Multifunctional police service terminal intelligent helmet

By integrating multi-view cameras and double-layer generative adversarial networks in police helmets, combining AR technology and thermal design, the shortcomings of existing equipment in information integration and comfort are solved, and the on-site judgment and response capabilities of law enforcement personnel are improved. It realizes efficient information display and equipment stability.

CN119073694APending Publication Date: 2024-12-06GUANGXI POLICE ACAD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411208971.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-30
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

Existing police helmet equipment has shortcomings in real-time, accuracy and multi-dimensional information integration, which is difficult to effectively integrate with augmented reality technology, and it is difficult to provide clear and accurate pictures in low-light environments and long-term wear, which affects the impact of Law enforcement officers’ judgment and reaction ability, while not fully considering comfort and heat dissipation issues.

Method used

A multi-functional police terminal smart helmet was designed, integrating a multi-view camera module and a double-layer generative adversarial network, and displaying multi-view fusion images in real time through the AR smart glasses module, combining convolutional neural network for feature extraction and image processing, and optimizing images Denoising, super-resolution reconstruction and low-light enhancement, as well as wearing comfort through thermal dissipation components.

Benefits of technology

It improves the on-site perception ability and command decision-making efficiency of law enforcement personnel, improves the real-time and accuracy of information display, enhances the spatial perception ability, and ensures the stability and comfort of the equipment in complex environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119073694A_ABST
    Figure CN119073694A_ABST
Patent Text Reader

Abstract

A multifunctional police service terminal intelligent helmet disclosed by the present invention comprises a device body, the device body comprises a helmet main body, the helmet main body is internally provided with an identification assembly, the identification assembly is arranged on the two sides of the inner wall of the helmet main body through two mounting shafts, the mounting shafts are hinged to a bracket, and the bracket is connected with the device body. A mounting shaft is fixedly mounted on the outer side of the helmet body, AR intelligent glasses are fixedly mounted at the end, away from the mounting shaft, of the support, a camera is fixedly mounted on the outer side of the helmet body, a battery bin is fixedly mounted on the outer side of the helmet body, a light supplementing assembly is fixedly mounted on the outer side of the helmet body outside the camera, and a heat dissipation assembly is arranged in the helmet body. According to the invention, the comprehensive judgment capability of on-site conditions and the commanding and decision-making efficiency are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of police helmets, and in particular to a multifunctional police terminal intelligent helmet. Background Art

[0002] In today's society, with the continuous advancement of technology and the increasing complexity of police needs, law enforcement officers have put forward higher requirements for the acquisition and processing of real-time information when performing their tasks. The rapid development of augmented reality technology, generative adversarial networks and multi-view fusion technology has made the application of technology in the field of police work possible, especially in the development of intelligent police terminal equipment. Smart helmets with real-time information transmission and multi-view synchronous display capabilities have gradually become the focus of research and development.

[0003] Although existing police helmet equipment has realized video surveillance and communication functions to some extent, it still has many shortcomings in terms of real-time information, accuracy and multi-dimensional information integration. At the same time, due to the single function of existing police helmet equipment, it can usually only provide limited visual information and is difficult to effectively combine with AR technology, resulting in law enforcement officers being unable to obtain enhanced spatial perception information in real time during on-site operations. The limitations are particularly prominent in complex and changeable law enforcement environments, especially in low-light environments or other visually restricted conditions. Existing equipment is difficult to provide clear and accurate images, further affecting the judgment and reaction capabilities of law enforcement officers. Most of the existing police helmet designs fail to fully consider the comfort of long-term wearing and the heat dissipation of electronic equipment. In actual applications, it will cause the equipment to overheat or be uncomfortable to wear, thereby affecting the continuous operation capabilities of law enforcement officers.

[0004] In view of the above-mentioned defects of the existing technology, there is an urgent need for a multifunctional police terminal smart helmet to enhance the on-site perception ability of law enforcement personnel, and the design should fully consider the comfort of long-term wearing and the heat dissipation performance of the equipment. Summary of the invention

[0005] One purpose of the present invention is to provide a multifunctional police terminal smart helmet, which improves the comprehensive judgment ability of on-site situations and the efficiency of command decision-making.

[0006] A multifunctional police terminal smart helmet according to an embodiment of the present invention comprises a device body (100), wherein the device body (100) comprises a helmet body (110), wherein an identification component (200) is arranged inside the helmet body (110), wherein the identification component (200) is arranged on both sides of the inner wall of the helmet body (110) via two mounting shafts (210), wherein a bracket (211) is hingedly connected to the mounting shaft (210), wherein an AR smart glasses (212) are fixedly mounted on the end of the bracket (211) away from the mounting shaft (210), wherein a camera (217) is fixedly mounted on the outer side of the helmet body (110), wherein a battery compartment (216) is fixedly mounted on the outer side of the helmet body (110), wherein a fill light component is fixedly mounted on the outer side of the helmet body (110) outside the camera (217), and wherein a heat dissipation component (300) is arranged inside the helmet body (110).

[0007] Optionally, a cleaning strip (215) that fits the AR smart glasses (212) is installed on one side of the inner wall of the helmet body (110), a slide rail (213) is fixedly installed on the inner wall of the helmet body (110), a slider is slidably installed inside the slide rail (213), a connecting piece (214) connected to the AR smart glasses (212) is fixedly installed on the outer side of the slider inside the slide rail (213), and the heat dissipation component (300) includes rubber columns (312) equidistantly installed on the inner wall of the helmet body (110), a heat dissipation frame (310) is fixedly installed at the bottom end of the rubber column (312), the heat dissipation frame (310) is evenly penetrated with through holes, and sponge strips (311) are equidistantly installed on the inner surface of the heat dissipation frame (310).

