Method for judging and optimizing image quality

Abstract

The invention discloses a method for judging and optimizing image quality, and relates to the technical field of image processing. The method comprises the following steps: S1, carrying out target detection on an input image by utilizing a target detection model, and identifying a key target area in the image; s2, extracting image data of the key target area according to a target detection result; s3, carrying out histogram analysis on the key target area, carrying out statistics on gray level pixel distribution, and judging the exposure state of the image based on the ratio of high-brightness pixels to low-brightness pixels; and S4, selecting a corresponding image optimization algorithm for processing according to the exposure state. According to the method, the key target in the image can be accurately extracted, the image quality is judged according to the target detection result, the definition and detail integrity of the key target in the image are ensured, and fine exposure adjustment is carried out on the target area in combination with homomorphic filtering and an adaptive histogram equalization technology.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to an image quality judgment optimization method. BACKGROUND

[0002] With the rapid development of unmanned aerial vehicle (UAV) inspection in the power industry, the quality requirements of the captured images for the inspection work are increasingly improved. The UAV inspection has become an important tool for the operation and management of the power industry due to its efficiency and safety. However, in practical applications, due to the influence of natural factors such as weather and light, the images captured by the UAV during autonomous inspection often have problems of overexposure or underexposure, resulting in loss of image details, decline of the quality of captured pictures, and even a large number of waste pictures. This not only affects the accuracy and efficiency of the inspection work, but also increases the difficulty of post-processing and data analysis of the images.

[0003] At present, most of the algorithms for image quality judgment and exposure adjustment are based on global analysis of the entire image. However, in the power industry inspection work, the focus is often on specific targets in the image, such as towers and equipment. Therefore, if a global adjustment method is used, although the exposure of the entire image may reach a relatively appropriate level, the exposure effect of the key targets concerned by the inspection work may not be ideal. This global adjustment method often cannot take into account the exposure needs of each part of the image, especially for the key targets, which may lose key details due to improper exposure, thereby affecting the accuracy and reliability of the inspection results.

[0004] Therefore, the existing image quality judgment and exposure adjustment algorithms have obvious deficiencies in the application of UAV inspection in the power industry, and a more accurate and efficient image quality optimization method is needed to meet the actual needs of the inspection work. SUMMARY

[0005] The present application aims to solve at least one of the technical problems existing in the prior art. To this end, one object of the present application is to provide an image quality judgment optimization method which can accurately extract key targets in an image and judge the image quality according to the target detection results, ensure the clarity and detail integrity of the key targets in the image, and combine the homomorphic filtering and adaptive histogram equalization techniques to finely adjust the exposure of the target regions, ensuring the clarity and detail integrity of the target regions.

[0006] In order to solve the above problems, the present application provides an image quality judgment optimization method, comprising the following steps: S1, using a target detection model to detect the input image, identifying the key target regions in the image, the key target regions including towers or equipment in power inspection; S2. Based on the target detection results, extract image data of key target regions; S3. Perform histogram analysis on key target areas, statistically analyze the grayscale pixel distribution, and determine the exposure status of the image based on the proportion of high-brightness pixels and low-brightness pixels, where the exposure status includes overexposure, underexposure, or normal. S4. Based on the exposure status, select the corresponding image optimization algorithm for processing: if it is determined to be overexposed, use the homomorphic filtering algorithm to adjust the image brightness; if it is determined to be underexposed, use the adaptive histogram equalization algorithm to adjust the image brightness; if it is determined to be normal, output the image directly.

[0007] Preferably, the target detection model in S1 is built based on deep learning technology and trained using a labeled power inspection image dataset. Data augmentation technology is used during training to improve the model's generalization ability.

[0008] Preferably, the specific method for determining the exposure state in S3 is as follows: If the proportion of high-brightness pixels in the key target area exceeds 30% of the total and the proportion of low-brightness pixels is less than 10%, it is judged as an overexposed image. If the number of low-brightness pixels in the key target area exceeds 30% of the total and the number of high-brightness pixels is less than 10%, it is judged as an underexposed image. If the above conditions are not met, the image is considered normal.

