Method and system for object detection

Abstract

Methods and systems for object detection. A computer-implemented method for object detection comprises the following steps performed by a computer hardware component: acquiring an image of an environment; determining at least one reflection in the image, the at least one reflection being at least one reflection of a light source of an occluded object; determining enhanced reflection information based on the at least one reflection; and detecting the occluded object based on the enhanced reflection information.

Description

Technical Field

[0001] This disclosure relates to methods and systems for object detection. Background Technology

[0002] Advanced driver assistance systems (ADAS) are expected to reliably detect other road users at greater distances. However, obstruction of direct line of sight can become a problem, as it renders all the most advanced sensors—such as lidar, radar, and cameras—"blind." This can lead to dangerous situations, as objects may appear suddenly, and users are often led to place too much trust in the assistance system when they do not fully understand its (physical) capabilities.

[0003] Therefore, there is a need to provide enhanced object detection methods and systems. Summary of the Invention

[0004] This disclosure provides a computer-implemented method, a computer system, a vehicle, and a non-transitory computer-readable medium.

[0005] In one aspect, this disclosure relates to a computer-implemented method for object detection, the method comprising the steps performed by computer hardware components: acquiring an image of an environment; determining at least one reflection in the image, the at least one reflection being at least one reflection of a light source of an occluded object; determining enhanced reflection information based on the at least one reflection; and detecting the occluded object based on the enhanced reflection information.

[0006] In other words, headlight reflections can be used to detect obscured vehicles.

[0007] According to various implementations, the obscured object may include or be a vehicle.

[0008] Reflection can be understood as the change in direction of an optical wavefront at the interface between two different media (e.g., between air and a parked vehicle), causing the wavefront to return to the medium of its origin.

[0009] According to one embodiment, the enhanced reflection information includes multiple reflection feature patterns, and the occluded object is detected based on the feature patterns. The feature patterns may include features such as pixel brightness, and / or reflection area, and / or pixel color value, and / or motion of the reflection relative to the reflector.

[0010] According to one implementation, the enhanced reflection information includes information about how the reflection changes over time, and the occluded object is detected based on the changes. For example, if the reflection intensity increases over time, this can be considered to indicate that the reflection is related to an approaching object.

[0011] According to one embodiment, the computer-implemented method further includes the steps performed by the computer hardware components: detecting visible objects in the environment; and determining whether the visible objects are static objects; wherein, if the visible objects are determined to be static objects, the at least one reflection includes at least one reflection on the determined visible object. Reflections on static objects may change solely due to relative motion between the observer (e.g., the vehicle itself) and the obscured object.

[0012] According to one embodiment, determining whether the visible object is a static object includes determining an image of the visible object and / or lidar measurements and / or radar measurements associated with the visible object. Based on the image and / or lidar / radar measurements, it can be determined whether the visible object is in motion.

[0013] According to one embodiment, determining whether the visible object is a static object includes determining the orientation of the visible object. For example, for a visible object that is a vehicle, if the vehicle's orientation is perpendicular to its direction of travel, it can be determined that the visible object is moving, as this may indicate that the visible object is turning into the road on which the vehicle is traveling.

[0014] According to one embodiment, the computer-implemented method further includes the following steps performed by the computer hardware components: determining context information; determining, based on the context information, whether to execute the computer-implemented method; and executing the computer-implemented method only when it is determined that the computer-implemented method should be executed. It has been found that executing the method only under specific circumstances can reduce computational workload, particularly when object detection results are not required.

[0015] According to one implementation, the contextual information includes at least one of the following: daytime time information, environmental traffic conditions, road width information, or visibility information.

[0016] According to one embodiment, the computer-implemented method further includes the following steps performed by the computer hardware components: determining tracking information of the at least one reflection; wherein the enhanced reflection information is determined based on the tracking information. Using not only a single reflection but also multiple reflections at different time steps, particularly tracking information about the reflections (e.g., the trajectory of the reflection), can improve detection accuracy.

[0017] According to one embodiment, the method is performed relative to the vehicle, and the method further includes at least one of the following steps performed by the computer hardware components: outputting a warning to the driver of the vehicle based on the detection of the obstructed object; or influencing a planning algorithm for the vehicle's autonomous driving functions to reduce the risk of collision with the obstructed object. For example, if an obstructed object is detected, a warning may be provided to the driver of the vehicle regarding the obstructed object, or the vehicle's trajectory may be altered to avoid a collision between the vehicle and the obstructed object.

