Road defect identification method based on target detection

Abstract

The invention relates to the technical field of road management, in particular to a road defect recognition method based on target detection, which comprises the following steps: setting a road image acquisition point location, periodically acquiring road images at the set road image acquisition point location based on unmanned aerial vehicle equipment, creating a cloud database, storing the collected road image by using a cloud database; the road images are called from the cloud database, and the difference degree between the road images is recognized; setting a road defect judgment interval, and obtaining an identification result of a difference degree between the road images; according to the method, the road images are collected and stored through the unmanned aerial vehicle equipment, so that the positions with defects on the road are judged by comparing the difference degrees of the road images, and the positions with the defects on the road are refined by synchronously combining image segmentation and further comparing the difference degrees; therefore, highway management personnel can be assisted to carry out highway defect problem maintenance work more efficiently.

Description

Technical Field

[0001] The present invention relates to the technical field of highway management, and particularly relates to a highway defect recognition method based on object detection. Background Art

[0002] Highway defects refer to various problems that occur during the use of highways and affect their normal functions and service lives. Common ones include pavement cracks caused by factors such as temperature and load; potholes, mostly caused by rain erosion and vehicle rolling; and subgrade settlement, resulting from poor foundation soil quality or improper filling.

[0003] The invention patent with the application number 202310624084.5 discloses a surface defect recognition method based on small object detection. The method includes: collecting real-time images of the object to be detected to obtain a set of real-time images; preprocessing the set of real-time images and using a feature extraction model to extract features from the preprocessed set of real-time images to obtain a set of image features; using a binary classifier to perform small object recognition on the set of image features, and obtaining a set of multiple recognition regions according to the recognition results; using a softmax loss function to perform loss analysis on the set of multiple recognition regions, and judging whether the loss value in the loss analysis result meets a preset threshold. If it meets, setting the set of multiple recognition regions as a set of target regions; collecting videos of the set of target regions within a preset time window to obtain multiple target region videos; inputting the multiple target region videos into a defect recognition module in the defect recognition model to output a set of multiple surface defects, where the defect recognition module is constructed based on the slowfast network; and inputting the set of multiple surface defects into a defect matching module in the defect recognition model to obtain a set of surface defect recognition results.

[0004] This application aims to solve the problem of "the technical problem that in the prior art, due to the poor processing effect of defect images, it is difficult to detect tiny defects, and the accuracy of defect recognition is not good".

[0005] However, for the scenario of highway defect recognition, due to the randomness of highway damage problems and the large span of highways, highway defect recognition has always been a huge project;

[0006] Therefore, a highway defect recognition method based on object detection is proposed. Summary of the Invention

[0007] Aiming at the above-mentioned shortcomings of the prior art, the present invention provides a highway defect recognition method based on object detection, which solves the technical problems proposed in the above background art.

[0008] To achieve the above object, the present invention is realized through the following technical solutions:

[0009] A method for identifying highway defects based on object detection, comprising:

[0010] Setting highway image acquisition points, periodically acquiring highway images at the set highway image acquisition points based on unmanned aerial vehicle equipment, creating a cloud database, and storing the acquired highway images in the cloud database using the cloud database;

[0011] The acquisition cycle applied in the highway image acquisition stage follows:

[0012]

[0013] where: T is the highway image acquisition cycle; T 0 is the acquisition cycle base; C MAX 、C MIN are the highest temperature and the lowest temperature in the area where the highway is located; n is the number of rainfall times in the area where the highway is located; x is a constant; h i is the rainfall amount of the i-th time in the area where the highway is located; h 0 is the rainfall amount that can cause damage to the highway; p is the number of historical freeze-thaw cycles in the area where the highway is located; p max is the limit freeze-thaw cycle number that the highway can withstand; L is the cumulative settlement of the highway in the current area where the highway is located; L max is the limit settlement that the highway can withstand: ω 1 、ω 2 、ω 3 、ω 4 are weights; γ is a normalization factor;

[0014] wherein, the weights ω 1 、ω 2 、ω 3 、ω 4 are all positive numbers, and the sum of the weights ω 1 、ω 2 、ω 3 、ω 4 is 1, and the normalization factor γ is used to control to be between 0.1 and 1.9;