[0008] Optionally, a central processing unit is embedded in the recognition component (200), and an AR smart glasses module, a multi-view camera module and a double-layer generative adversarial network module are integrated in the central processing unit. The recognition method of the recognition component (200) includes:

[0009] S1. Integrate an AR smart glasses module into a police helmet, overlay display images in the wearer's field of view, and obtain the wearer's first-person video footage in real time;

[0010] S2, installing a multi-view camera module on a police helmet to capture a multi-angle video image set of the police officer's surrounding environment;

[0011] S3, inputting the first-person video image and the multi-angle video image set into the built-in double-layer generative adversarial network of the helmet for processing, wherein the first generator of the first-layer generative adversarial network generates a denoised, super-resolution reconstructed and low-light enhanced image from the multi-angle video image set;

[0012] S4, a first discriminator of the first layer of the generative adversarial network distinguishes the difference between the enhanced multi-angle video image set generated by the first generator and the original multi-angle video image set, and optimizes the realism of the generated multi-angle video image set;

[0013] S5, a second generator of the second-layer generative adversarial network, fusing the multi-angle video image set processed by the first-layer generative adversarial network to generate a multi-view fused image with enhanced spatial perception capability;

[0014] S6, the second discriminator of the first layer of the generative adversarial network trains the multi-view fusion image generated by the second generator to be visually coherent;

[0015] S7. The multi-view fusion image processed by the double-layer generative adversarial network in step S6 is superimposed and displayed in real time in the AR smart glasses module, and is displayed synchronously with the wearer's first-person video image to optimize the wearer's on-site perception ability.

[0016] Optionally, the S1 includes the following steps:

[0017] S11, configuring an image processing unit in the AR smart glasses module of the police helmet, the image processing unit is used to receive real-time video data acquired by the multi-view camera module in the camera (217), and convert the real-time video data into a digital signal suitable for AR display, and generate a first-view video picture V1;

[0018] S12, using the image overlay processor in the AR smart glasses module to augment the reality information I AR It is superimposed with the first-view video image V1 to generate a superimposed display image V′1:

[0019] V′1=V1+α1×I AR ;

[0020] Among them, V′1 is the displayed image after superposition, V1 is the first-view video screen, and I AR is the augmented reality information, α1 is the image superposition weight coefficient;

[0021] S13, projecting the superimposed display image V′1 into the wearer's AR smart glasses (212) in real time through the display unit in the AR smart glasses module;

[0022] S14. Using the sensor unit integrated in the AR smart glasses (212), the head posture and sight direction of the wearer are monitored in real time, and the display position and angle of the displayed image V′1 are dynamically adjusted.

[0023] Optionally, S2 includes the following steps:

[0024] S21. A plurality of camera units are configured in the multi-view camera module of the police helmet, each camera unit being set in a different direction to capture video images of different angles of the surrounding environment of the police officer, and generate a multi-angle video image set:

[0025] D={V 21 ,V 22 ,,V 2n};

[0026] S22, preprocessing the original multi-angle video image set obtained by each camera unit, including image denoising, color correction and resolution enhancement, to generate a video image V′ 2i , where i is the camera unit number, V′ 2i is the original video image, V′ 2i is a pre-processed video image;

[0027] S23, the pre-processed multi-angle video image set D′={V′ 21 ,V′ 22 ,…,V′ 2n} is stored in the image data storage unit built into the helmet, and the multi-angle video image set D′ is transmitted in real time to the image processing unit of the police helmet through the built-in communication module.

[0028] Optionally, S3 includes the following steps:

[0029] S31, input the first-view display image V′1 and the preprocessed multi-angle video image set D′ into the first-layer generator G1 of the two-layer generative adversarial network model built into the helmet. The first-layer generator uses a convolutional neural network to extract features from the input data to obtain a feature mapping set

[0030] S32, in the first generator G1, the feature mapping set Processing is performed to generate a denoised, super-resolution reconstructed and low-light enhanced image set D″;

[0031] S33, the enhanced image set D″ generated by the first layer generator G1 is input into the first layer discriminator D1, and the first layer discriminator calculates the difference measurement function L by comparing the features of the generated image set D″ with the original image set D′. diff (D″,D′):

[0032]

[0033] in, is the feature extraction function in the first layer discriminator D1, L diffis the difference measure between the enhanced image and the original image, n is the number of images, D1∥·∥ 2 It is the Euclidean distance metric, which is used to quantify the similarity between images;

[0034] S34, based on the difference metric function L calculated by the first layer discriminator D1 diff (D″,D′) is the parameter of the first layer generator G1 Perform back-propagation update:

[0035]

[0036] Where η is the learning rate, is the parameter vector of the first layer generator. The parameter vector is adjusted by the gradient descent method to reduce the difference between the enhanced image and the original image, and the performance of the two-layer generative adversarial network model is gradually optimized until the generated image set D″ reaches the preset quality standard.