[0009] Preferably, in step S4, for overexposed images, a homomorphic filtering algorithm is used to adjust the brightness of the original image. The specific calculation formula is as follows: ; in, For pixels within the original image, For the pixels in the filtered image, This is a contrast control parameter, typically set to a value of 1. This is the gamma correction parameter used to adjust brightness, with a value range of [0.5, 2]. Average brightness of the image; For underexposed images, adaptive histogram equalization is used to adjust the brightness of the original image. First, the image is divided into multiple 16×16 blocks, the histogram of each block is calculated, histogram equalization is performed, and then bilinear interpolation is used to merge the processed blocks into a complete image. The calculation of histogram equalization can be expressed as the following formula: ; For normal images, no adjustments are needed; the image can be output directly.

[0010] Preferably, the adaptive histogram equalization algorithm in S4 includes: dividing the image into multiple small blocks, performing histogram equalization on each small block, and merging them into a complete image using a bilinear interpolation method.

[0011] The advantages of this invention compared to the prior art are: This invention identifies key target regions in an image using a target detection model, such as power poles and transformers in power line inspections, and determines the image's exposure status based on histogram analysis of these target regions. Compared to traditional global exposure adjustment methods, this invention can more accurately identify and focus on key areas in an image, thereby enabling targeted exposure adjustments for these areas. This improves the visibility and clarity of key equipment in the image and reduces background interference, allowing inspection personnel to more quickly identify potential problems or anomalies. This invention addresses the intelligent algorithm selection for images with varying exposure levels. Traditional image processing methods often employ a single global algorithm, which struggles to handle diverse and complex image scenarios. This invention, however, precisely determines the image's exposure state and selects the most suitable optimization algorithm for processing. This makes key information in the image more prominent and clearer, effectively avoiding the over-processing or under-processing problems that traditional methods may cause. It improves the optimization algorithm's effectiveness and the quality of the adjusted image, providing inspection personnel with more reliable and accurate image data support, and offering new ideas and methods for solving similar problems. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart of the image quality judgment optimization algorithm of the present invention. Detailed Implementation

[0014] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0015] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "joining" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0016] The present invention will now be described in further detail with reference to the accompanying drawings.

[0017] This invention provides an image quality assessment and optimization method, offering a method for extracting targets within pole images and assessing their quality based on a target detection model. The specific steps are as follows: 1. Utilize deep learning technology to construct an object detection model; 2. Collect a large number of labeled power industry inspection images as training datasets. The datasets should include key targets such as power poles and equipment under various weather and lighting conditions. 3. Train the model through iterative optimization algorithms to enable it to accurately identify key targets in images; 4. During training, data augmentation techniques are used to increase data diversity and improve the model's generalization ability; 5. Input the inspection image to be detected into the trained target detection model; 6. The model outputs the target's location (i.e., bounding box coordinates) and category information (tower or equipment type); 7. Based on the location information output by the model, extract the key target regions from the image: 8. Check whether the target detection model has successfully detected the key target. If the model cannot detect the target (e.g., the output is empty), the image quality is directly determined to be unqualified.

[0018] In addition, the image quality determination and optimization method of the present invention also provides an image exposure optimization method based on a target region image, the specific steps of which are as follows: 1. Based on the target detection results of the above-mentioned invention points, determine the key target regions in the image; 2. Divide the pixel value range of the target area (usually 0-255) into several gray levels (e.g., 256 levels), count the number of pixels in each gray level, and obtain a histogram. ,in Grayscale; 3. Set the high-brightness pixel threshold Gray levels greater than or equal to Pixels are defined as high-brightness pixels. Typically, It can be set to 200 or higher, depending on the specific image. 4. Set low-brightness pixel threshold Gray levels less than or equal to The area is defined as low-brightness pixels. Typically, It can be set to 50 or lower, depending on the specific image. 5. Determine the image exposure based on the histogram of the pixels in the target area. If the proportion of high-brightness pixels in the area exceeds 30% of the total and the proportion of low-brightness pixels is less than 10%, the image is considered overexposed. If the proportion of low-brightness pixels in the area exceeds 30% of the total and the proportion of high-brightness pixels is less than 10%, the image is considered underexposed. If the image does not meet either of these two conditions, the image quality is considered normal. 6. For overexposed images, a homomorphic filtering algorithm is used to adjust the brightness of the original image. The specific calculation formula is as follows: ; in, For pixels within the original image, For the pixels in the filtered image, This is a contrast control parameter (usually set to 1). This is the gamma correction parameter (used to adjust brightness, with a value range of [0.5, 2]). This represents the average brightness of the image (the average value of all pixel values ​​in the image). 7. For underexposed images, adaptive histogram equalization is used to adjust the brightness of the original image. First, the image is divided into multiple 16×16 blocks, the histogram of each block is calculated, histogram equalization is performed, and then bilinear interpolation is used to merge the processed blocks into a complete image. The calculation of histogram equalization can be expressed as the following formula: ; For normal images, no adjustments are needed; the image can be output directly.