[0018] According to one embodiment, detecting the occluded object includes determining at least one of the distance of the occluded object, the speed of the occluded object, or the direction of movement of the occluded object.

[0019] According to one embodiment, the method is performed relative to the vehicle, and the method further includes the following step performed by the computer hardware component: determining a collision risk between the vehicle and the occluded object based on the detection of the occluded object.

[0020] In another aspect, this disclosure relates to a computer system comprising multiple computer hardware components configured to perform several or all of the steps of the computer-implemented methods described herein. The computer system may be part of a vehicle.

[0021] The computer system may include multiple computer hardware components (e.g., a processor, such as a processing unit or processing network, at least one memory, such as a memory unit or memory network, and at least one non-transitory data storage device). It should be understood that additional computer hardware components may be provided and used in the computer system to perform the steps of the computer-implemented method. The non-transitory data storage device and / or the memory unit may include a computer program for instructing the computer to perform several or all steps or aspects of the computer-implemented method described herein, for example, using the processing unit and the at least one memory unit.

[0022] In another aspect, this disclosure relates to a vehicle that includes a computer system as described herein and sensors configured to acquire the images.

[0023] In another aspect, this disclosure relates to a non-transitory computer-readable medium comprising instructions for performing (in other words, including instructions that, when executed by a computer system (e.g., the computer system described herein), cause the computer system to perform) several or all of the steps or aspects of the computer-implemented method described herein. The computer-readable medium may be configured as: an optical medium, such as an optical disc (CD) or digital versatile disk (DVD); a magnetic medium, such as a hard disk drive (HDD); a solid-state drive (SSD); a read-only memory (ROM), such as flash memory; and so on. Furthermore, the computer-readable medium may be configured as a data storage device accessible via a data connection (e.g., an internet connection). The computer-readable medium may be, for example, an online data warehouse or cloud storage.

[0024] This disclosure also relates to a computer program for instructing a computer to perform several or all of the steps or aspects of the computer-implemented method described herein. Attached Figure Description

[0025] Exemplary embodiments and functions of this disclosure are illustrated in conjunction with the following drawings, in which:

[0026] Figure 1 A schematic diagram of a system for detecting an obscured vehicle based on headlight reflection, according to various embodiments, is shown.

[0027] Figure 2 A schematic diagram illustrating an example of a bottleneck road condition is shown.

[0028] Figure 3 It schematically shows the situation in relation to Figure 2 The diagram shows the output of the bright spot detector / tracker in the same scenario.

[0029] Figure 4 The flowcharts illustrating various embodiments of methods for object detection are shown schematically; and

[0030] Figure 5 A computer system having multiple computer hardware components configured to perform steps of a computer-implemented method for object detection according to various embodiments is schematically shown.

[0031] List of reference numerals

[0032] 100 Schematic diagrams of systems for detecting obscured vehicles based on headlight reflection, according to various embodiments.

[0033] 102 Blocks used for detecting parked vehicles

[0034] 104 Blocks for identifying applicable scenarios

[0035] 106 Decision Blocks

[0036] 108 Highlight Detection / Classification Blocks

[0037] 110 Blocks for direction estimation

[0038] 112 Blocks for risk estimation and output

[0039] 114 cameras

[0040] 116 One or more radars

[0041] 118 One or more lidar units

[0042] 120 Road Map

[0043] 122 Environmental Classifier

[0044] 124 Object Detector

[0045] 126 Trackers and / or classifiers

[0046] 128 detectors

[0047] 130 This car is in motion

[0048] 132 classifier

[0049] 134 Direction Estimation Block

[0050] 136 Simplified Chassis Model

[0051] Risk Level 138

[0052] 200. Schematic diagrams depicting examples of bottleneck road conditions.

[0053] 300 in relation to Figure 2 The output diagram of the bright spot detector / tracker in the same scene is shown.

[0054] 302. Illustration of the box containing the output of the bright spot detector / tracker.

[0055] Figure 400 shows flowcharts of methods for object detection according to various embodiments.

[0056] 402 Steps for acquiring an image of the environment

[0057] 404 The step of determining at least one reflection in an image, wherein the at least one reflection is at least one reflection of a light source of an occluded object.