[0037] Optionally, the S32 includes the following steps:

[0038] S321, in the first generator G1, use the denoising function D denoise For feature map set Each image feature map in Perform denoising to remove the noise components in the image and obtain the denoised feature map

[0039]

[0040] Among them, K is the number of components of the Gaussian mixture model, w k is the weight of the Gaussian component, μ k and σ k are the mean and variance of the Gaussian components, is the input feature map, is the feature map after denoising;

[0041] S322, feature mapping after denoising Use super-resolution reconstruction function Processing is performed to optimize the resolution of the image and obtain the feature map after super-resolution reconstruction

[0042]

[0043] Among them, * represents the convolution operation, H is the convolution kernel, λ is the regularization parameter, is the image gradient, is the feature map after super-resolution reconstruction;

[0044] S323, feature mapping after super-resolution reconstruction Use low light enhancement function ε lle Processing is performed to optimize the visual effect of the image under low light conditions and obtain the final enhanced feature map

[0045]

[0046] Among them, ∈ is a small amount to avoid division by zero, α2 is a brightness adjustment parameter, It is the feature map after low-light enhancement processing;

[0047] S324, feature mapping after denoising, super-resolution reconstruction and low-light enhancement Perform summary processing to generate the final enhanced image set D":

[0048]

[0049] in, is the probability distribution function of the feature map, D″ is the final generated enhanced image set, Indicates that all images are processed and summarized. The final enhancement result for each image.

[0050] Optionally, S5 includes the following steps:

[0051] S51, input the enhanced image set D″ processed by the first layer of the generative adversarial network into the second generator G2 of the second layer of the generative adversarial network model, and the second generator G2 extracts features from the input enhanced image set to generate a feature mapping set

[0052] S52, using the multi-view fusion function in the second generator G2 For feature map set Fusion is performed to generate a multi-view fusion feature map with enhanced spatial perception capabilities

[0053]

[0054] Among them, γ ij is the adaptive weight for multi-view fusion, * represents the convolution operation, Ω j is the convolution kernel in the fusion process, exp is the exponential function, μ ij and σ ij are the mean and standard deviation parameters in the Gaussian mixture model, which are used to describe the similarity and importance between different feature maps;

[0055] S53, fused multi-view feature mapping Reconstruct and generate a multi-view fusion image V with enhanced spatial perception ability fusion :

[0056]

[0057] Among them, τ is the reconstruction function, δ l is the reconstruction weight, ReLU is the linear rectification function, η ij is the weight parameter of the fusion feature map, Γ l To reconstruct the convolution kernel, ξ l is the bias term, and ζ is the adjustment term in the reconstruction process.

[0058] Optionally, the S7 includes the following steps:

[0059] S71, the multi-view fusion image V processed by the double-layer generative adversarial network in step S6 fusion Synchronize with the wearer's first-person video image V1 in real time, using the synchronization mapping function S sync Align the two in space and time to obtain a synchronous fusion image V sync :

[0060] V sync =S sync (V fusion ,V1)=α3·V fusion +(1-α3)·V1;

[0061] Among them, α3 is the fusion weight coefficient, which is used to control the fusion degree of the multi-view fusion image and the first-view video picture, V sync is the synchronized fusion image;

[0062] S72, using the display processing unit in the AR smart glasses module to synchronize the fusion image V sync The real-time superimposed display is in the wearer's field of view, and the image display position and angle are adjusted according to the wearer's line of sight and head posture;

[0063] S73, using the real-time feedback mechanism in the AR smart glasses module to continuously monitor the wearer's environmental changes and movements, and dynamically adjust the synchronous fusion image V sync The display content and display method can optimize the wearer's on-site perception ability.

[0064] The beneficial effects of the present invention are:

[0065] (1) The present invention successfully solves the problems of information transmission delay and single perspective in the prior art by integrating a multi-view camera module and a two-layer generative adversarial network model. The first-layer generative adversarial network can denoise, super-resolution reconstruct, and low-light enhance video image sets from multiple angles, thereby generating a high-quality image set. Subsequently, the second-layer generative adversarial network further fuses the enhanced image set to generate a multi-view fused image with stronger spatial perception capability, thereby improving the comprehensive judgment capability of the on-site situation and the efficiency of command decision-making.

[0066] (2) The present invention utilizes the AR smart glasses module to synchronously overlay and display the multi-perspective fusion image with the first-person perspective video of the law enforcement personnel, thereby improving the real-time and accuracy of information display and enhancing the spatial perception ability of the law enforcement personnel at the scene. In the complex and changeable law enforcement environment, the synchronous fusion display effectively makes up for the limitations of the traditional single-perspective display, enabling law enforcement personnel to obtain comprehensive environmental information while responding quickly.

[0067] (3) In the design of a two-layer generative adversarial network, the present invention combines a convolutional neural network for feature extraction, denoising and super-resolution reconstruction, and introduces a multi-view fusion function to achieve adaptive weight adjustment of feature mapping, thereby improving the accuracy of image processing, reducing the complexity of calculation, and ensuring the real-time generation and display of multi-view fusion images.

[0068] (4) In the design of the smart helmet, the present invention fully considers the comfort of long-term wearing and the heat dissipation of internal electronic equipment. Through the combined design of the heat dissipation frame and the rubber column, good air circulation is ensured during the use of the helmet, heat accumulation is avoided, and the wearing comfort and stability of the equipment are significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0070] Figure 1 The present invention is a front view cross-sectional schematic diagram of the structure of a multifunctional police terminal smart helmet.

