An image enhancement method for security equipment detection of workers

Abstract

The present application belongs to the field of image enhancement, and particularly relates to an image enhancement method for security equipment detection of staff. The method first collects images containing security equipment and pre-processes; then fuses regional brightness features and multi-modal data to construct an input vector, and realizes improved dynamic brightness compensation with the aid of neural network and reinforcement learning; then improves adaptive frequency domain multi-scale CLAHE, covering adaptive local window, dynamic histogram clipping and multi-scale fused CLAHE; finally detects security equipment by using the enhanced image. The present application can avoid over-enhancement or information loss of traditional methods, precisely enhance details of different regions, optimize overall and local contrast of the image, improve the recognizability of security equipment in complex environment, and can be adaptively optimized during detection, with stable and reliable enhancement effect, effectively solving the security equipment detection problem in actual scene.

Description

Technical Field

[0001] This invention belongs to the field of image enhancement, and in particular relates to an image enhancement method for detecting security equipment used by workers. Background Technology

[0002] In today's security work, personnel rely heavily on security equipment, making it crucial to accurately and promptly detect whether they are correctly wearing and using it. However, image acquisition in real-world scenarios faces numerous challenges, resulting in inconsistent image quality and significantly complicating security equipment detection. Light intensity and angle vary considerably under different working environments. Outdoors, under strong direct sunlight, images are prone to overexposure, causing the loss of some details on the security equipment; while in dimly lit indoor or nighttime scenes, images are too dark to clearly distinguish equipment features. This uncertainty in lighting significantly reduces the accuracy of image-based security equipment detection. Work environments are often filled with various interfering factors. The background may contain objects similar in color and shape to the security equipment, or obstructions may partially or completely block it. Furthermore, the diversity of clothing styles, postures, and movements among different personnel increases image complexity, further complicating the accurate detection of security equipment. Existing image enhancement methods have limitations. For example, simple histogram equalization, while improving overall brightness and contrast to some extent, can easily lead to over-enhancement or loss of image details, resulting in image distortion. Some brightness compensation methods based on global adjustments cannot make targeted adjustments according to the characteristics of different regions in the image, making it difficult to meet the needs of accurate detection of security equipment under complex backgrounds and lighting conditions. Traditional frequency domain processing methods lack adaptability in window selection and contrast adjustment, and cannot effectively cope with the differences between different images. Summary of the Invention

[0003] In view of the technical problems existing in the background art, the present invention proposes an image enhancement method for detecting security equipment for workers.

[0004] To achieve the above objectives, the technical solution adopted by the present invention includes the following steps:

[0005] S1. First, collect image information containing staff security equipment and preprocess the images;

[0006] S2. Next, the regional brightness features and multimodal data are fused to construct the input vector, and the improved dynamic adjustment of image brightness compensation is realized.

[0007] S3. Improved adaptive frequency domain multi-scale CLAHE, including improved adaptive local windowing for frequency domain energy distribution, improved dynamic histogram clipping for local contrast perception, and improved multi-scale fusion CLAHE.

[0008] The improved adaptive local window for frequency domain energy distribution adaptively selects different local window sizes by analyzing the energy distribution of different regions of the image in the frequency domain. First, it calculates the frequency domain energy distribution within the local region: ,in Is the image in Fourier transform of position It is a high-pass filter. Representing local high-frequency energy, and then performing adaptive window selection: ,in This is the maximum window size. It is the minimum window size. It is the highest high-frequency energy in the entire image;

[0009] The improved local contrast perception dynamic histogram cropping involves calculating local contrast and dynamically adjusting the histogram cropping threshold for adaptive cropping. The adjustment of the adaptive cropping threshold is as follows: ,in These are the maximum and minimum cutting limits, respectively. For local contrast;

[0010] The improved multi-scale CLAHE fusion method enhances the image using three different window scales (small, medium, and large) and then performs weighted fusion. First, the image is equalized. ,in Indicates window sizes at different scales. This represents three different scales: small, medium, and large. Then, the final enhanced image is fused using adaptive fusion coefficients. ,in For adaptive fusion coefficients;

[0011] S4. Finally, the enhanced images are used to inspect the staff's security equipment to ensure accurate identification under different lighting conditions and complex environments.

[0012] Preferably, the construction of the input vector in step S2 is based on obtaining the average brightness through regional brightness analysis. and brightness standard deviation Add image color feature vectors 'm' represents the dimension of the color feature vector, and the depth feature vector is... The overall image contrast is Then the input vector X of the neural network is: , where N represents the different regions divided.

[0013] As a preferred option, average brightness ,in Indicates the region The number of pixels, the standard deviation of brightness .

