A and 6B shows thermal images without and with hotspots, respectively;

[0032] Figure 7 shows a hotspot detecting evaluation graph plotting performance against parameter variation for hotspot detection with a size filter;

[0033] Figures 8A and 8B show an RGB image and a thermal image, respectively, of a solar panel with artificial hotspots added for training;

[0034] Figures 9A, 9B, and 9C, show thermal images with hotspots detected at different speeds of a vehicle carrying a thermal camera capturing the thermal images;

[0035] Figure 10 shows a plan view of a solar farm including a plurality of rows of solar panels;

[0036] Figure 11 shows a thermal image of a solar panel with false positive hotspots detected outside the solar panel;

[0037] Figures 12A and 12B show an RGB image and a thermal image of a solar panel showing hotspots caused by bird droppings;

[0038] Figure 13 shows a display showing results from the method of detecting a hotspot in a solar panel; and

[0039] Figure 14 shows a flow chart summarising a computer-implemented method of detecting a hotspot on a solar panel according to one or more embodiments.DESCRIPTION OF EMBODIMENTS

[0040] At least some of the example embodiments described herein may be constructed, partially or wholly, using dedicated special-purpose hardware. Terms such as ‘component’, ‘module’ or ‘unit’ used herein may include, but are not limited to, a hardware device, such as circuitry in the form of discrete or integrated components, a Field Programmable Gate Array (FPGA) or Application Specific Integrated Circuit (ASIC), which performs certain tasks or provides the associated functionality. The hardware may be an embedded system forming a computer in that it involves processing. In some embodiments, the described elements may be configured to reside on a tangible, persistent, addressable storage medium and may be configured to execute on one or more processors. These functional elements may in some embodiments include, by way of example, components, such as software components, program code, drivers, firmware, microcode, circuitry, data, databases, data structures, tables, arrays, and variables. Although the example embodiments have been described with reference to the components, modules and units discussed herein, such functional elements may be combined into fewer elements or separated into additional elements. Various combinations of optional features have been described herein, and it will be appreciatedthat described features may be combined in any suitable combination. In particular, the features of any one example embodiment may be combined with features of any other embodiment, as appropriate, except where such combinations are mutually exclusive. Throughout this specification, the term “comprising” or “comprises” means including the component(s) specified but not to the exclusion of the presence of others.

[0041] The embodiments described herein may be embodied as sets of instructions stored as electronic data in one or more storage media. Specifically, the instructions may be provided on a transitory or non-transitory computer-readable media. When executed by the processor, the processor is configured to perform the various methods described in the following embodiments. In this way, the methods may be computer-implemented methods. In particular, the processor and a storage including the instructions may be incorporated into a vehicle. The vehicle may be an autonomous vehicle (AV).

[0042] Whilst the following embodiments provide specific illustrative examples, those illustrative examples should not be taken as limiting, and the scope of protection is defined by the claims. Features from specific embodiments may be used in combination with features from other embodiments without extending the subject-matter beyond the content of the present disclosure.

[0043] With reference to Figure 1 , an AV 10 may include a plurality of sensors 12. The sensors 12 may be mounted on a roof of the AV 10. The sensors 12 may be communicatively connected to a computer 14. The computer 14 may be onboard the AV 10. The computer 14 may include a processor 16 and a memory 18. The memory may include the non-transitory computer-readable media described above. Alternatively, the non-transitory computer-readable media may be located remotely and may be communicatively linked to the computer 14 via the cloud 20. The computer 14 may be communicatively linked to one or more actuators 22 for control thereof to move the AV 10. The actuators may include, for example, a motor, a braking system, a power steering system, etc.

[0044] The sensors 12 may include various sensor types. Examples of sensor types include LiDAR sensors, RADAR sensors, and cameras, e.g. red, green, blue (RGB) cameras, and thermal imaging cameras. Each sensor type may be referred to as a sensor modality. Each sensor type may record data associated with the sensor modality. For example, the LiDAR sensor may record LiDAR modality data.

[0045] The data may capture various scenes that the AV 10 encounters. For example, a scene may be a visible scene around the AV 10 and may include roads, buildings, weather, objects (e.g. other vehicles, pedestrians, animals, etc.), etc.

[0046] With reference to Figure 2, a computer-implemented method of detecting an anomaly on a solar panel comprises receiving 30 a thermal image of the solar panel, image processing 32, localisation of the anomaly 34, and outputting 36 an output image including the location of the anomaly. The image processing 32 histogram threshold filtering 38, size filtering 40, and temperature difference filtering 42. The anomaly in some embodiments may be a hotspot. Anomalies may include hotspots.