[0015] Retrieve road images from the cloud database and identify the degree of difference between the road images; set a road defect determination interval, obtain the recognition result of the degree of difference between the road images, compare the recognition result with the set road defect determination interval, and determine whether the recognition result is within the road defect determination interval; if the determination result is no, end, wait for the road images in the cloud database to be updated and stored, and after the update and storage, perform the determination operation again; if the determination result is yes, obtain the two road images from which the recognition result is derived, compare the sub-images of the two road images pairwise according to the corresponding positions, and determine the positions with defects in the road images based on the comparison results of the degrees of difference of the sub-images; obtain the positions with defects in the determined road images, and generate a defect position message to feedback to the user terminal.

[0016] Furthermore, the road image acquisition points are customized by the user terminal. After the road images are acquired, they are synchronously marked based on the coordinates and acquisition time of the road image acquisition points. There are the same number of differentiated storage intervals as the number of road image acquisition points set in the cloud database. The cloud database differentiates and stores the road images collected by the drone equipment based on the differentiated storage intervals, so that the coordinates marked on the road images stored in each differentiated storage interval are the same;

[0017] Among them, the acquisition period applied in the road image acquisition stage is set based on the environmental parameters of the area where the road where the road image acquisition point is located belongs. The environmental parameters include: temperature, rainfall, number of freeze-thaw cycles, and subgrade settlement.

[0018] Furthermore, the acquisition period base T 0 is customized by the user terminal. When calculating the road image acquisition period T, the environmental parameters of the area where the road belongs used are the environmental parameters within the historical time threshold specified by the user terminal;

[0019] f(h i -h 0 ) represents a constraint function. If h i -h 0 ≥0, then f(h i -h 0 ) = h i -h 0 . If h i -h 0 <0, then f(h i -h 0 ) = 0. The rainfall h 0 that can cause damage to the road is customized by the user terminal. The constant x takes the cumulative number of times when f(h i -h 0 ) = 0;

[0020] The maximum number of freeze-thaw cycles p that the road can withstand max and the maximum settlement L that the road can withstand max are user-defined;

[0021] The value of the normalization factor γ follows that the higher the average daily traffic flow of the road history, the smaller the value of the normalization factor γ, and vice versa, the larger the value of the normalization factor γ.

[0022] Furthermore, in the stage of retrieving road images, any storage interval in the cloud database is used as the retrieval target to perform the retrieval operation. All the stored road images in the retrieved storage interval are retrieved at one time, and the marked contents of all the retrieved road images are traversed synchronously. Based on the traversal results, two road images with the earliest and the most recent acquisition times in the marked contents are selected as the targets for identifying the degree of difference, and the degree of difference between the two road images is identified:

[0023]

[0024] In the formula: DISS(a, b) is the degree of difference between the two road images; m is the total number of contour points in the two road images; d pos (a j , b j ) is the Euclidean distance between the two road images at the j-th contour point; α is the weight; u is the curvature sequence obtained after aligning the two road images by the DTW algorithm; is the curvature value of the v-th sequence point of each of the two road images;

[0025] Among them, before calculating the degree of difference between the two road images, they are synchronously converted into contour images. The larger the value of DISS(a, b), the greater the degree of difference between the two road images, and vice versa, the smaller the degree of difference between the two road images. The value of the weight α ranges from 0 to 1.

[0026] Furthermore, when the determination result is no, all the road images stored in the storage interval where the road image from which the recognition result is derived is located are synchronously deleted, and the deletion target is all the road images except the road image with the earliest marked acquisition time;

[0027] In the stage of segmenting the road images, the segmentation operation of the road images is performed based on the road image segmentation processing logic;

[0028] Among them, when the two road images for which the segmentation processing is performed are acquired, the acquisition distance and angle are exactly the same.