[0058] 406 Step of determining enhanced reflection information based on at least one reflection

[0059] 408 Steps for detecting occluded objects based on the enhanced reflection information

[0060] 500 Computer systems according to various implementation methods

[0061] 502 processor

[0062] 504 memory

[0063] 506 Non-transitory data storage device

[0064] 508 camera

[0065] 510 Distance Sensor

[0066] 512 connections Detailed Implementation

[0067] Advanced driver assistance systems (ADAS) are expected to reliably detect other road users at greater distances. However, obstruction of direct line of sight can become a problem, as it renders all the most advanced sensors—such as lidar, radar, and cameras—"blind." This can lead to dangerous situations, as objects may appear suddenly, and users are often led to place too much trust in the assistance system when they do not fully understand its (physical) capabilities.

[0068] To overcome the problem of drivers switching to low beams too late, thus dazzling oncoming drivers, or not using high beams at all to avoid dazzling other drivers, current automatic high beam algorithms evaluate camera images to detect the headlights and taillights of other vehicles and can switch to low beams without requiring a quick manual reaction from the driver. However, to avoid false alarms and overly frequent switching, these algorithms must wait to see multiple frames of headlights to obtain a high-confidence classification. During this period, other drivers will be dazzled.

[0069] According to various implementations, an object detection method is provided that, like many human drivers, can predict other vehicles by observing indirect light scattering (e.g., by recognizing that the road is illuminated by vehicles that are not yet visible).

[0070] According to various implementations, methods and systems are provided for estimating obscured oncoming traffic participants using headlight reflections from parked vehicles to address bottleneck sections and similar scenarios in urban environments for autonomous driving. This can provide the ability to decelerate / stop at bottleneck sections, intersections (planning / strategy), and single-lane tunnels, and prevent deadlock situations, thereby improving safety and passenger comfort (e.g., eliminating the need for abrupt deceleration, reducing the need for driver intervention due to the autonomous driving system's inability to handle deadlock situations, as reversing from a deadlock scenario is highly undesirable for autonomous driving).

[0071] Figure 1 A schematic diagram 100 of a system for detecting an obscured vehicle based on headlight reflection (e.g., on a parked vehicle) according to various embodiments is shown.

[0072] Block 102 for detecting parked vehicles may include an object detector 124 and a tracker and / or classifier 126. The object detector 124 may receive input data from a camera 114 and / or one or more radars 116 and / or one or more lidars 118, and may detect objects. If the object detector 124 detects a car, its output data may be provided to the tracker and / or classifier 126. The tracker and / or classifier 126 may determine whether the detected car is a static car (in other words: a parked car). The output of block 102 for detecting parked vehicles may be provided to block 104 for identifying applicable scenarios and decision block 106.

[0073] According to various implementations, in order to interpret reflections in parked vehicles, suitable sensors can be used to detect them. These sensors may include, for example, radar sensors, lidar sensors, and / or camera sensors (the latter encompassing both stereo vision and monocular vision), or any combination of these sensors. Radar sensors may have the advantage of obtaining single-frame speeds. LiDAR may provide particularly accurate distance and direction information. Camera sensors may allow for the most accurate classification. Fusion of all sensor types may allow for the most reliable and detailed detection.

[0074] According to various implementations, to obtain reliable information from reflections from these vehicles, it can be ensured that these vehicles are not moving but static, parked, or currently stationary. Vehicle movement can interfere with the information conveyed by reflections from the target vehicle because the movement of the reflecting vehicle affects light reflection independently of the target vehicle. Therefore, after detecting a vehicle, it can be classified as either moving or stationary. If a radar sensor is used, it may provide the strongest signal because it can use the Doppler frequency shift signal to estimate the object's velocity. LiDAR and camera sensors can determine the object's velocity by comparing two or more sensor frames.

[0075] Another available source of information for classifying cars as stationary or moving is the location and orientation of objects in a high-resolution (HD) map. Vehicles perpendicular to the street are likely leaving a property entrance or exit, or vehicles at an angle to the road may be leaving a parking space.

[0076] Block 104, used to identify the applicable scenario, can detect the driving scenario as a bottleneck section, an intersection, a private driveway, or any other scenario from a set of predefined scenarios. The input to block 104 can include the road map 120 and / or the output of the environment classifier 122. The output of block 104 can be provided to decision block 106.