[0047] The histogram thresholding filtering 38 involves generating a mask of high- temperature pixels in the image by filtering out ones that are below a certain threshold. The threshold is defined according to the percentage of high-intensity pixels we want to keep from the range of pixel values. In other words, the histogram thresholding filter 18 comprises generating a mask of one or more regions within the thermal image that are above a temperature threshold, where the temperature threshold is set according to a percentage of pixels to be retained from the thermal image. For instance, there may be a histogram threshold which may be 10%. The temperature threshold may be the temperature at which 90% of the pixels are less than or equal to. In this way, the hottest 10% of pixels form the mask. Therefore, generally, the temperature threshold is variable and is based on the histogram threshold and the range of temperature values of the pixels in the thermal image.

[0048] The size filtering 40 involves comparing the area (in pixels) of each candidate hotspot, e.g. a candidate hotspot, to predefined upper and lower bounds. Candidate hotspots are then discarded if they have unusual shapes or invalid sizes. This is achieved by applying a size filter to the mask to retain any of the one or more regions having a size less than or equal to a predetermined size and to disregard any of the one or more regions having a size larger than the predetermined size.

[0049] The temperature difference filter 42 involves comparing the temperature of each hotspot candidate to the solar panel surface temperature. Candidates meeting a minimum temperature difference may be retained. Candidates not meeting the minimum temperature difference may be discarded in some embodiments and retained in others. This may be achieved by identifying any retained region as a hotspot if a temperature difference between the retained region and the solar panel outside the one or more mask regions is greater than or equal to a temperature difference threshold, and identifyingany retained region as a potential hotspot if the temperature difference between the retained region and the solar panel outside the one or more mask regions is less than the temperature difference threshold. The anomaly may be a hotspot and the potential anomaly may be a potential hotspot.

[0050] The hotspots, or harmful hotspots, may have a temperature difference of 3 degree Celsius or more compared to the panel’s surface temperature. Their presence classifies the solar panel as defective. Potential hotspots, or non-harmful hotspots, exhibit a temperature difference to the surface of the panel of less than 3 degree Celsius. It is useful to notice and record these hotspots for continued monitoring and possible human inspection. However, the potential hotspots do not necessarily render the panel as defective.

[0051] The localisation of the hotspot 34, or localisation (GPS) fusion, is described in more detail below. However, in summary, an autonomy stack on the computer 14 of the AV 10 is able to localise the AV 10 in its environment based on positions of objects that are sensed by the sensors 12 and their positions on a map which is also stored on the storage 18 of the computer 14. The position of the hotspot relative to the AV 10 can be used in conjunction with the localisation of the AV 10 to determine a position of the hotspot and thus the solar panel with the hotspot.

[0052] For training purposes, several thermal images were collected from the thermal imaging camera and labelled as a first step to provide a ground truth dataset of detections. The thermal imaging camera was equipped with a lens that offers a 42 degree by 32 degree field of view. Different image modes were explored during sensor exploration, which were enumerated below.

[0053] Mono16: Monochrome 16 bits images (with linear temperature signal per pixel).

[0054] Mono8: Monochrome 8 bits Y800 images.

[0055] YU V422_8_U YVY: Colourized YUV422 images.

[0056] The mono16 image mode was chosen since it offered the best input for the detection algorithm by including temperature readings in each image pixel. While the resolution of the temperature signal in the camera was set to 10 mK.

[0057] With reference to Figure 3A, a solar panel 44 was used to capture the thermal images in a first data collection. Figure 3B shows a thermal image of the solar panel 44.

[0058] With reference to Figure 4A, a solar panel 44 was used to capture the thermal images in a second data collection. As shown in Fig. 4A, the camera 12 was mounted on a mobile trolley 46, positioned 120 cm off the ground, pointing directly at the solar panels 44. To simulate an autonomous inspection vehicle, the trolley was moved alongside the solar panel array to collect thermal images. Figure 4B shows a thermal image of the solar panel 44.

[0059] With reference to Figure 5, data from the first and second data collections was used to tune the parameters of the size filter 40 (Figure 2).

[0060] In the initial training, the size filter 40 was deactivated. Figure 5 shows the results of testing the hotspot detection when varying the histogram threshold and the minimum temperature difference threshold. Each threshold, or parameter, choice was executed against all collected images from previous steps, which enabled the assessment of results against ground truth labels and generated precision and recall metrics. The recall metric is above 80% in most cases. Precision, however, is below expected values due to false positive detections. In the following training step, activating the size filter 40 resolves this issue.

