A wearable night vision system based on AI image processing method and AR glasses

Abstract

The present invention discloses a wearable night vision system based on an AI image processing method and AR glasses, and relates to the field of image processing technology. The system obtains night vision images through AI-ISP technology, extracts spatial features based on the night vision images, wherein the spatial features include feature information in the night vision images, and constructs an environmental model based on the spatial features; obtains the real-time position of the AR glasses, performs image understanding according to the real-time position corresponding to the environmental model to obtain image information, combines the image information with the environmental model to obtain an image model; and corresponds the image model to the viewing angle of the AR glasses to obtain a night vision image. In the process of the AR glasses changing, the connection line between the AR glasses and the sensing point will also change. Since the image model has been constructed in advance, the angle relationship between the connection line and the AR glasses can facilitate the subsequent rapid and accurate presentation of the image in the image model collected by the AR glasses according to the viewing angle of the AR glasses, and can improve the imaging processing efficiency and imaging quality of the AR glasses.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a wearable night vision system based on an AI image processing method and AR glasses. Background Art

[0002] AR glasses are a device that superimposes virtual images on real scenes. Users can see a mixture of virtual reality and the real world through the glasses. AR glasses contain cameras, gyroscopes, accelerometers, etc., which are used to perceive the user's head posture, position, and movement status in real time, and then cooperate with AR glasses to collect night vision images. In the use of AR glasses, AI image processing is an important technical means for cooperation, especially in the night vision operation of AR glasses. AI image processing can intelligently understand the image content in the night vision full-color image. However, at present, AR glasses need to process the collected night vision images during the night vision process. However, since the position of AR glasses is usually changing, it is difficult to guarantee the efficiency and quality of imaging in image processing. Summary of the invention

[0003] The purpose of the present invention is to provide a wearable night vision system based on AI image processing method and AR glasses to solve the shortcomings of the background technology.

[0004] In order to achieve the above object, the present invention provides the following technical solution: an AI-based image processing method, comprising the following steps:

[0005] Acquire a night vision image, extract spatial features based on the night vision image, wherein the spatial features include feature information in the night vision image, and construct an environment model based on the spatial features;

[0006] Obtain the real-time position of the AR glasses, perform image understanding based on the real-time position corresponding to the environmental model to obtain image information, and combine the image information with the environmental model to obtain an image model;

[0007] The image model is matched to the viewing angle of the AR glasses to obtain a night vision image.

[0008] In a preferred embodiment, the step of acquiring night vision images includes:

[0009] Construct a spatial model and set a spatial coordinate system on the spatial model;

[0010] Based on the images collected by AR glasses, the dark light images are restored to full-color images through AI-ISP technology, and multiple features in the full-color images are extracted and marked in the spatial coordinate system in the spatial model;

[0011] The state of multiple features in the spatial model is used as a night vision image.

[0012] In a preferred embodiment, the step of constructing an environment model based on spatial features includes:

[0013] Extracting features from the night vision image to construct feature information in the night vision image as spatial features, wherein the feature information is a spatial model and a spatial position of the feature in the night vision image;

[0014] The features in the night vision image are replaced by the corresponding spatial features and placed in the spatial model to obtain the environmental model.

[0015] In a preferred embodiment, the step of combining the image information with the environment model to obtain the image model includes:

[0016] The real-time position of the AR glasses is obtained in real time, a new space model is configured according to the real-time position of the AR glasses, and the new space model is combined with the environment model as a combined model;

[0017] Overlaying the combined model based on the spatial coordinate system in the environment model;

[0018] Supplement the spatial features of the combined model and spatially locate the spatial features;

[0019] According to the real-time position of the AR glasses and the full-color images collected by the AR glasses, the acquisition relationship between the AR glasses and the spatial features in the combined model is updated in real time;

[0020] The combined model of the acquisition relationship between AR glasses and spatial features is established as the image model.

[0021] In a preferred embodiment, the step of acquiring the real-time position of the AR glasses in real time, configuring a new space model according to the real-time position of the AR glasses, and combining the new space model with the environment model as a combined model includes:

[0022] Obtain the horizontal displacement between the real-time position of the AR glasses and the center point of the previous environment model in real time, and match the new space model when the horizontal displacement exceeds the preset displacement;

[0023] The overlapping parts of the new spatial model and the environmental model are merged, and the non-overlapping parts of the new spatial model and the environmental model are reconstructed and combined to obtain a combined model.

