An image semantic feature-based power image data screening method

Abstract

The application relates to the technical field of image screening, and discloses a power image data screening method based on image semantic features, which comprises the following steps: step one, removing bad image data in collected images; step two, image data background removal processing; step three, selecting images containing the shapes according to the color and shape of the required screening target and storing the images into a data set; step four, selecting target pictures with defects or faults; and step five, completing screening and outputting image data. After the initial screening treatment of all images, the color semantic features of the target are enhanced and screened, then the shape semantic features of the target are enhanced and screened, single image data containing only the target is obtained, and then the single image of the normal target is compared, the images of the target with faults or defects are screened and output, so that the equipment fault position can be accurately judged and recognized at last.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image screening, in particular to a power image data screening method based on image semantic features. BACKGROUND

[0002] The engineering project site of the power construction company is far away from the urban area, the conditions are harsh, the communication means is backward, there is no company special local area network coverage, the telecom and mobile public network signal is poor, the station contact relies on the intercom, the station contact with the general office relies on the mobile phone, the digital management and office conditions are poor, with the development of intelligent inspection, the power image data screened out from the large number of inspection power images collected by aerial photography or robots has great effect on timely fault repair.

[0003] Because the image data photographed by the inspection equipment in each inspection process is more, and the types and quantities of the equipment existing in each photographed image are different, there are many interference factors, the artificial intelligence cannot be compared with the accurate automatic identification, and then the electrical equipment defects or faults cannot be accurately judged, finally the rapid maintenance work of the fault equipment is affected, and the timely recovery of the power is greatly affected. Therefore, we propose a power image data screening method based on image semantic features. SUMMARY

[0004] The purpose of the present application is to provide a power image data screening method based on image semantic features, after the preliminary screening treatment of all images, the color semantic features of the target are enhanced and screened, then the shape semantic features of the target are enhanced and screened, only the single image data containing the target is obtained, then compared with the single image of the normal target, the images of the target with faults or defects are screened out and output, which is convenient for the accurate judgment and identification of the equipment fault position, and solves the problems in the background art.

[0005] To achieve the above purpose, the present application provides the following technical scheme: a power image data screening method based on image semantic features, comprising the following steps:

[0006] Step 1: the images collected by the inspection equipment are preliminarily screened and treated, and the bad image data such as blurred, repeated, overexposed, underexposed and the like in the collected images are removed;

[0007] Step 2: the background of all image data is removed, such as the sky background, the ground or grassland, forest background and the like, only the substantive content in the image is retained;

[0008] Step three: the basic color or common color of the target to be screened is enhanced, the contrast of the color or multiple colors in the image is strengthened, the color or multiple colors are displayed more obviously, and the contents of other colors are deleted, and the pictures are selected, and the shapes are further screened according to the shape of the target to be screened, and the images containing the shapes are selected and stored in the data set;

[0009] Step four: the color enhancement part of the image screened and selected in the above step is restored to the initial state, the image data without other contents is obtained, and the restored image data is compared with the target image without failure, defect and other contents, and the target picture with defects or failures is selected;

[0010] Step five: finally, the image data is completely restored to the initial state, the screening is completed, and the image data is output.

[0011] As a preferred embodiment of the present application, the inspection device in step one can be a camera installed on a tower, a camera installed on a drone or other camera devices for shooting power equipment.

[0012] As a preferred embodiment of the present application, the color enhanced in step three is the fixed color of the target, or when the target has multiple colors, the colors are enhanced together.

[0013] As a preferred embodiment of the present application, the target shape screened in step three is the shape contained in any position of the image to be screened or the shape of the whole image.

[0014] As a preferred embodiment of the present application, the single target is compared with the normal single target in step four, and the content with defects can be accurately screened.

[0015] As a preferred embodiment of the present application, the image data finally output in step five needs to be checked by artificial inspection, and the failure problem of the power equipment is finally accurately judged.

[0016] As a preferred embodiment of the present application, the processing process and screening process in steps one to five are completed in an artificial intelligence system based on a computer.