[0029] Furthermore, the road image segmentation processing logic is the setting of the number of segments of the road image, and the road image segmentation processing logic is expressed as:

[0030] Set the number of highway image segmentations to 2*2, 3*3, 4*4, ...;

[0031] Take the sub-images obtained from the same positions in two highway images after segmentation as the comparison targets, apply the calculation formula for the difference degree between the two highway images, calculate the comparison targets, and simultaneously apply the highway defect determination interval for determination;

[0032] When there are calculation results in the corresponding sub-images obtained from the segmentation of two highway images that meet the highway defect determination interval and there are also determination results that do not meet the highway defect determination interval, end.

[0033] Furthermore, the positions with defects in the highway image are the positions of the sub-images that meet the highway defect determination interval in the highway image;

[0034] After determining the positions with defects in the highway image, configure real-world coordinates for the highway image synchronously. Combine the positions with defects in the highway image with the real-world coordinates configured for the highway image to determine the positions with defects in the corresponding area of the highway image.

[0035] Furthermore, the content of the defect position message is the corresponding real-world coordinates of the defect position. After the defect position message is generated, it is transmitted to the cloud database synchronously, and the user terminal retrieves the defect position message from the cloud database.

[0036] Adopting the technical solution provided by the present invention, compared with the known public technologies, it has the following beneficial effects:

[0037] The present invention provides a method for identifying highway defects based on object detection. During the execution of this method, highway images are collected and stored by drone equipment, so as to determine the positions with defect problems on the highway by comparing the sequential difference degrees of the highway images. Synchronously, combined with image segmentation and further comparison of the difference degrees, refine the positions with defect problems on the highway, thereby assisting highway management personnel to carry out highway defect problem maintenance work more efficiently, effectively improving the efficiency and intelligence level of highway defect identification, ensuring that the highway pavement can be comprehensively monitored, and guaranteeing the functionality and safety of the highway. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0039] Figure 1It is a schematic flow diagram of a highway defect recognition method based on object detection. Detailed implementation mode

[0040] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0041] The present invention will be further described below with reference to embodiments.

[0042] Embodiment:

[0043] A highway defect recognition method based on object detection in this embodiment, as Figure 1 shown, includes:

[0044] Set the highway image acquisition points, periodically acquire highway images at the set highway image acquisition points based on the unmanned aerial vehicle (UAV) equipment, create a cloud database, and use the cloud database to store the acquired highway images;

[0045] The highway image acquisition points are user-defined. After the highway images are acquired, they are synchronously marked based on the coordinates and acquisition time of the highway image acquisition points. There are the same number of differentiated storage intervals as the number of highway image acquisition points set in the cloud database. The cloud database differentiates and stores the highway images acquired by the UAV equipment based on the differentiated storage intervals, so that the coordinates marked on the highway images stored in each differentiated storage interval are the same;

[0046] Among them, the acquisition period applied in the highway image acquisition stage is set based on the environmental parameters of the area where the highway where the highway image acquisition points are located belongs. The environmental parameters include: temperature, rainfall, number of freeze-thaw cycles, and subgrade settlement;

[0047] The acquisition period applied in the highway image acquisition stage follows:

[0048]

[0049] In the formula: T is the highway image acquisition period; T 0 is the acquisition period base; C MAX , C MIN are the highest temperature and the lowest temperature in the area where the highway belongs; n is the number of rainfall times in the area where the highway belongs; x is a constant; h i is the rainfall amount of the i-th time in the area where the highway belongs; h 0Rainfall that can cause damage to the road; p is the number of historical freeze-thaw cycles in the area where the road is located; p max is the maximum number of freeze-thaw cycles that the road can withstand; L is the cumulative settlement of the road in the area where the road is currently located; L max is the maximum settlement that the road can withstand: ω 1 、ω 2 、ω 3 、ω 4 are weights; γ is a normalization factor;

[0050] Among them, the weights ω 1 、ω 2 、ω 3 、ω 4 are all positive numbers, and the sum of the weights ω 1 、ω 2 、ω 3 、ω 4 is 1, and the normalization factor γ is used to control to be between 0.1 and 1.9;

[0051] The acquisition cycle base T 0 is user-defined. When calculating the road image acquisition cycle T, the environmental parameters of the area where the road is located are the environmental parameters within the historical time threshold specified by the user;