[0024] In a preferred embodiment, the step of supplementing the spatial features of the combined model and spatially locating the spatial features includes:

[0025] Obtain a full-color image of the non-overlapping part of the new space model and the environment model, extract multiple features in the full-color image and mark the corresponding spatial coordinate system in the combined model;

[0026] The multiple features added in the combined model are spatially constructed to obtain the spatial features added in the combined model, and the spatial features added in the combined model are placed in the spatial model instead of the corresponding features;

[0027] The center points of multiple spatial models in the combined model are obtained, the center points of the multiple spatial models are respectively mapped downward to obtain root points, and the spatial positioning between the root points and the corresponding spatial features is established.

[0028] In a preferred embodiment, the step of updating the acquisition relationship between the AR glasses and the spatial features in the combined model in real time according to the real-time position of the AR glasses and the full-color image acquired by the AR glasses includes:

[0029] Configure the root point with the corresponding sensing point, and connect the root point to the AR glasses based on the sensing point;

[0030] According to the real-time position of the AR glasses, the connection line between the AR glasses and the spatial feature sensing point is updated to obtain the collection relationship between the AR glasses and the spatial features.

[0031] The present invention also provides a wearable night vision system for AR glasses, comprising:

[0032] An acquisition module, used for acquiring night vision images, extracting spatial features based on the night vision images, wherein the spatial features include feature information in the night vision images, and constructing an environment model based on the spatial features;

[0033] A construction module is connected to the acquisition module and is used to obtain the real-time position of the AR glasses, perform image understanding according to the real-time position corresponding to the environment model to obtain image information, and combine the image information with the environment model to obtain an image model;

[0034] The imaging module is connected to the building module and is used to match the image model to the viewing angle of the AR glasses to obtain a night vision image.

[0035] In a preferred embodiment, the acquisition module includes:

[0036] A construction unit, used for constructing a spatial model and setting a spatial coordinate system on the spatial model;

[0037] The marking unit is used to restore the dark light image into a full-color image based on the image collected by the AR glasses through the AI-ISP technology, and extract multiple features in the full-color image and mark the corresponding spatial coordinate system in the spatial model;

[0038] A management unit is used for taking the state of a plurality of features in a spatial model as a night vision image.

[0039] In a preferred embodiment, the acquisition module further includes:

[0040] An extraction unit, used to extract features in the night vision image and construct feature information in the night vision image as a spatial feature, wherein the feature information is a spatial model and a spatial position of the feature in the night vision image;

[0041] The setting unit is used to replace the features in the night vision image with the corresponding spatial features and place them in the spatial model to obtain the environment model.

[0042] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0043] The present invention uses an environmental model construction method to update and construct a combined model of full-color images collected by AR glasses within the range of a spatial model, and then establishes a collection relationship between the AR glasses and spatial features in the updated combined model. The combined model with the established collection relationship between the AR glasses and the spatial features is used as an image model, so that the night vision image display effect of the AR glasses can be obtained, so that the image can be displayed more clearly.

[0044] As the AR glasses change, the connection line between the AR glasses and the sensing point will also change. Since the image model has been constructed in advance, the angle relationship between the connection line and the AR glasses can facilitate the subsequent rapid and accurate presentation of the image in the image model captured by the AR glasses based on the viewing angle of the AR glasses, thereby improving the imaging processing efficiency of the AR glasses. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0046] Figure 1 The figure is a flow chart of the method of the present invention.

[0047] Figure 2 It is a system block diagram of the present invention. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] Example 1, please refer to Figure 1 As shown, the AI-based image processing method described in this embodiment includes the following steps:

[0050] S1. Acquire a night vision image, extract spatial features based on the night vision image, wherein the spatial features include feature information in the night vision image, and construct an environment model based on the spatial features;

[0051] S2, obtaining the real-time position of the AR glasses, performing image understanding according to the real-time position corresponding to the environment model to obtain image information, and combining the image information with the environment model to obtain an image model;

[0052] S3, matching the image model to the viewing angle of the AR glasses to obtain a night vision image;

[0053] According to the angular relationship between the connecting line and the spatial feature, a plane display image of the spatial feature from the perspective of the connecting line is given;

[0054] Based on the perspective of the AR glasses (representing the angle of the connection line), a plane display image of the spatial features of the corresponding connection line can be obtained, and the plane display images of multiple spatial features under the perspective of the AR glasses are integrated within the acquisition range under the perspective of the AR glasses as a night vision image;