[0019] To more clearly illustrate the specific embodiments of the present invention, an example is provided below: This method achieves precise adjustment and optimization of image exposure through steps such as target detection, histogram analysis, exposure state judgment, and intelligent algorithm selection, thereby improving image quality and inspection efficiency. The specific steps are as follows: 1. Use target detection models to detect targets in UAV inspection images and identify key target areas in the images, such as power equipment like power poles and transformers; 2. Based on the target detection results of the model, obtain the image within the target area; 3. Calculate the histogram of the target area, count the number of pixels at each gray level, and obtain the gray-level histogram of the target area; 4. Analyze the image exposure status based on the grayscale histogram of the target area, including the pixel ratio of high-brightness areas and low-brightness areas; 5. Based on the pixel ratio of high-brightness areas and low-brightness areas, determine whether the image is overexposed or underexposed. If the proportion of high-brightness areas is too high and the proportion of low-brightness areas is too low, it is judged as an overexposed image. If the proportion of low-brightness areas is too high and the proportion of high-brightness areas is too low, it is judged as an underexposed image. 6. Select the appropriate optimization algorithm for processing based on the image's exposure status; 6.1 For overexposed images, homomorphic filtering is used to eliminate uneven lighting and enhance image details. During homomorphic filtering, the contrast and brightness parameters are adjusted to make key information in the image stand out more. 6.2 For underexposed images, adaptive histogram equalization is used to improve the contrast and sharpness of the image. In the process of adaptive histogram equalization, the image is divided into multiple small blocks, histogram equalization is performed on each small block, and the processed small blocks are merged into a complete image by interpolation. 6.3 For normal images, the original image is output directly.

[0020] Finally, any aspects not fully described in this invention utilize existing mature products and technologies.

[0021] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A method for image quality assessment and optimization, characterized in that, Includes the following steps: S1. Use a target detection model to perform target detection on the input image and identify key target regions in the image. The key target regions include power poles or equipment in power inspection. S2. Based on the target detection results, extract the image data of the key target region; S3. Perform histogram analysis on the key target area, count the gray level pixel distribution, and determine the exposure status of the image based on the proportion of high brightness pixels and low brightness pixels, wherein the exposure status includes overexposure, underexposure or normal. S4. Based on the exposure state, select the corresponding image optimization algorithm for processing: if it is determined to be overexposed, use the homomorphic filtering algorithm to adjust the image brightness; if it is determined to be underexposed, use the adaptive histogram equalization algorithm to adjust the image brightness; if it is determined to be normal, output the image directly.

2. The method for image quality determination and optimization according to claim 1, characterized in that: The target detection model in S1 is built based on deep learning technology and trained using a labeled power inspection image dataset. Data augmentation technology is used during training to improve the model's generalization ability.

3. The method for image quality determination and optimization according to claim 1, characterized in that: The specific method for determining the exposure state in S3 is as follows: If the proportion of high-brightness pixels in the key target area exceeds 30% of the total and the proportion of low-brightness pixels is less than 10%, it is judged as an overexposed image. If the number of low-brightness pixels in the key target area exceeds 30% of the total and the number of high-brightness pixels is less than 10%, it is judged as an underexposed image. If the above conditions are not met, the image is considered normal.

4. The method for image quality determination and optimization according to claim 3, characterized in that: In S4 For overexposed images, a homomorphic filtering algorithm is used to adjust the brightness of the original image. The specific calculation formula is as follows: ； in, For pixels within the original image, For the pixels in the filtered image, This is a contrast control parameter, typically set to a value of 1. This is the gamma correction parameter used to adjust brightness, with a value range of [0.5, 2]. Average brightness of the image; For underexposed images, adaptive histogram equalization is used to adjust the brightness of the original image. First, the image is divided into multiple 16×16 blocks, the histogram of each block is calculated, histogram equalization is performed, and then bilinear interpolation is used to merge the processed blocks into a complete image. The calculation of histogram equalization can be expressed as the following formula: ； For normal images, no adjustments are needed; the image can be output directly.

5. The method for image quality determination and optimization according to claim 1, characterized in that: The adaptive histogram equalization algorithm in S4 includes: dividing the image into multiple small blocks, performing histogram equalization on each small block, and merging them into a complete image using bilinear interpolation.

Images