[0077] To reduce workload and false alarm notifications, methods according to various implementations can be applied only in appropriate contexts. These contexts can be determined based on one or more of the following:

[0078] 1. Daytime Information: For example, the method can be performed only during suitable daytime hours and / or when suitable lighting conditions are present. To assess light reflection on the body of a parked vehicle, the ambient brightness must be low enough to enhance reflection, and the ambient ratio can be determined to a reasonable value.

[0079] 2. Environment: Since the system is designed for narrow roads and a sufficient density of parked cars, which is only reasonably possible in urban environments, information about this characteristic can be used.

[0080] 3. Road width: If the road is wide enough for oncoming vehicles to pass without problems, then early information about oncoming vehicles may not be of much value. Methods according to various implementations should only be used when the road is too narrow for two vehicles to pass without risk, and one vehicle may have to yield to the other.

[0081] 4. Visibility limitations: If the road is curved, the line of sight to oncoming vehicles may be reduced, and information about vehicles that are not yet visible can improve safety. On the other hand, if the road is straight, the visibility range may be increased, and oncoming vehicles can be seen at a sufficiently large distance without having to perform the methods according to various implementations.

[0082] According to various implementation methods, object detection methods can be applied to intersections. Even if the road itself is straight, the view of the intersection may be obstructed by buildings or parked vehicles.

[0083] To identify the circumstances under which the method should be activated according to various implementation methods, one or more of the following information sources can be used:

[0084] A. Lighting / Daytime. This information source can be obtained through a light condition sensor used for automatic headlight switching. Existing camera modules can replace this sensor.

[0085] B. Road Map. This map provides information about the shape of the road (straight or curved) and whether there are intersecting roads / intersections. The location of obstructing elements (e.g., buildings, trees, hillsides) can be part of the map. Additionally, the map provides information about the vehicle's surroundings.

[0086] C. Radar / LiDAR / Camera-based Environment Classifiers. Based on any of these sensors, machine learning-trained classifiers can provide information about the environment categories, rather than map-based information.

[0087] D. A list of detected cars from block 102 used to detect parked vehicles. The distribution of parked cars can provide information about the road shape, especially when visibility is obstructed.

[0088] The lighting conditions under which the method should be implemented according to various implementation methods can be determined based on the information source A mentioned above.

[0089] The environmental category conditions under which the method should be performed according to various implementation methods can be determined based on the information source B or C mentioned above.

[0090] The road width conditions for implementing the method according to various implementation methods can be determined based on the combination of information sources B and D mentioned above, because even a sufficiently wide road may only leave one lane for one vehicle to pass.

[0091] The visibility range conditions for implementing this method according to various implementation methods can be determined based on the aforementioned information sources B and D. Each road on the map provides lane centerline information; constructing a line of sight along these lane lines determines the current visible range. The visibility range conditions are only satisfied when this range is below a given threshold.

[0092] According to various implementations, the method can be activated if all four conditions (lighting conditions, environmental category conditions, road width conditions, and visibility range conditions) are satisfied.

[0093] Decision block 106 can trigger reflection-based occluded vehicle detection based on its input data, which may include operations performed by bright spot detection / classification block 108 and orientation estimation block 134.

[0094] Decision block 106 can make decisions about subsequent processing steps and trigger them. Suppose that the processing in block 102, which is used to detect parked vehicles, detects a stationary vehicle, and the situation classification in block 104, which is used to identify applicable situations, finds that the current environment is suitable for further processing, then the operations in the following blocks (bright spot detection / classification block 108 and block 110 for orientation estimation) can be triggered.

[0095] The bright spot detection / classification block 108 may include a detector 128 for detecting bright spots on a reflective vehicle and a classifier 132 for classifying moving / static light sources. The detector 128 may provide its output to the classifier 132. The classifier 132 may receive information about the vehicle's motion 130 as further input and provide its output to a block 110 for orientation estimation.

[0096] According to various implementation methods, reflected light spots on stationary vehicles can be detected. The headlights of oncoming vehicles reflect significantly more light onto the body of parked vehicles than other typical urban light sources (such as...). Figure 2 (As shown). It can capture the detection direction of a parked car relative to the road because it contains information about the expected reflection intensity and the position on the parked vehicle, and can be used to further reduce the false alarm rate.

[0097] According to various implementation methods, existing trajectories can be updated using detection results and vehicle data (e.g., vehicle motion and / or yaw rate).