[0014] Preferably, the dynamic adjustment of image brightness compensation in step S2 is achieved by processing the obtained input vector X through a neural network to output a brightness compensation parameter vector for each region. The j-th output of the neural network Represented as: ,in These are the weights connecting the i-th neuron in the input layer and the j-th neuron in the output layer. It is the bias of the j-th neuron in the output layer; where, when reinforcement learning is introduced, the weights and biases are dynamically adjusted according to the reward feedback, and the weights and biases are updated using the stochastic gradient method. When the security equipment in the image undergoes dynamic changes that trigger the re-division of the region, the features of each region are recalculated and the compensation parameters are adjusted. The compensation parameters of the new region are obtained by the updated neural network based on the newly updated input vector. The compensated image is... ,in For the corresponding new compensation parameters, For the new area.

[0015] Preferably, the adaptive fusion coefficient in step S3 The specific form is: ,in Represents global contrast at different scales.

[0016] Compared with existing technologies, the advantages and positive effects of this invention are as follows: It constructs an input vector using regional brightness features and multimodal data, calculates brightness compensation parameters using a neural network, and continuously optimizes the image through reinforcement learning. This ensures that the image maintains optimal brightness under different lighting conditions, avoiding over-enhancement or information loss caused by traditional global adjustment methods. By analyzing the frequency domain energy distribution of local image regions, it adaptively selects the local window size, enabling targeted enhancement processing for different regions, improving detail representation, and effectively reducing noise impact. Based on local contrast characteristics, it adaptively adjusts the histogram cropping threshold to ensure more accurate contrast enhancement in different regions, while avoiding loss of detail information due to over-equalization, thus improving image clarity. By combining CLAHE enhancement results with small, medium, and large window sizes and using adaptive fusion coefficients for weighting, both overall image detail and local contrast are optimized, improving the recognizability of security equipment in complex backgrounds. Reinforcement learning is introduced during the detection process. When security equipment in the image adjusts due to changes in posture, lighting, or environment, the system dynamically updates the neural network weights, achieving adaptive optimization and ensuring stable and reliable enhancement effects over long-term operation. Detailed Implementation

[0017] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described below with reference to embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0018] Numerous specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways than those described herein, and therefore the invention is not limited to the specific embodiments disclosed in the following specification.

[0019] In this example, in today's security work, workers rely heavily on security equipment, making it crucial to accurately and promptly detect whether they are correctly wearing and using this equipment. However, image acquisition in real-world scenarios faces numerous challenges, resulting in inconsistent image quality and posing significant difficulties for security equipment inspection. Therefore, an image enhancement method for inspecting security equipment worn by workers is proposed.

[0020] First, to ensure image quality, image information containing staff and security equipment is collected and preprocessed. Preprocessing steps include noise reduction, normalization, and image alignment to ensure consistent quality and format for subsequent image processing.

[0021] To address the issue of uneven image brightness under varying lighting conditions, existing dynamic brightness compensation methods primarily rely on global adjustment strategies, such as histogram equalization, gamma correction, or simple linear transformations. These methods have significant drawbacks when processing images with uneven brightness, mainly manifested in their inability to adaptively adjust local areas, leading to over-enhancement or information loss in certain regions. Furthermore, they fail to incorporate image content information, relying solely on fixed rules for adjustment, making them ill-suited for complex environments such as scenes with strong light, shadows, or multiple light sources. However, this strategy improves upon this by fusing regional brightness features with multimodal data and utilizing neural networks for adaptive optimization of brightness compensation parameters. Therefore, the input vector is constructed by fusing regional brightness features and multimodal data to achieve improved dynamic image brightness compensation. The input vector is constructed by obtaining the average brightness based on regional brightness analysis. and brightness standard deviation Add image color feature vectors 'm' represents the dimension of the color feature vector, and the depth feature vector is... The overall image contrast is Then the input vector X of the neural network is: , where N represents the different regions divided. The average brightness... ,in Indicates the region The number of pixels, the standard deviation of brightness .

[0022] Furthermore, the dynamic adjustment of image brightness compensation is achieved by processing the obtained input vector X through a neural network and outputting a brightness compensation parameter vector for each region. The j-th output of the neural network Represented as: ,in These are the weights connecting the i-th neuron in the input layer and the j-th neuron in the output layer. It is the bias of the j-th neuron in the output layer; where, when reinforcement learning is introduced, the weights and biases are dynamically adjusted according to the reward feedback, and the weights and biases are updated using the stochastic gradient method. When the security equipment in the image undergoes dynamic changes that trigger the re-division of the region, the features of each region are recalculated and the compensation parameters are adjusted. The compensation parameters of the new region are obtained by the updated neural network based on the newly updated input vector. The compensated image is... ,in For the corresponding new compensation parameters, This is a new region. The advantage of this improvement is that it can adaptively adjust to the brightness differences in different regions, thereby achieving balanced illumination while avoiding over-enhancement and information loss. Furthermore, it can continuously optimize compensation parameters in dynamically changing environments, enabling the system to maintain efficient brightness adjustment capabilities even when facing changes in lighting conditions.