[0052] f(h i -h 0 ) represents a constraint function. If h i -h 0 ≥0, then f(h i -h 0 ) = h i -h 0 , if h i -h 0 <0, then f(hi - h 0 ) = 0. The rainfall h 0 that can cause damage to the road is user-defined, and the constant x takes the cumulative number of times when f(h i -h 0 ) = 0;

[0053] The maximum number of freeze-thaw cycles p max that the road can withstand and the maximum settlement L max that the road can withstand are user-defined;

[0054] The value of the normalization factor γ follows that the higher the average daily traffic flow of the road history, the smaller the value of the normalization factor γ, and vice versa, the larger the value of the normalization factor γ;

[0055] Through the above logical formula calculation, the acquisition cycle applied in the highway image acquisition stage is designed to ensure that the highway image acquisition cycle can conform to the frequency characteristics of highway defect generation.

[0056] Retrieve highway images from the cloud database and identify the degree of difference between the highway images.

[0057] In the stage of retrieving highway images, any differentiated storage interval in the cloud database is used as the retrieval target to execute the retrieval operation. All stored highway images in the differentiated storage interval are retrieved at one time, and the marked content of all retrieved highway images is traversed synchronously. Based on the traversal result, two highway images with the earliest and latest acquisition times in the marked content are selected as the targets for identifying the degree of difference, and the degree of difference between the two highway images is identified:

[0058]

[0059] In the formula: DISS(a,b) is the degree of difference between the two highway images; m is the total number of contour points in the two highway images; d pos (a j ,b j ) is the Euclidean distance between the two highway images at the j-th contour point; α is the weight; u is the curvature sequence obtained after aligning the two highway images through the DTW algorithm; is the curvature value of the v-th sequence point of each of the two highway images;

[0060] Among them, before calculating the degree of difference between the two highway images, they are synchronously converted into contour images. The larger the value of DISS(a,b), the greater the degree of difference between the two highway images; conversely, it means the smaller the degree of difference between the two highway images. The weight α takes values in the range of 0 to 1;

[0061] Through the above logical formula, the calculation formula for the degree of difference of highway images is further defined, providing support for determining the defect location in highway images.

[0062] When the judgment result is no, all highway images stored in the differentiated storage interval where the highway image from which the recognition result is derived is located are synchronously deleted. The deletion target is all highway images other than the highway image with the earliest marked acquisition time;

[0063] In the stage of segmenting highway images, the segmentation operation of highway images is performed based on the highway image segmentation processing logic.

[0064] Among them, when the two highway images for which the segmentation process is performed are acquired, the acquisition distance and angle are exactly the same;

[0065] Set the highway defect determination interval, obtain the recognition result of the difference degree between highway images, compare the recognition result with the set highway defect determination interval, and determine whether the recognition result is within the highway defect determination interval;

[0066] If the determination result is no, end, wait for the highway images in the cloud database to be updated and stored, and after the update and storage, execute the determination operation again;

[0067] If the determination result is yes, obtain the two highway images from which the recognition result is derived, perform segmentation processing on the two highway images, and record the segmented images as sub-images. Compare the sub-images of the two highway images pairwise according to the corresponding positions, and determine the defective positions in the highway images based on the comparison results of the difference degrees of the sub-images;

[0068] Obtain the defective positions existing in the determined highway images, collect all the obtained defective positions, and generate a defective position message to feedback to the user terminal.

[0069] In this embodiment, through the execution of the above method, a brand-new, intelligent and comprehensive recognition method is provided for highway defect recognition. When there are defect problems on the highway, based on this method, it can be quickly recognized and located, assisting highway management users to carry out highway defect repair work more quickly, and ensuring the functionality and safety of the highway.

[0070] As Figure 1 shown, the segmentation processing logic of highway images is the setting of the number of segments of highway images, and the segmentation processing logic of highway images is expressed as:

[0071] Set the number of segments of highway images to 2*2, 3*3, 4*4,...;

[0072] Take the sub-images from the same positions in the two highway images as the comparison targets, apply the calculation formula of the difference degree between the two highway images to calculate the comparison targets, and simultaneously apply the highway defect determination interval for determination;

[0073] When the calculation results of the corresponding sub-images obtained by segmenting the two highway images meet the highway defect determination interval and there are determination results that do not meet the highway defect determination interval, end.