[0071] Figure 2 It is a front view schematic diagram of the identification component structure of a multifunctional police terminal smart helmet of the present invention.

[0072] Figure 3 The present invention is a front view cross-sectional schematic diagram of a heat dissipation component structure of a multifunctional police terminal smart helmet.

[0073] Figure 4 A multifunctional police terminal intelligent helmet of the present invention Figure 1A schematic diagram of the enlarged local structure at center A;

[0074] Figure 5 This is a flow chart of a multifunctional police terminal smart helmet of the present invention.

[0075] The accompanying drawings are marked as follows: 100, device body; 110, helmet body; 200, identification component; 210, mounting axis; 211, bracket; 212, AR smart glasses; 213, slide rail; 214, connecting piece; 215, cleaning strip; 216, battery compartment; 217, camera; 300, heat dissipation component; 310, heat dissipation rack; 311, sponge strip; 312, rubber column. DETAILED DESCRIPTION

[0076] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0077] refer to Figure 1-5 A multifunctional police terminal smart helmet includes a device body 100, which includes a helmet body 110. An identification component 200 is arranged inside the helmet body 110. The identification component 200 is arranged on both sides of the inner wall of the helmet body 110 through two installation shafts 210. A bracket 211 is hinged on the installation shaft 210, and an AR smart glasses 212 is fixedly installed on the end of the bracket 211 away from the installation shaft 210. A camera 217 is fixedly installed on the outside of the helmet body 110, a battery compartment 216 is fixedly installed on the outside of the helmet body 110, a fill light component is fixedly installed on the outside of the helmet body 110 outside the camera 217, and a heat dissipation component 300 is arranged inside the helmet body 110.

[0078] In this embodiment, a cleaning strip 215 that fits the AR smart glasses 212 is installed on one side of the inner wall of the helmet body 110, a slide rail 213 is fixedly installed on the inner wall of the helmet body 110, a slider is slidably installed inside the slide rail 213, a connecting piece 214 connected to the AR smart glasses 212 is fixedly installed on the outer side of the slider inside the slide rail 213, and the heat dissipation component 300 includes rubber columns 312 equidistantly installed on the inner wall of the helmet body 110, a heat dissipation frame 310 is fixedly installed on the bottom end of the rubber column 312, the heat dissipation frame 310 is evenly penetrated with through holes, and sponge strips 311 are equidistantly installed on the inner surface of the heat dissipation frame 310.

[0079] In this embodiment, a central processing unit is embedded in the recognition component 200, and an AR smart glasses module, a multi-view camera module and a double-layer generative adversarial network module are integrated in the central processing unit. The recognition method of the recognition component 200 includes:

[0080] S1. Integrate an AR smart glasses module into a police helmet, overlay display images in the wearer's field of view, and obtain the wearer's first-person video footage in real time;

[0081] S2, installing a multi-view camera module on a police helmet to capture a multi-angle video image set of the police officer's surrounding environment;

[0082] S3, inputting the first-person video image and the multi-angle video image set into the built-in double-layer generative adversarial network of the helmet for processing, wherein the first generator of the first-layer generative adversarial network generates a denoised, super-resolution reconstructed and low-light enhanced image from the multi-angle video image set;

[0083] S4, a first discriminator of the first layer of the generative adversarial network distinguishes the difference between the enhanced multi-angle video image set generated by the first generator and the original multi-angle video image set, and optimizes the realism of the generated multi-angle video image set;

[0084] S5, a second generator of the second-layer generative adversarial network, fusing the multi-angle video image set processed by the first-layer generative adversarial network to generate a multi-view fused image with enhanced spatial perception capability;

[0085] S6, the second discriminator of the first layer of the generative adversarial network trains the multi-view fusion image generated by the second generator to be visually coherent;

[0086] S7. The multi-view fusion image processed by the double-layer generative adversarial network in step S6 is superimposed and displayed in real time in the AR smart glasses module, and is displayed synchronously with the wearer's first-person video image to optimize the wearer's on-site perception ability.

[0087] In this implementation, S1 includes the following steps:

[0088] S11, configuring an image processing unit in the AR smart glasses module of the police helmet, the image processing unit is used to receive real-time video data acquired by the multi-view camera module in the camera 217, and convert the real-time video data into a digital signal suitable for AR display, and generate a first-view video picture V1;

[0089] S12, using the image overlay processor in the AR smart glasses module to augment the reality information I AR It is superimposed with the first-view video image V1 to generate a superimposed display image V′1:

[0090] V′1=V1+α1×I AR ;

[0091] Among them, V′1 is the displayed image after superposition, V1 is the first-view video screen, and I ARis the augmented reality information, α1 is the image superposition weight coefficient;

[0092] S13, projecting the superimposed display image V′1 to the wearer's AR smart glasses 212 in real time through the display unit in the AR smart glasses module;

[0093] S14. Through the sensor unit integrated in the AR smart glasses 212, the head posture and sight direction of the wearer are monitored in real time, and the display position and angle of the displayed image V′1 are dynamically adjusted.