[0023] Next, considering that the existing CLAHE method, with its fixed local window, static contrast cropping threshold, and single-scale processing, performs poorly under complex lighting conditions and cannot adaptively adjust to the characteristics of different regions of the image, it may over-enhance noise in high-frequency detail-rich areas while providing insufficient enhancement in low-frequency areas. This invention improves the adaptive local window for frequency domain energy distribution, the dynamic histogram cropping for improved local contrast perception, and the multi-scale fusion CLAHE. The improved adaptive local window for frequency domain energy distribution adaptively selects different local window sizes by analyzing the energy distribution of different regions of the image in the frequency domain. First, it calculates the frequency domain energy distribution within the local region: ,in Is the image in Fourier transform of position It is a high-pass filter. Representing local high-frequency energy, and then performing adaptive window selection: ,in This is the maximum window size. It is the minimum window size. It represents the highest high-frequency energy across the entire image. The improved local contrast perception dynamic histogram cropping involves calculating local contrast and dynamically adjusting the histogram cropping threshold for adaptive cropping. The adjustment of the adaptive cropping threshold is as follows: ,in These are the maximum and minimum cutting limits, respectively. This refers to local contrast. The improved multi-scale CLAHE fusion method enhances CLAHE using three different window scales (small, medium, and large) and then performs weighted fusion. First, the image is equalized. ,in Indicates window sizes at different scales. This represents three different scales: small, medium, and large. Then, the final enhanced image is fused using adaptive fusion coefficients. ,in For adaptive fusion coefficients, ,in Represents global contrast at different scales.

[0024] Finally, the enhanced images are used to inspect the staff's security equipment to ensure accurate identification under different lighting conditions and complex environments.

[0025] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments for application in other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An image enhancement method for detecting security equipment used by staff, characterized in that, Includes the following steps: S1. First, collect image information containing staff security equipment and preprocess the images; S2. Next, the regional brightness features and multimodal data are fused to construct the input vector, and the improved dynamic adjustment of image brightness compensation is realized. S3. Improved adaptive frequency domain multi-scale CLAHE, including improved adaptive local windowing for frequency domain energy distribution, improved dynamic histogram clipping for local contrast perception, and improved multi-scale fusion CLAHE. The improved adaptive local window for frequency domain energy distribution adaptively selects different local window sizes by analyzing the energy distribution of different regions of the image in the frequency domain. First, it calculates the frequency domain energy distribution within the local region: ,in Is the image in Fourier transform of position It is a high-pass filter. Representing local high-frequency energy, and then performing adaptive window selection: ,in This is the maximum window size. It is the minimum window size. It is the highest high-frequency energy in the entire image; The improved local contrast perception dynamic histogram cropping involves calculating local contrast and dynamically adjusting the histogram cropping threshold for adaptive cropping. The adjustment of the adaptive cropping threshold is as follows: ,in These are the maximum and minimum cutting limits, respectively. For local contrast, For overall image contrast; The improved multi-scale CLAHE fusion method enhances the image using three different window scales (small, medium, and large) and then performs weighted fusion. First, the image is equalized. ,in Indicates window sizes at different scales. This represents three different scales: small, medium, and large. Then, the final enhanced image is fused using adaptive fusion coefficients. ,in For adaptive fusion coefficients; S4. Finally, the enhanced images are used to inspect the staff's security equipment to ensure accurate identification under different lighting conditions and complex environments.

2. The image enhancement method for detecting security equipment for workers according to claim 1, characterized in that, The construction of the input vector in step S2 is based on obtaining the average brightness from the regional brightness analysis. and brightness standard deviation Add image color feature vectors 'm' represents the dimension of the color feature vector, and the depth feature vector is... The overall image contrast is Then the input vector X of the neural network is: , where N represents the different regions divided.

3. The image enhancement method for detecting security equipment for workers according to claim 1, characterized in that, The dynamic adjustment of image brightness compensation in step S2 is achieved by processing the obtained input vector X through a neural network and outputting the output as a brightness compensation parameter vector for each region. The j-th output of the neural network Represented as: ,in These are the weights connecting the i-th neuron in the input layer and the j-th neuron in the output layer. It is the bias of the j-th neuron in the output layer; where, when reinforcement learning is introduced, the weights and biases are dynamically adjusted according to the reward feedback, and the weights and biases are updated using the stochastic gradient method. When the security equipment in the image undergoes dynamic changes that trigger the re-division of the region, the features of each region are recalculated and the compensation parameters are adjusted. The compensation parameters of the new region are obtained by the updated neural network based on the newly updated input vector. The compensated image is... ,in For the corresponding new compensation parameters, For the new area.

4. The image enhancement method for detecting security equipment for workers according to claim 1, characterized in that, The adaptive fusion coefficient in step S3 The specific form is: ,in Represents global contrast at different scales.

Images