[0055] After determining the viewing angle of the AR glasses, a plane display image of the spatial features corresponding to the connecting lines is obtained according to the corresponding connecting lines under the viewing angle of the AR glasses. After determining the viewing angle of the AR glasses, it has a collection range under the viewing angle, so the plane display images corresponding to multiple spatial features under the collection range are integrated within the collection range according to the positions of the corresponding spatial features to obtain a night vision image;

[0056] As described in the above steps S1-S3, AR glasses are a device that superimposes virtual images on real scenes. Users can see a mixture of virtual reality and the real world through the glasses. In the use of AR glasses, AI image processing is an important means of cooperation, especially in the night vision operation of AR glasses. AI image processing can intelligently understand the image content in the full-color image collected by night vision. However, at present, AR glasses need to process the collected night vision images during the night vision process. Since AR glasses usually change position, the image processing efficiency is low and it is difficult to ensure the efficiency and quality of imaging processing. In the present application, during the change of AR glasses, the connection line between the AR glasses and the sensing point will also change. Since the image model has been constructed in advance, the angle relationship between the connection line and the AR glasses can facilitate the subsequent rapid and accurate presentation of the image in the image model collected by the AR glasses according to the viewing angle of the AR glasses, thereby improving the imaging processing efficiency and imaging quality of the AR glasses.

[0057] In one embodiment, the step S11 of acquiring night vision images includes:

[0058] S111, constructing a spatial model, and setting a spatial coordinate system on the spatial model;

[0059] S112, based on the image collected by AR glasses, the dark light image is restored to a full-color image through AI-ISP technology, and multiple features in the full-color image are extracted and marked in the spatial coordinate system in the spatial model;

[0060] S113, taking the state of the multiple features in the space model as a night vision image;

[0061] As described in the above steps S111-S113, a spatial model is constructed. The spatial model here is a model of the spatial range corresponding to the farthest range of the full-color image acquisition. Here, the spatial model is constructed with the position of the AR glasses as the center. The spatial model here is a blank virtual spatial model. It can be constructed through three-dimensional modeling software such as 3ds Max, Maya, Blender, etc., such as polygonal modeling, NURBS modeling, etc., to create a spatial model with high realism and complex details. Various decorative elements can be added to the blank space to simulate a real indoor or outdoor environment. In this application, they are used to add features in the night vision image later. The shape of the spatial model for night vision here is determined by the shape of the farthest range that can be collected by the full-color image. In order to facilitate the carrying and combination of the spatial model with subsequent images, it is necessary to set a spatial coordinate system on the spatial model to understand the position in the spatial model. Since the AR glasses are used at night here, the images collected by the AR glasses can be restored in real time with 25 to 30 frames of color through AI-ISP technology in an environment with extremely low illumination (such as <0.01Lux), without the need for white light fill light or infrared fill light, with a color saturation of ≥100%, and ensuring that there is no obvious drag, perspective or ghosting of moving objects. The conversion process here is: AI-ISP uses a deep learning algorithm and uses massive scenes and data as a training data set (here, the training data set is obtained by collecting historical data); pre-processing operations are performed on the collected images, such as using artificial intelligence (AI) and image signal processing (ISP); artificial intelligence can analyze and identify the content in the image by learning and understanding a large amount of image data, thereby optimizing and enhancing the image. Image signal processing refers to various processing and adjustments to images, such as noise reduction, contrast enhancement, and repair of insufficient light, to improve image quality and clarity. The advantage of AI-ISP technology is that it can further improve the effect of image processing through the support of intelligent algorithms and big data.Compared with traditional ISP technology, AI-ISP technology can more accurately identify and analyze the content in the image, and adjust various parameters of the image according to different scenarios and needs, making the image details clearer and the color more accurate. The AI ​​image processing here is integrated in the chip of AR glasses, and the AR glasses process the picture through the AI ​​chip; select a suitable model (such as CycleGAN) for training, and optimize the conversion effect by continuously adjusting the network parameters and training strategies; apply the trained model to the new full-color image for necessary post-processing operations (such as smoothing, color correction, etc.), and then extract the features in the full-color image and sort them according to the feature location. The position corresponds to the spatial coordinate system marked in the spatial model. The feature here is the feature subject in the full-color image. A feature subject represents an object image in the full-color image. The feature extraction can use machine learning methods to take the state of multiple features in the spatial model as the night vision image. The position of the feature in the full-color image here can be obtained in the spatial model through the full-color image. The night vision image here is the state of the feature in the spatial model after conversion into a full-color image as the night vision image, which can facilitate the subsequent image understanding and processing of the night vision image, and better improve the night vision ability of AR glasses. The AR glasses realize the image display function, which is more convenient and comfortable than traditional eyepieces. At the same time, AR glasses are translucent, which allows virtual information to merge with the real world, allowing users to see the night vision effect and the real environment at the same time;.