[0094] In this implementation, S2 includes the following steps:

[0095] S21. A plurality of camera units are configured in the multi-view camera module of the police helmet, each camera unit being set in a different direction to capture video images of different angles of the surrounding environment of the police officer, and generate a multi-angle video image set:

[0096] D={V 21 ,V 22 ,,V 2n};

[0097] S22, preprocessing the original multi-angle video image set obtained by each camera unit, including image denoising, color correction and resolution enhancement, to generate a video image V′ 2i , where i is the camera unit number, V′ 2i is the original video image, V′ 2i is a pre-processed video image;

[0098] S23, the pre-processed multi-angle video image set D′={V′ 21 ,V′ 22 ,…,V′ 2n} is stored in the image data storage unit built into the helmet, and the multi-angle video image set D′ is transmitted in real time to the image processing unit of the police helmet through the built-in communication module.

[0099] In this implementation, S3 includes the following steps:

[0100] S31, input the first-view display image V′1 and the preprocessed multi-angle video image set D′ into the first-layer generator G1 of the two-layer generative adversarial network model built into the helmet. The first-layer generator uses a convolutional neural network to extract features from the input data to obtain a feature mapping set

[0101] S32, in the first generator G1, the feature mapping set Processing is performed to generate a denoised, super-resolution reconstructed and low-light enhanced image set D″;

[0102] S33, the enhanced image set D″ generated by the first layer generator G1 is input into the first layer discriminator D1, and the first layer discriminator calculates the difference measurement function L by comparing the features of the generated image set D″ with the original image set D′. diff (D″,D′):

[0103]

[0104] in, is the feature extraction function in the first layer discriminator D1, L diff is the difference measure between the enhanced image and the original image, n is the number of images, D1∥·∥ 2 It is the Euclidean distance metric, which is used to quantify the similarity between images;

[0105] S34, based on the difference metric function L calculated by the first layer discriminator D1 diff (D″,D′) is the parameter of the first layer generator G1 Perform back-propagation update:

[0106]

[0107] Where η is the learning rate, is the parameter vector of the first layer generator. The parameter vector is adjusted by the gradient descent method to reduce the difference between the enhanced image and the original image, and the performance of the two-layer generative adversarial network model is gradually optimized until the generated image set D″ reaches the preset quality standard.

[0108] In this implementation, S32 includes the following steps:

[0109] S321, in the first generator G1, use the denoising function D denoise For feature map set Each of

[0110] Image feature map Perform denoising to remove the noise components in the image and obtain the denoised feature map

[0111]

[0112] Among them, K is the number of components of the Gaussian mixture model, w k is the weight of the Gaussian component, μ k and σ k are the mean and variance of the Gaussian components, is the input feature map, is the feature map after denoising;

[0113] S322, feature mapping after denoising Use super-resolution reconstruction function Processing is performed to optimize the resolution of the image and obtain the feature map after super-resolution reconstruction

[0114]

[0115] Among them, * represents the convolution operation, H is the convolution kernel, λ is the regularization parameter, is the image gradient, is the feature map after super-resolution reconstruction;

[0116] S323, feature mapping after super-resolution reconstruction Use low light enhancement function ε lle Processing is performed to optimize the visual effect of the image under low light conditions and obtain the final enhanced feature map

[0117]

[0118] Among them, ∈ is a small amount to avoid division by zero, α2 is a brightness adjustment parameter, It is the feature map after low-light enhancement processing;

[0119] S324, feature mapping after denoising, super-resolution reconstruction and low-light enhancement Perform summary processing to generate the final enhanced image set D":

[0120]

[0121] in, is the probability distribution function of the feature map, D″ is the final generated enhanced image set, Indicates that all images are processed and summarized. The final enhancement result for each image.

[0122] In this implementation, S5 includes the following steps:

[0123] S51, input the enhanced image set D″ processed by the first layer of the generative adversarial network into the second generator G2 of the second layer of the generative adversarial network model, and the second generator G2 extracts features from the input enhanced image set to generate a feature mapping set

[0124] S52, using the multi-view fusion function in the second generator G2 For feature map set Fusion is performed to generate a multi-view fusion feature map with enhanced spatial perception capabilities

[0125]

[0126] Among them, γ ij is the adaptive weight for multi-view fusion, * represents the convolution operation, Ω j is the convolution kernel in the fusion process, exp is the exponential function, μ ij and σ ij are the mean and standard deviation parameters in the Gaussian mixture model, which are used to describe the similarity and importance between different feature maps;

[0127] S53, fused multi-view feature mapping Reconstruct and generate a multi-view fusion image V with enhanced spatial perception ability fusion :

[0128]

[0129] Among them, τ is the reconstruction function, δ l is the reconstruction weight, ReLU is the linear rectification function, η ij is the weight parameter of the fusion feature map, Γ l To reconstruct the convolution kernel, ξ l is the bias term, and ζ is the adjustment term in the reconstruction process.

[0130] In this implementation, S7 includes the following steps:

[0131] S71, the multi-view fusion image V processed by the double-layer generative adversarial network in step S6 fusion Synchronize with the wearer's first-person video image V1 in real time, using the synchronization mapping function S sync Align the two in space and time to obtain a synchronous fusion image V sync :

[0132] V sync =S sync (V fusion ,V1)=α3·V fusion +(1-α3)·V1;

[0133] Among them, α3 is the fusion weight coefficient, which is used to control the fusion degree of the multi-view fusion image and the first-view video picture, V sync is the synchronized fusion image;

[0134] S72, using the display processing unit in the AR smart glasses module to synchronize the fusion image V syncThe real-time superimposed display is in the wearer's field of view, and the image display position and angle are adjusted according to the wearer's line of sight and head posture;

[0135] S73, using the real-time feedback mechanism in the AR smart glasses module to continuously monitor the wearer's environmental changes and movements, and dynamically adjust the synchronous fusion image V sync The display content and display method can optimize the wearer's on-site perception ability.