[0074] Through the above settings, the segmentation processing logic of highway images is further defined, providing support for further refinement of the defective positions on the highway in the above method.

[0075] It should be noted that when segmenting the two highway images, the number of segments is set to be the same, that is, both highway images are segmented into 2*2, 3*3, 4*4 or other numbers to facilitate pairwise comparison of sub-images according to the corresponding positions.

[0076] As Figure 1 shown, the position with defects in the road image is the position of the sub-image that conforms to the road defect determination interval in the road image;

[0077] After the position with defects in the road image is determined, real coordinates are configured for the road image synchronously. The position with defects in the road image combines the real coordinates configured for the road image to determine the position with defects in the corresponding area of the road image;

[0078] The content of the defect position message is the corresponding real coordinates of the defect position. After the defect position message is generated, it is transmitted to the cloud database synchronously, and the user terminal retrieves the defect position message from the cloud database.

[0079] Through the above settings, based on the configuration of the real coordinates of the road image, a further real positioning of the defect position in the road image is carried out.

[0080] In summary, during the execution of the method in the above embodiments, the road image is collected and stored by the drone device, so as to determine the position with defect problems on the road by comparing the degree of difference between the previous and subsequent road images. Synchronously, combined with image segmentation and further comparison of the degree of difference, the position with defect problems on the road is refined, so as to assist road management personnel to carry out road defect problem maintenance work more efficiently, effectively improving the efficiency and intelligent level of road defect recognition, ensuring that the road surface can be comprehensively monitored, and guaranteeing the functionality and safety of the road.

[0081] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements will not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A road defect recognition method based on target detection, characterized in that: include: Set highway image collection points, periodically collect highway images based on the UAV equipment at the set highway image collection points, create a cloud database, and use the cloud database to store the collected highway images; Retrieving highway images from a cloud database and identifying differences between highway images; Setting a road defect determination interval, obtaining recognition results of the degree of difference between road images, and comparing the recognition results with the set road defect determination interval to determine whether the recognition results are within the road defect determination interval; If the judgment result is no, the process ends and the cloud database waits for the highway image to be updated and stored. After the update and storage, the judgment operation is performed again. If the result of the determination is yes, two highway images from which the recognition result is derived are obtained, the two highway images are segmented, and the segmented images are recorded as sub-images. The sub-images of the two highway images are compared in pairs according to corresponding positions, and the position of the defect in the highway image is determined according to the result of the comparison of the sub-image differences; The defect locations existing in the determined highway image are obtained, and a defect location message is generated and fed back to the user end.

2. A road defect identification method based on target detection according to claim 1, characterized in that: The highway image acquisition points are customized by the user end. After the highway image is acquired, the highway image is synchronously marked based on the coordinates of the highway image acquisition points and the acquisition time. The cloud database is provided with the same number of differentiated storage intervals as the highway image acquisition points. The cloud database differentiates and stores the highway images acquired by the drone device based on the differentiated storage intervals, so that the coordinates of the highway image marks stored in each differentiated storage interval are the same. Among them, the collection cycle applied in the highway image collection stage is set based on the environmental parameters of the highway area where the highway image collection point is located. The environmental parameters include: temperature, rainfall, number of freeze-thaw cycles, and base settlement.

3. A method for identifying highway defects based on target detection according to claim 2, characterized in that: The acquisition cycle applied in the highway image acquisition phase is subject to: Where: T is the highway image acquisition cycle; T0 is the acquisition cycle base; C MAX , C MIN is the highest and lowest temperature in the area where the highway belongs; n is the number of rainfalls in the area where the highway belongs; x is a constant; h i is the i-th rainfall in the area where the highway belongs; h0 is the rainfall that can cause damage to the highway; p is the number of historical freeze-thaw cycles in the area where the highway belongs; p max is the maximum number of freeze-thaw cycles that the highway can withstand; L is the cumulative settlement of the highway in the area where the current highway belongs; L max is the limit settlement that the highway can bear: ω1, ω2, ω3, ω4 are weights; γ is the normalization factor; Among them, the weights ω1, ω2, ω3, and ω4 are all positive numbers, and the sum of the weights ω1, ω2, ω3, and ω4 is 1. The normalization factor γ is used to control Between 0.1 and 1.