[0017] Compared with the prior art, the present application has the following advantages:

[0018] The application obtains single image data containing only the target by preliminarily screening all images, enhancing and screening the color semantic features of the target, then enhancing and screening the shape semantic features of the target, comparing the single image data with normal target single images, screening out images of the target with faults or defects and outputting, so as to accurately judge and identify the equipment fault. BRIEF DESCRIPTION OF DRAWINGS

[0019] Other features, objects and advantages of the application will become more apparent from the following detailed description of non-limiting embodiments, read in conjunction with the accompanying drawings:

[0020] Figure 1 A flow chart of the power image data screening method based on image semantic features. DETAILED DESCRIPTION

[0021] In order to make the technical means, creative features, purposes and effects of the application easy to understand, the application is further described below in conjunction with specific embodiments.

[0022] Embodiment one

[0023] Please refer to Figure 1 The application provides a technical solution: a power image data screening method based on image semantic features, comprising the following steps:

[0024] Step one: image acquisition by inspection equipment, the inspection equipment can be a camera installed on a tower, a camera installed on a drone or other position camera device for shooting power equipment, the images collected by the inspection equipment are preliminarily screened, and bad image data such as blurred, repeated, overexposed, underexposed and the like are removed;

[0025] Step two: background removal processing of all image data, such as sky background, ground or grassland, forest background and the like, only the substantive content in the image is retained;

[0026] Step three: enhancement processing of the basic color or common color of the target to be screened, the contrast of the color or multiple colors in the image is enhanced, the enhanced display color is the fixed color of the target, or when the target has multiple colors, the colors are enhanced together to make them more obviously displayed and delete the content of other colors, and the images are selected, the shapes are further screened in the selected images, the shape of the target to be screened is screened, when any position in the image to be screened contains the shape or the whole shape is the shape to be screened, the images containing the shape are selected and stored in the data set;

[0027] Step four: restore the image color enhancement part selected by the above screening to the initial state, obtain the image data excluding other contents, and compare the restored image data with the target image without faults, defects and other contents, and select the target image with defects or faults;

[0028] After the above steps three and four, the screening and selection of one target are realized, then the content of the target is changed, and the screening and selection of another target are performed again, so that the power equipment targets or cables and the like required to be screened and obtained are screened finally, and corresponding image data is output, and the equipment faults can be judged.

[0029] Step five: finally, the image data is completely restored to the initial state, the screening is completed, and the image data is output, and the finally output image data needs to be checked manually, and finally the fault problems of the power equipment are accurately judged, and then the equipment repair and other maintenance work can be performed.

[0030] Further, in step four, the single target is compared with the normal single target, the image data of the normal target can be sampled when the power equipment is normally working, and the clearest image is stored in the artificial intelligence system, so that the image data can be called at any time, and the content with defects can be screened more accurately.

[0031] Further, the processing and screening processes in steps one to five are completed in the artificial intelligence system based on the computer.

[0032] Embodiment two

[0033] Please refer to Figure 1 The present application provides a technical solution: a power image data screening method based on image semantic features, comprising the following steps:

[0034] Step one: image collection is performed by using an inspection device, the inspection device can be a camera installed on a tower, a camera installed on a drone, or other camera devices for shooting power equipment, and the collected images of the inspection device are centrally processed for preliminary screening, and bad image data such as blurred, repeated, overexposed, underexposed and the like is removed;

[0035] Step two: all image data is processed for background removal, such as sky background, ground or grassland, forest background and the like, only the substantive content in the image is retained, and when the background is processed, the color of the background to be processed is first discussed, and the background image can be quickly cleaned up;

[0036] Step three: the target color is screened out, and other colors except the basic color or common color are weakened, reducing the contrast of other colors in the image, thereby increasing the contrast of the target color. The color displayed by the weakening process is other colors except the fixed color of the target, or when the target has multiple colors, other colors are weakened, making them weakened and deleted. Then select these pictures, and further screen the shapes according to the shape of the target to be screened. The target shape is the shape contained in any position of the image to be screened or the overall shape. Select the image containing these shapes and store it in the data set.