[0136] Embodiment 1:

[0137] In this embodiment 1, in August 2024, a large-scale drill was held in the central square of City B. The drill simulated an incident in which a suspect took hostages and threatened to detonate a bomb. After receiving the order, the special police team of the Municipal Public Security Bureau quickly dispatched a team to the scene to deal with it. The special police team wore the multifunctional intelligent police helmet of the present invention. Through the helmet system, the on-site information can be transmitted to the command center in real time to ensure the timeliness and accuracy of decision-making.

[0138] When the SWAT team arrived at the scene, the square had been blocked off and a large number of onlookers gathered around. Time was running out, and the command center asked the SWAT team to quickly locate and identify the suspect. The team members wearing smart police helmets activated the multi-view camera module. The helmet system immediately captured and processed image data from all angles of the square. Through the two-layer generative adversarial network, the first layer generator quickly processed the images captured on the scene. The SWAT team found that in a sheltered area in the southeast corner, a man was behaving abnormally, and his body shape was highly consistent with the description of the suspect provided by the command center. After image denoising and super-resolution reconstruction, the final image clearly showed the man's facial features.

[0139] During further observation, the team members discovered through the helmet system that a black package next to the man had abnormal thermal imaging characteristics. Real-time data analysis showed that the temperature changes of the package did not conform to the temperature curve of conventional objects and it was very likely to contain an explosive device. The SWAT team used the multi-perspective fusion technology of the helmet to integrate the images from various cameras and generated a complete three-dimensional scene model, clearly marking the suspect's location, the specific location of the package, and the surrounding environment.

[0140] Based on this information, the command center directed the team members to surround the suspect from different directions and formulated an action plan. The AR smart glasses module of the smart police helmet displayed the scene after multi-perspective fusion, and projected the instructions in real time into the team members' field of vision. At the scene, the team members used the real-time feedback mechanism of the smart helmet to dynamically adjust their position and angle to ensure the tightness of the encirclement.

[0141] During the operation, the system continued to highlight the suspect and the package he was carrying through the visual attention enhancement algorithm. The team members judged the specific location and dynamic changes of the potential threat. When the suspect tried to flee the scene, the smart helmet immediately updated his movement trajectory and displayed the trajectory in the team members' field of view through the AR glasses module. The suspect was subsequently successfully controlled and the package was quickly transferred to a safe area for bomb disposal.

[0142] In order to verify the actual effect of the intelligent police helmet of the present invention, detailed data of the drill was recorded and compared with the performance of traditional police equipment in similar situations:

[0143] Table 1 Comparison data between the present invention and traditional police equipment under similar circumstances

[0144]

[0145] As can be seen from Table 1 above, the accuracy of traditional equipment in identifying the facial features of suspects in low-light environments is about 65%. With the assistance of the smart helmet system of the present invention, after super-resolution reconstruction and low-light enhancement processing, the recognition accuracy is increased to 92%, effectively reducing the misjudgment and missed judgment. Traditional equipment can usually only provide image coverage from a single perspective, resulting in a lack of comprehensive information support when formulating action plans. The smart police helmet integrates images from different angles through multi-perspective fusion technology to generate a 360-degree scene model without blind spots, which can monitor suspects and the environment in all directions, improving the safety and success rate of the action. In the same drill scenario, the average time from target discovery to action completion of traditional equipment is 18 minutes, while the smart police helmet system shortens this time to 12 minutes, and the reaction speed is increased by 33%. During the drill, the heat dissipation system of the smart police helmet performed well. Even if it worked continuously for more than 1 hour in a high temperature environment of 30 degrees Celsius, the internal temperature of the equipment remained within the normal range, and the wearer reported a high degree of comfort. In a similar environment, the internal temperature of the traditional helmet equipment was high. After using it for more than 40 minutes, the wearer had obvious discomfort, which affected its operating efficiency.

[0146] Through the application examples of this exercise, it can be clearly seen that the intelligent police helmet system of the present invention has demonstrated excellent performance in complex and high-pressure law enforcement scenarios. Whether in terms of real-time information acquisition, image processing accuracy, multi-perspective fusion effect, or improving operational efficiency and ensuring the comfort of law enforcement personnel, the intelligent police helmet far exceeds traditional equipment. The successful application of the system in the exercise has fully demonstrated its feasibility and effectiveness. It is expected to be widely used in more complex police tasks in the future and become an important scientific and technological means to ensure public safety.

[0147] The invention successfully solves the problems of information transmission delay and single perspective in the prior art by integrating a multi-view camera module and a two-layer generative adversarial network model. The first-layer generative adversarial network can denoise, super-resolution reconstruct and low-light enhance video image sets from multiple angles, thereby generating a high-quality image set. Subsequently, the second-layer generative adversarial network further fuses the enhanced image set to generate a multi-view fused image with stronger spatial perception capability, thereby improving the comprehensive judgment ability of the on-site situation and the efficiency of command decision-making.

[0148] The present invention utilizes the AR smart glasses module to synchronously overlay and display the multi-perspective fusion image with the first-perspective video screen of the law enforcement personnel, thereby improving the real-time and accuracy of information display and enhancing the spatial perception ability of the law enforcement personnel at the scene. In the complex and changeable law enforcement environment, the synchronous fusion display effectively makes up for the limitations of the traditional single-perspective display, enabling law enforcement personnel to quickly respond while obtaining comprehensive environmental information.