9.

4. A method for identifying highway defects based on target detection according to claim 3, characterized in that: The acquisition cycle base T0 is customized by the user end. When calculating the highway image acquisition cycle T, the environmental parameters of the area to which the highway belongs are the environmental parameters within the historical time threshold specified by the user end. f(h i -h0) represents the constraint function, h i -h0≥0, then f(h i -h0)=h i -h0,h i -h0<0, then f(h i -h0)=0, the rainfall h0 that can cause damage to the road is defined by the user, and the constant x is taken as f(h i -h0)=0 cumulative times; The maximum number of freeze-thaw cycles that a highway can withstand p max and the ultimate settlement L that the highway can withstand max Customized by the user; The value of the normalization factor γ follows the following rule: the higher the historical average daily traffic volume of the highway, the smaller the value of the normalization factor γ; conversely, the larger the value of the normalization factor γ.

5. The method for identifying highway defects based on target detection according to claim 2, characterized in that: In the road image retrieval stage, any differentiated storage interval in the cloud database is used as the retrieval target to perform the retrieval operation, and all road images stored in the differentiated storage interval are retrieved at a time, and the marked contents of all retrieved road images are traversed synchronously. Based on the traversal results, the two road images with the earliest and most recent acquisition time in the marked contents are selected as the difference degree recognition targets to identify the difference degree between the two road images: Where: DISS(a,b) is the difference between the two highway images; m is the total number of contour points in the two highway images; d pos (a j ,b j ) is the Euclidean distance between the two highway images at the jth contour point; α is the weight; u is the curvature sequence obtained after the two highway images are aligned by the DTW algorithm; is the curvature value of the vth sequence point of each of the two highway images; Among them, the two highway images are synchronously converted into contour images before performing the difference calculation. The larger the DISS (a, b) value is, the greater the difference between the two highway images is. Conversely, the smaller the difference between the two highway images is, and the weight α is in the range of 0 to 1.

6. A method for identifying highway defects based on target detection according to claim 5, characterized in that: When the determination result is no, the highway images stored in the differentiated storage interval where the highway image from which the recognition result is derived is synchronously deleted, and the deletion target is all highway images except the highway image with the earliest acquisition time marked; In the stage of segmenting the highway image, the segmentation operation of the highway image is performed based on the highway image segmentation processing logic; The two highway images that are segmented are collected at exactly the same distance and angle.

7. A method for identifying highway defects based on target detection according to claim 6, characterized in that: The segmentation processing logic of the highway image is the setting of the number of segmentations of the highway image. The segmentation processing logic of the highway image is expressed as: Set the number of road image segments to 2*2, 3*3, 4*4, ...; The sub-images obtained from the segmentation and coming from the same position in the two highway images are used as the comparison targets, and the difference degree calculation formula between the two highway images is applied to calculate the comparison targets, and the highway defect determination interval is simultaneously applied to make the determination; When there are calculation results of the corresponding sub-images obtained by segmenting the two highway images that meet the highway defect determination interval and there are determination results that do not meet the highway defect determination interval, the process ends.

8. The method for identifying highway defects based on target detection according to claim 1, characterized in that: The position where the defect exists in the highway image is the position of the sub-image that meets the highway defect determination interval in the highway image; After the position of the defect in the highway image is determined, the real coordinates are configured for the highway image synchronously, and the position of the defect in the highway image is combined with the real coordinates configured for the highway image to determine the position of the defect in the corresponding area of ​​the highway image.

9. A road defect identification method based on target detection according to claim 1 or 8, characterized in that: The defect location message content is the corresponding real coordinates of the defect location. After being generated, the defect location message is synchronously transmitted to the cloud database, and the user end retrieves the defect location message from the cloud database.

Images