[0098] According to various implementation methods, if the detection result does not match any existing trajectory, a new trajectory can be created.

[0099] Depending on the implementation method, the trajectory of each light source can be classified as moving or static.

[0100] According to various implementation methods, the trajectory likelihood of the reflected headlights can be integrated.

[0101] The orientation estimation block 134 can receive the simplified chassis model 136 as input and provide its output to the orientation estimation block 110.

[0102] Block 110 for direction estimation can estimate the direction of the occluded vehicle and can provide the reflection angle and distance estimates as outputs to block 112 for risk estimation and output.

[0103] Given the output of the bright spot detection / classification block 108 and the estimated direction of the reflecting vehicle, the block 110 for direction estimation can estimate the reflection angle, the driving direction, and the distance to oncoming vehicles within a certain confidence range.

[0104] Block 112, used for risk estimation and output, can provide risk level 138 as output.

[0105] Given the output of block 110 for direction estimation, the probability of an occluded oncoming vehicle entering the planned trajectory of the vehicle can be estimated in situations such as bottlenecks, intersections, or single-lane tunnels. Using this information, a risk level estimate 138 can be issued for use by the planning algorithm to adjust the autonomous driving behavior of the vehicle accordingly or to provide warnings to the driver.

[0106] Figure 2 A scenario diagram 200 depicting an example of a bottleneck section is shown. The headlight reflections of an oncoming vehicle that is obscured are clearly visible several seconds before the bottleneck section itself even becomes visible (top) and remain visible throughout (middle, bottom). Figure 2 An example scenario is given where an obstacle is placed on a road in a residential area, and the reflection of headlights on a parked car can be observed well before oncoming traffic or the bottleneck section itself becomes visible.

[0107] Figure 3 It shows that in relation to Figure 2 The output diagram 300 of the bright spot detector / tracker in the same scenario shown (illustrated by box 302) illustrates robust detection of reflections from an occluded oncoming car headlights.

[0108] Figure 4 A flowchart 400 illustrating a method for object detection according to various embodiments is shown. In 402, an image of the environment can be acquired. In 404, at least one reflection in the image can be determined, which is at least one reflection of a light source from an occluded object. In 406, enhanced reflection information can be determined based on the at least one reflection. In 408, the occluded object can be detected based on the enhanced reflection information.

[0109] According to various embodiments, the enhanced reflection information may include multiple characteristic patterns of reflection, and the occluded object is detected based on the characteristic patterns.

[0110] According to various embodiments, the enhanced reflection information may include information about how the reflection changes over time, and wherein the occluded object is detected based on the changes.

[0111] According to various embodiments, the method may further include: detecting a visible object in the environment; and determining whether the visible object is a static object; wherein, if the visible object is determined to be a static object, the at least one reflection includes at least one reflection on the determined visible object.

[0112] According to various embodiments, determining whether the visible object is a static object may include determining an image of the visible object and / or lidar measurements and / or radar measurements associated with the visible object.

[0113] According to various implementation methods, determining whether the visible object is a static object may include determining the orientation of the visible object.

[0114] According to various embodiments, the method may further include: determining context information; determining, based on the context information, whether to execute the computer-implemented method; and executing the computer-implemented method only when it is determined that the computer-implemented method should be executed.

[0115] According to various implementation methods, the contextual information may include at least one of the following: daytime time information, environmental traffic conditions, road width information, or visibility information.

[0116] According to various embodiments, the method may further include: determining tracking information of the at least one reflection; wherein the enhanced reflection information is determined based on the tracking information.

[0117] According to various embodiments, the method can be performed relative to the vehicle, and the method may further include: outputting a warning to the driver of the vehicle based on detecting the occluded object; or influencing the planning algorithm of the vehicle's autonomous driving function to reduce the risk of collision with the occluded object.

[0118] According to various embodiments, detecting (408) the occluded object may include determining at least one of the distance of the occluded object, the speed of the occluded object, or the direction of movement of the occluded object.

[0119] According to various embodiments, the method can be performed relative to the vehicle, and the method may further include: determining the collision risk between the vehicle and the occluded object based on detecting the occluded object.

[0120] Steps 402, 404, 406, 408, and each of the other steps mentioned above can be executed by computer hardware components.