[0037] Step four: compare the image data selected by the above screening with the target image without defects, no defects and no other content, and select the target image with defects or faults.

[0038] After the above steps three and four, the screening and selection of one target are realized, then the content of the target is changed, and the screening and selection of another target are performed again. Repeat the operation, finally the power equipment target or cable and other targets that need to be screened and obtained can be screened, and the corresponding image data can be output. The device fault can be judged.

[0039] Step five: finally, the image data is completely restored to the initial state, the screening is completed, and the image data is output. The finally output image data also needs to be checked manually, and finally the fault problem of the power equipment is accurately judged. Then the repair and other maintenance work of the equipment can be carried out.

[0040] Further, in step four, a single target is compared with a normal single target. The image data of the normal target can be sampled when the power equipment is working normally, and the clearest image is stored in the artificial intelligence system for easy retrieval and use. The content with defects can be accurately screened out.

[0041] Further, the processing and screening processes in steps one to five are completed in the artificial intelligence system based on computer.

[0042] The foregoing merely illustrates the principles of the application and application of its leading features. This application is not limited to the illustrative embodiments shown and described herein. Rather, the scope of the present application is defined by the appended claims, and other embodiments of this application will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. Therefore, to the extent that there is disclosed herein an illustrative implementation that could be embodied in any of a number of varied forms, the intent is to support a claim scope that includes any such forms as fall within the scope of claims, and their equivalents. It is therefore required by the claims to refer to any such patentable design that comes within the scope of a claim.

[0043] Furthermore, it should be understood that although the description above relates to embodiments, not every embodiment according to the application contains all features needed to implement the invention. The description is provided for purposes of illustration only and merely implies one of various alternatives to the present invention. Those skilled in the art will recognize immediately that the description is illustrative only and is not intended to be limiting on the overall scope of the present application. Therefore, the scope of the present application is defined not by the detailed description of the embodiments but by the appended claims, and their equivalents.

Claims

1. A method for screening electric power image data based on image semantic features, characterized in that: The following steps are involved: Step 1: Perform preliminary screening on the images collected by the inspection equipment to remove blurry, repeated, overexposed, and underexposed image data; Step 2: All image data is processed for background removal, including the sky background, ground or grass, and forest background, retaining only the essential content of the image; Step 3: Enhance the basic colors or common colors of the target to be screened, increase the contrast of the color or multiple colors in the image, make them more obvious, and delete the content of other colors at the same time, and select these images. Among the selected images, further perform shape screening, screen the target shape as needed, select images containing these shapes and save them into the dataset; Step 4: Restore the color-enhanced portion of the image after the screening to its original state, thereby obtaining image data without other content. The restored image data is then compared with a target image without faults, defects, or other content, to select the target image with defects or faults. Step 5: Finally, the image data is completely restored to its original state, the screening is completed, and the image data is output.

2. The method for screening electric power image data based on image semantic features according to claim 1, characterized in that: The inspection equipment in step 1 may be a camera mounted on a tower, a camera mounted on a drone, or a camera device for photographing power equipment at other locations.

3. The method for screening electric power image data based on image semantic features according to claim 1, characterized in that: The enhanced color displayed in step 3 is a fixed color of the target, or when the target has multiple colors, these colors are enhanced together.

4. The method for screening electric power image data based on image semantic features according to claim 1, characterized in that: The target shape screened in step three is the content that needs to be screened when the shape is contained at any position in the image to be screened or the entire image is the shape.

5. The method for screening electric power image data based on image semantic features according to claim 1, characterized in that: In step 4, the single target is compared with the normal single target, so that defective content can be screened out more accurately.

6. The method for screening electric power image data based on image semantic features according to claim 1, characterized in that: The image data finally output in step 5 needs to be manually checked to accurately determine the fault problem of the power equipment.

7. The method for screening electric power image data based on image semantic features according to claim 1, characterized in that: The processing and screening processes in steps 1 to 5 are all completed in a computer-based artificial intelligence system.

Images