[0061] With reference to Figures 6A and 6B, after comparing the size of all hotspots observed from the first and second data collections to only the false positives, an upper and a lower size limit were introduced in the size filter 40 (respectively 30% and 0.3% of the image area) as shown in each of Figures 6A and 6B. In this way, the predetermined size is between 0.3% and 30% of a total area of the thermal image. These infrequent cases correspond to non-faulty solar panels as they do not generate any individual heated cells within the solar panels. The system was tuned to ignore these cases to increase precision, reducing the false positive detection rate. The results from testing hotspot detection with the size filter 40 presented in Figures 6A and 6B. We observed fewer false positive cases, and the precision increased to 75%.

[0062] Minor white stains 50 in Figure 6B did not cause hotspots but appear to be something on top of the solar panel 44. Since data collection occurred during a hot summer season, the whole solar panel in Figure 6A was evenly heated. The scale on the grayscale colour bars 52 represents pixel temperatures in degrees Celsius.

[0063] With reference to Figure 7, the training and fine-tuning of the hotspot detection system resulted in a set of optimal parameters. The selected parameters focus was put into detecting hotspots that have a 3°C temperature difference to the rest of the solar panel surface. The best range of parameter sets was found when the histogramthreshold was 10% and the minimum temperature difference was 7 degrees Celsius, and when the histogram threshold was 20% and the minimum temperature difference was 4 degrees Celsius. This was for the size filtering had an upper bound of 30%. In this way, it can be said that the percentage of pixels to be retained is between 5% to 25% or more preferably either 10% or 20%. The temperature difference threshold is between 3 and 7 degrees Celsius, and preferably is 3 degrees Celsius. In other words, the temperature threshold is at least 3 degrees Celsius.

[0064] With reference to Figures 8A and 8B, after the algorithm had been developed and tuned, the system’s extrinsic parameters were tested (namely speed and camera mount positions) while giving a demo at a testing site. The data collected in this exercise was done using the platform and mounting presented below on two solar panels with artificial hotspots.

[0065] The first external parameter is the speed at which the mobile platform carrying the inspection can move without affecting detection results. The mobile platform in this instance was either a manual or an autonomous vehicle. In this way, the method comprises receiving a thermal image (Figure 8B) of the solar panel 44 by capturing, by a thermal imaging camera mounted on an autonomous vehicle, the thermal image of the solar panel 44.

[0066] It should be noted that a known method for hotspot detection is the usage of drones (fixed-wing or multi-rotor) equipped mounted with thermal and RGB cameras. This approach has the advantage of quickly covering large areas, by inspecting multiple rows at the same time while flying at low altitudes. However, on the other hand, the drawbacks of using a drone for inspection can be enumerated as follows. There is limited operational time to weather conditions, battery life, and operator (human) cognitive duration, when using drones. Furthermore, when using drones, there is an inability to determine the exact panel where a hotspot occurs due to visual aliasing and poor localisation / association. Additionally, when using drones, there is limited onboard embedding space and payload capacity for additional equipment to perform multiple tasks at the same time.

[0067] Another possibility is to use a ground vehicle, or land-based vehicle, equipped with a thermal camera that can continuously patrol the entire site. Despite being slower in inspection, this approach can offer a more proximate view over the inspected elements; including more than only solar panel hotspots (inverters, sub-stations, cables, etc.). A ground vehicle can also provide more payload capacity for embedding sensorsand actuators, to perform multiple tasks while searching for hotspots (cleaning, basic maintenance, monitoring equipment, etc.).

[0068] In this way, the mobile asset is a land-based vehicle, wherein capturing the thermal image of the solar panel comprises driving the land based vehicle past the solar panel, and capturing, using the thermal imaging camera, the thermal image while driving. This method occurs in the training phase and also during run-time, in-use.

[0069] During training, two artificial hotspots 54 were created using PVC duct tape stuck to the surface of the solar panel, as shown below. This action aims to create a groundtruth hotspot that the detection system must pick up.

[0070] Figure 9A shows the artificial hotspots 54 being detected at 2.3 miles per hour (mph), or 3.7 kilometres per hour (kph). Figure 9B shows the artificial hotspots 54 being detected at 5 mph or 8 kph. Figure 9C shows the artificial hotspot 54 being detected at 7 mph or 11 kph. The two marks on each thermal image of Figures 9A through 9C show the detection of the artificial hotspots 54 because they have a temperature difference of more than 3 degrees Celsius compared to the temperature of the solar panel 44.