[0121] Figure 5 A computer system 500 is shown, having multiple computer hardware components configured to perform steps of a computer-implemented method for object detection according to various embodiments. The computer system 500 may include a processor 502, a memory 504, and a non-transitory data storage device 506. A camera 508 and / or a distance sensor 510 (e.g., a radar sensor and / or a lidar sensor) may be provided as part of the computer system 500 (e.g., ...). Figure 5 (as shown), or provided outside of computer system 500.

[0122] Processor 502 can execute instructions provided in memory 504. Non-transitory data storage device 506 can store computer programs, including instructions that can be transferred to memory 504 and then executed by processor 502. Camera 508 can be used to acquire images of the environment. Proximity sensor 510 can be used to determine whether a visible object is a static object.

[0123] Processor 502, memory 504, and non-transitory data storage device 506 can be interconnected, for example, by exchanging electrical signals via electrical connection 512, such as a cable or computer bus or any other suitable electrical connection. Camera 508 and / or distance sensor 510 can be connected to computer system 500, for example, via an external interface, or can be provided as a component of the computer system (in other words: within the computer system, for example, via electrical connection 512).

[0124] The terms “connection” or “link” are intended to include direct “connection” (e.g., via a physical link) or direct “link”, as well as indirect “connection” or indirect “link” (e.g., via a logical link).

[0125] It should be understood that the content described for one of the above methods can be similarly applied to computer system 500.

Claims

1. A computer-implemented method for object detection, the method comprising the following steps performed by computer hardware components: - Obtain an image of the (402) environment; - Determine (404) at least one reflection in the image, wherein the at least one reflection is at least one reflection of a light source of the occluded object; - Determine (406) enhanced reflection information based on the at least one reflection; and - Detect (408) the occluded object based on the enhanced reflection information.

2. The computer-implemented method according to claim 1, in, The enhanced reflection information includes multiple reflection feature patterns, and the occluded object is detected based on the feature patterns.

3. The computer-implemented method according to claim 1 or 2, in, The enhanced reflection information includes information about how the reflection changes over time, and wherein the occluded object is detected based on the changes.

4. The computer-implemented method according to claim 1, further comprising the following steps performed by the computer hardware components: Detect visible objects in the environment; and Determine whether the visible object is a static object; in, If the visible object is determined to be a static object, then the at least one reflection includes at least one reflection on the determined visible object.

5. The computer-implemented method according to claim 4, in, Determining whether the visible object is a static object includes determining the image of the visible object and / or the lidar measurement value associated with the visible object and / or the radar measurement value associated with the visible object.

6. The computer-implemented method according to claim 4 or 5, in, Determining whether the visible object is a static object includes determining the orientation of the visible object.

7. The computer-implemented method of claim 1, further comprising the following steps performed by the computer hardware components: Determine contextual information; Based on the context information, determine whether to execute the computer-implemented method; and The computer-implemented method is executed only when it is determined that the computer-implemented method should be executed.

8. The computer-implemented method according to claim 7, in, The contextual information includes at least one of the following: daytime information, environmental traffic conditions, road width information, or visibility information.

9. The computer-implemented method of claim 1, further comprising the following steps performed by the computer hardware components: Determine the tracking information of the at least one reflection; in, The enhanced reflection information is determined based on the tracking information.

10. The computer-implemented method according to claim 1, wherein, The method is performed relative to the vehicle, and the method further includes at least one of the following steps performed by the computer hardware component: Based on the detection of the obstructed object, a warning is issued to the driver of the vehicle; or The planning algorithm affects the autonomous driving function of the vehicle to reduce the risk of collision with the obscured object.

11. The computer-implemented method according to claim 1, in, Detecting (408) the occluded object includes determining at least one of the distance of the occluded object, the speed of the occluded object, or the direction of movement of the occluded object.

12. The computer-implemented method according to claim 1, in, The method is performed relative to the vehicle, and the method further includes the following steps performed by the computer hardware component: Based on the detection of the obstructed object, the collision risk between the vehicle and the obstructed object is determined.

13. A computer system (500) comprising a plurality of computer hardware components configured to perform a computer-implemented method according to any one of claims 1 to 12.

14. A vehicle comprising a computer system (500) according to claim 13 and a sensor (508) configured to acquire the image.

15. A non-transitory computer-readable medium comprising instructions for performing a computer-implemented method according to any one of claims 1 to 12.

Images