[0149] In the design of a two-layer generative adversarial network, the present invention combines a convolutional neural network for feature extraction, denoising and super-resolution reconstruction, and introduces a multi-view fusion function to achieve adaptive weight adjustment of feature mapping, thereby improving the accuracy of image processing, reducing the complexity of calculation, and ensuring the real-time generation and display of multi-view fusion images.

[0150] In the design of the smart helmet, the present invention fully considers the comfort of long-term wearing and the heat dissipation of internal electronic equipment. Through the combined design of the heat dissipation frame and the rubber column, good air circulation of the helmet during use is guaranteed, heat accumulation is avoided, and the wearing comfort and the stability of the equipment are significantly improved.

[0151] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A multifunctional police terminal intelligent helmet, comprising a device body (100), wherein the device body (100) comprises a helmet body (110), characterized in that: An identification component (200) is provided inside the helmet body (110), and the identification component (200) is arranged on both sides of the inner wall of the helmet body (110) through two installation shafts (210), a bracket (211) is hinged on the installation shaft (210), and an AR smart glasses (212) are fixedly installed on the end of the bracket (211) away from the installation shaft (210), a camera (217) is fixedly installed on the outside of the helmet body (110), a battery compartment (216) is fixedly installed on the outside of the helmet body (110), and a fill light component is fixedly installed on the outside of the helmet body (110) outside the camera (217), and a heat dissipation component (300) is provided inside the helmet body (110).

2. The multifunctional police terminal smart helmet according to claim 1 is characterized in that: A cleaning strip (215) that fits the AR smart glasses (212) is installed on one side of the inner wall of the helmet body (110); a slide rail (213) is fixedly installed on the inner wall of the helmet body (110); a slider is slidably installed inside the slide rail (213); a connecting piece (214) that is connected to the AR smart glasses (212) is fixedly installed on the outer side of the slider inside the slide rail (213); the heat dissipation component (300) includes rubber columns (312) that are equidistantly installed on the inner wall of the helmet body (110); a heat dissipation frame (310) is fixedly installed at the bottom end of the rubber column (312); through holes are evenly penetrated on the heat dissipation frame (310); and sponge strips (311) are equidistantly installed on the inner surface of the heat dissipation frame (310).

3. The multifunctional police terminal smart helmet according to claim 1 is characterized in that: The recognition component (200) is embedded with a central processing unit, and the central processing unit is integrated with an AR smart glasses module, a multi-view camera module and a double-layer generative adversarial network module. The recognition method of the recognition component (200) includes: S1. Integrate an AR smart glasses module into a police helmet, overlay display images in the wearer's field of view, and obtain the wearer's first-person video footage in real time; S2, installing a multi-view camera module on a police helmet to capture a multi-angle video image set of the police officer's surrounding environment; S3, inputting the first-person video image and the multi-angle video image set into the built-in double-layer generative adversarial network of the helmet for processing, wherein the first generator of the first-layer generative adversarial network generates a denoised, super-resolution reconstructed and low-light enhanced image from the multi-angle video image set; S4, a first discriminator of the first layer of the generative adversarial network distinguishes the difference between the enhanced multi-angle video image set generated by the first generator and the original multi-angle video image set, and optimizes the realism of the generated multi-angle video image set; S5, a second generator of the second-layer generative adversarial network, fusing the multi-angle video image set processed by the first-layer generative adversarial network to generate a multi-view fused image with enhanced spatial perception capability; S6, the second discriminator of the first layer of the generative adversarial network trains the multi-view fusion image generated by the second generator to be visually coherent; S7. The multi-view fusion image processed by the double-layer generative adversarial network in step S6 is superimposed and displayed in real time in the AR smart glasses module, and is displayed synchronously with the wearer's first-person video image to optimize the wearer's on-site perception ability.

4. The multifunctional police terminal smart helmet according to claim 3 is characterized in that: The S1 comprises the following steps: S11, configuring an image processing unit in the AR smart glasses module of the police helmet, the image processing unit is used to receive real-time video data acquired by the multi-view camera module in the camera (217), and convert the real-time video data into a digital signal suitable for AR display, and generate a first-view video picture V1; S12, using the image overlay processor in the AR smart glasses module to augment the reality information I AR It is superimposed with the first-view video image V1 to generate a superimposed display image V′1: V′1=V1+α1×I AR ; Among them, V′1 is the displayed image after superposition, V1 is the first-view video screen, and I AR is the augmented reality information, α1 is the image superposition weight coefficient; S13, projecting the superimposed display image V′1 into the wearer's AR smart glasses (212) in real time through the display unit in the AR smart glasses module; S14. Using the sensor unit integrated in the AR smart glasses (212), the head posture and sight direction of the wearer are monitored in real time, and the display position and angle of the displayed image V′1 are dynamically adjusted.