[0071] With reference to Figures 9A through 9C, when the mobile asset travelled at speeds up to 5 mph, the detection of the hotspots 54 was clear and the shape and size of the hotspots 54 were captured in the thermal image accurately. The shape and size of the artificial hotspots 54 started to degrade above 5 mph. Therefore, 5 mph was selected as an upper speed limit for the mobile asset. In other embodiments, another speed limit could be applied, e.g. 7 mph. In this way, in some embodiments, driving the land based autonomous vehicle past the solar panel comprises driving the land based autonomous vehicle past the solar panel at a 5 miles per hour or less.

[0072] With reference to Figure 10, the autonomous vehicle 10 traverses a route 56 around a solar farm. The solar farm includes a plurality of rows of solar panels 44. Each row of includes a plurality of solar panels 44. When traversing a training route, the AV 10 may detect the artificial hotspots 54. When traversing the training route, or even when the route is a run-time route, the AV 10 may detect actual hotspots 58.

[0073] During the training route traversals, all hotspot detections were associated with RGB images, tagged with Universal Transverse Mercator (UTM) coordinates and the data was saved in both text and media files in real-time. Hence, a full report of all hotspots is available for download after the vehicle finishes its mission during each deployment.Moreover, given a wireless connection between the vehicle and the site operations, these results can be transmitted in real-time to the site operator.

[0074] With reference to Figure 11 , on occasion, certain hotspot detections may be false positive detections. Some of these false positive detections 60 may occur on the solar panel 44. Those false positive detections 60 may be disregarded as hotspots if their size is outside the lower or upper bounds of the size filter. Other false positive detections 62 may occur off the solar panel 44. Even if these false positive detections are within the upper and lower bounds of the size filter, they may be identified by recognising that they are not on the solar panel 44. This may be achieved as follows.

[0075] The method may include performing segmentation on the thermal image to identify pixels associated with the solar panel and pixels not associated with the solar panel; comparing pixels associated with the identified hotspot to the segmentation; classifying the identified hotspot as a hotspot if the pixels of the identified hotspot correspond to a position associated with the solar panel; and disregarding the identified hotspot if the pixels of the identified hotspot correspond to a position not associated with the solar panel.

[0076] With reference to Figures 12A and 12B, bird droppings 64 cover half of a solar panel cell. However, the bird droppings 64 do not generate any heating of the cell. Although bird droppings on panels can cause hotspots, the data collected in the training routes shows that it is a minority of cases. Specifically, related to detection, these data imply that an RGB camera along tuned to detect bird dropping would not be sufficient to detect hotspots. A thermal camera is required to ensure that hotspots are detected with high accuracy.

[0077] With reference to Figure 13, the method may also comprise sending a signal indicating presence of a hotspot based on the or each retained region being retained, wherein the sending the signal indicating the presence of the hotspot comprises sending a signal identifying any retained region as a hotspot or a potential hotspot. More specifically, the sending the signal indicating the hotspot identification comprises sending a signal to a display device 66 to display simultaneously the image (e.g. RGB image) 68 of the solar panel, the thermal image 70 of the solar panel, and the position 72 of the solar panel, and optionally the identification of the retained region as being a hotspot 74 or a potential hotspot 76. The potential hotspot may be considered a non-harmful hotspot in some embodiments. Other features may be presented too including the vehicle speed78, the mode of operation 80 of the vehicle, e.g. manual or autonomy, the index 82 of the data, and a date and time of detection 84.

[0078] In order to generate the position 72 of the solar panel, the mobile asset uses a map including the positions of all solar panels, the sensors 12 (Figure 1), and a localisation function of the autonomy stack. As described above, an autonomy stack on the computer 14 of the AV 10 is able to localise the AV 10 in its environment based on positions of objects that are sensed by the sensors 12 and their positions on a map which is also stored on the storage 18 of the computer 14. The position of the hotspot relative to the AV 10 can be used in conjunction with the localisation of the AV 10 to determine a position of the hotspot and thus the solar panel with the hotspot. The position of the solar panel is provided in the form of GPS co-ordinates in some embodiments. In other embodiments, the position may be provided in the form of the row and / or column and / or solar panel number.

[0079] In this way, the method comprises receiving, from one or more sensors of the autonomous vehicle, positions of objects in an environment of the autonomous vehicle, and a map including positions of the objects; and localising a position of the autonomous vehicle based on the positions of the objects and the map. The method may also comprise determining a position of the solar panel associated with the identified hotspot based on the localised position of the autonomous vehicle and the position of the identified hotspot in the thermal image.