5. The multifunctional police terminal smart helmet according to claim 3 is characterized in that: The S2 comprises the following steps: S21. A plurality of camera units are configured in the multi-view camera module of the police helmet, each of which is set in a different direction to capture video images of different angles of the surrounding environment of the police officer, and generate a multi-angle video image set: D={V 21 ,V 22 ,,V 2n }; S22, preprocessing the original multi-angle video image set obtained by each camera unit, including image denoising, color correction and resolution enhancement, to generate a video image V′ 2i , where i is the camera unit number, V 2i is the original video image, V′ 2i is a pre-processed video image; S23, the pre-processed multi-angle video image set D′={V′ 21 ,V′ 22 ,…,V′ 2n } is stored in the image data storage unit built into the helmet, and the multi-angle video image set D′ is transmitted in real time to the image processing unit of the police helmet through the built-in communication module.

6. The multifunctional police terminal smart helmet according to claim 3 is characterized in that: The S3 comprises the following steps: S31, input the first-view display image V′1 and the preprocessed multi-angle video image set D′ into the first-layer generator G1 of the two-layer generative adversarial network model built into the helmet. The first-layer generator uses a convolutional neural network to extract features from the input data to obtain a feature mapping set S32, in the first generator G1, the feature mapping set Processing is performed to generate a denoised, super-resolution reconstructed and low-light enhanced image set D″; S33, the enhanced image set D″ generated by the first layer generator G1 is input into the first layer discriminator D1, and the first layer discriminator calculates the difference measurement function L by comparing the features of the generated image set D″ with the original image set D′. diff (D″,D′): in, is the feature extraction function in the first layer discriminator D1, L diff is the difference measure between the enhanced image and the original image, n is the number of images, D1∥·∥ 2 It is the Euclidean distance metric, which is used to quantify the similarity between images; S34, the difference metric function L calculated according to the first layer discriminator D1 diff (D″,D′) is the parameter of the first layer generator G1 Perform back-propagation update: Where η is the learning rate, is the parameter vector of the first layer generator. The parameter vector is adjusted by the gradient descent method to reduce the difference between the enhanced image and the original image, and the performance of the two-layer generative adversarial network model is gradually optimized until the generated image set D″ reaches the preset quality standard.

7. The multifunctional police terminal smart helmet according to claim 3 is characterized in that: The S32 comprises the following steps: S321, in the first generator G1, use the denoising function D denoise For feature map set Each image feature map in Perform denoising to remove the noise components in the image and obtain the denoised feature map Among them, K is the number of components of the Gaussian mixture model, w k is the weight of the Gaussian component, μ k and σ k are the mean and variance of the Gaussian components, is the input feature map, is the feature map after denoising; S322, feature mapping after denoising Use super-resolution reconstruction function Processing is performed to optimize the resolution of the image and obtain the feature map after super-resolution reconstruction Among them, * represents the convolution operation, H is the convolution kernel, λ is the regularization parameter, is the image gradient, is the feature map after super-resolution reconstruction; S323, feature mapping after super-resolution reconstruction Use low light enhancement function ε lle Processing is performed to optimize the visual effect of the image under low light conditions and obtain the final enhanced feature map Among them, ∈ is a small amount to avoid division by zero, α2 is a brightness adjustment parameter, It is the feature map after low-light enhancement processing; S324, feature mapping after denoising, super-resolution reconstruction and low-light enhancement Perform summary processing to generate the final enhanced image set D": in, is the probability distribution function of the feature map, D″ is the final generated enhanced image set, Indicates that all images are processed and summarized. The final enhancement result for each image.

8. The multifunctional police terminal smart helmet according to claim 3 is characterized in that: The S5 comprises the following steps: S51, input the enhanced image set D″ processed by the first layer of the generative adversarial network into the second generator G2 of the second layer of the generative adversarial network model, and the second generator G2 extracts features from the input enhanced image set to generate a feature mapping set S52, using the multi-view fusion function in the second generator G2 For feature map set Fusion is performed to generate a multi-view fusion feature map with enhanced spatial perception capabilities Among them, γ ij is the adaptive weight for multi-view fusion, * represents the convolution operation, Ω j is the convolution kernel in the fusion process, exp is the exponential function, μ ij and σ ij are the mean and standard deviation parameters in the Gaussian mixture model, which are used to describe the similarity and importance between different feature maps; S53, fused multi-view feature mapping Reconstruct and generate a multi-view fusion image V with enhanced spatial perception ability fusion : Among them, τ is the reconstruction function, δ l is the reconstruction weight, ReLU is the linear rectification function, η ij is the weight parameter of the fusion feature map, Γ l To reconstruct the convolution kernel, ξ l is the bias term, and ζ is the adjustment term in the reconstruction process.

9. The multifunctional police terminal smart helmet according to claim 3 is characterized in that: The S7 comprises the following steps: S71, the multi-view fusion image V processed by the double-layer generative adversarial network in step S6 fusion Synchronize with the wearer's first-person video image V1 in real time, using the synchronization mapping function S sync Align the two in space and time to obtain a synchronous fusion image V sync : V sync =S sync (V fusion ,V1)=α3·V fusion +(1-α3)·V1; Among them, α3 is the fusion weight coefficient, which is used to control the fusion degree of the multi-view fusion image and the first-view video picture, V sync is the synchronized fusion image; S72, using the display processing unit in the AR smart glasses module to synchronize the fusion image V sync The real-time superimposed display is in the wearer's field of view, and the image display position and angle are adjusted according to the wearer's line of sight and head posture; S73, using the real-time feedback mechanism in the AR smart glasses module to continuously monitor the wearer's environmental changes and movements, and dynamically adjust the synchronous fusion image V sync The display content and display method can optimize the wearer's on-site perception ability.