[0080] With reference to Figure 14, a computer-implemented method of detecting an anomaly in a solar panel can be summarised as including a plurality of steps comprising: receiving S100 a thermal image of a solar panel; generating S102 a mask of one or more regions that are above a temperature threshold within the thermal image; applying S104 a size filter to the mask to retain any of the one or more regions having a size less than or equal to a predetermined size and to disregard any of the one or more regions having a size larger than the predetermined size; and sending S106 a signal indicating presence of anomaly based on the or each retained region being retained.

[0081] While the invention has been illustrated and described in detail in the drawings and foregoing description, such illustration and description are to be considered illustrative or exemplary and not restrictive; the invention is not limited to the disclosed embodiments.

[0082] Other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings,the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality. A single processor or other unit may fulfil the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measured cannot be used to advantage. Any reference signs in the claims should not be construed as limiting the scope.

Claims

CLAIMS1. A computer-implemented method of detecting an anomaly in a solar panel, the computer-implemented method comprising: receiving a thermal image of a solar panel; generating a mask of one or more regions that are above a temperature threshold within the thermal image; applying a size filter to the mask to retain any of the one or more regions having a size less than or equal to a predetermined size and to disregard any of the one or more regions having a size larger than the predetermined size; and sending a signal indicating presence of an anomaly based on the or each retained region being retained.

2. The computer-implemented method of Claim 1 , wherein the anomaly is a hotspot.

3. The computer-implemented method of Claim 2, further comprising: identifying any retained region as a hotspot if a temperature difference between the retained region and the solar panel outside the one or more mask regions is greater than or equal to a temperature difference threshold; identifying any retained region as a potential hotspot if the temperature difference between the retained region and the solar panel outside the one or more mask regions is less than the temperature difference threshold, wherein the sending the signal indicating the presence of the hotspot comprises sending a signal identifying any retained region as a hotspot or a potential hotspot.

4. The computer-implemented method of any preceding claim, wherein the temperature threshold is set according a percentage of pixels to be retained from the thermal image.

5. The computer-implemented method of Claim 4, wherein the percentage of pixels to be retained is between 5% and 25%.

6. The computer-implemented method of Claim 5, wherein the percentage of pixels to be retained is about 10% or about 20%.

7. The computer-implemented method of any preceding claim, wherein the temperature difference threshold is between 3 and 7 degrees Celsius.

8. The computer-implemented method of any preceding claim, wherein the predetermined size is between 0.3% and 30% of a total area of the thermal image.

9. The computer-implemented method of any preceding claims, wherein the receiving a thermal image of the solar panel comprises: capturing, by a thermal imaging camera mounted on an autonomous vehicle, the thermal image of the solar panel.

10. The computer-implemented method of Claim 8, wherein the autonomous vehicle is a land-based vehicle, and wherein the capturing the thermal image of the solar panel comprises: driving the land based autonomous vehicle past the solar panel; and capturing, using the thermal imaging camera, the thermal image while driving.

11. The computer-implemented method of Claim 9, wherein the driving the land based autonomous vehicle past the solar panel comprises driving the land based autonomous vehicle past the solar panel at a 5 miles per hour or less.

12. The computer-implemented method of any of Claims 8 to 10, further comprising: receiving, from one or more sensors of the autonomous vehicle, positions of objects in an environment of the autonomous vehicle, and a map including positions of the objects; and localising a position of the autonomous vehicle based on the positions of the objects and the map.

13. The computer-implemented method of Claim 11 , further comprising:determining a position of the solar panel associated with the identified anomaly based on the localised position of the autonomous vehicle and the position of the identified anomaly in the thermal image.

14. The computer-implemented method of Claim 13, further comprising: capturing, using a camera of the autonomous vehicle, an image of the solar panel, wherein the sending the signal indicating the anomaly identification comprises: sending a signal to a display device to display simultaneously the image of the solar panel, the thermal image of the solar panel, and the position of the solar panel, and optionally the identification of the retained region as being a anomaly or a potential anomaly.

15. The computer-implemented method of any preceding claim, further comprising: performing segmentation on the thermal image to identify pixels associated with the solar panel and pixels not associated with the solar panel; comparing pixels associated with the identified anomaly to the segmentation; classifying the identified anomaly as a anomaly if the pixels of the identified hotspot correspond to a position associated with the solar panel; and disregarding the identified anomaly if the pixels of the identified hotspot correspond to a position not associated with the solar panel.

16. A transitory, or non-transitory, computer-readable medium, having instructions stored thereon that when executed by one or more processors, cause the one or more processors to perform the computer-implemented method of any preceding claim.

17. An autonomous vehicle, comprising: a thermal imaging camera; a processor; andstorage including instructions stored thereon that, when executed by the processor, cause the processor to perform the computer-implemented method of any of Claims 1 to 14.

Images