[0062] In one embodiment, the step S12 of constructing the environment model based on the spatial features includes:

[0063] S121, extracting features from the night vision image to construct feature information in the night vision image as spatial features, wherein the feature information is a spatial model and a spatial position of the features in the night vision image;

[0064] S122, replacing the features in the night vision image with corresponding spatial features and placing them in the spatial model to obtain an environment model;

[0065] As described in the above steps S121 and S122, the features in the night vision image are extracted to construct the feature information in the night vision image as a spatial feature. The feature is constructed in three dimensions through multiple features extracted from the full-color image. The specific operations are as follows: first, a basic geometric body (such as a cube, a sphere, etc.) is created in the 3D modeling software as the starting point for feature modeling, the basic geometric body is converted into an editable polygon, and a sub-object is selected. In the editable polygon mode, the sub-object levels such as vertices, edges, boundaries, polygons (faces) and elements can be selected for editing, and the position, direction and size of the selected sub-object can be adjusted with basic transformation tools; the extrusion command is applied to the selected face to create a new volume or shape. The extrusion operation can be combined with moving, rotating and scaling to further adjust the shape of the extruded part; a new line is added between two points or two lines to increase the details of the model or smooth the transition; a new line is added to the model to split the face or create a new polygon. The cutting operation is very flexible, and different numbers of lines can be added at any position; the selected face is chamfered to produce a rounded effect. The chamfering operation is similar to a combination of extrusion and scaling; the selected points or lines are chamfered to increase the details and three-dimensional sense of the model; two or more close points are merged to reduce the number of vertices and optimize the model structure. The welding operation requires setting a welding threshold to control the accuracy of the merge; a part of the model is separated from the main body to form a separate object. The separation operation can be used to create complex combined models or extract specific parts of the model to obtain a spatial model of the feature. The spatial model of the feature is constructed by searching the image of the feature to find the corresponding object as the target object, and obtaining the model corresponding to the target object for use. At the same time, the position information of multiple features is obtained according to the state of multiple features in the spatial model. The position information here is the position of the feature in the spatial coordinate system in the spatial model. The spatial model of the feature is then replaced by the feature according to the position of the feature in the spatial coordinate system in the spatial model to obtain an environmental model, which can ensure the restoration of the image collected by the AR glasses, thereby improving the night vision ability of the AR glasses and having a good image processing effect;

[0066] In one embodiment, the step S2 of combining the image information with the environment model to obtain the image model includes:

[0067] S21, obtaining the real-time position of the AR glasses in real time, configuring a new space model according to the real-time position of the AR glasses, and combining the new space model with the environment model as a combined model;

[0068] S22, covering the combined model based on the spatial coordinate system in the environment model;

[0069] S23, supplementing the spatial features of the combined model and spatially locating the spatial features;

[0070] S24, updating the acquisition relationship between the AR glasses and the spatial features in the combined model in real time according to the real-time position of the AR glasses and the full-color image acquired by the AR glasses;

[0071] S25, using the established combined model of the acquisition relationship between the AR glasses and the spatial features as an image model;

[0072] As described in the above steps S22-S25, in actual use, since the AR glasses will constantly change, the collected night vision images are different, so the night vision images collected by the AR glasses need to be changed accordingly. Here, the real-time position of the AR glasses needs to be obtained first. The spatial model is a blank three-dimensional spatial model. Multiple spatial models need to be reserved. The corresponding spatial model needs to be configured according to the real-time position of the AR glasses. The configuration condition here is that when the real-time position change of the AR glasses exceeds the preset displacement, the configuration of a new spatial model is enabled, which means that the change displacement of the AR glasses is large and the combined model needs to be rebuilt. After that, the combined model has a large range. On the basis of the environmental model, the spatial The coordinate system covers the combined model, and then the entire combined model can be positioned. Then, the acquisition relationship between the AR glasses and the spatial features in the combined model is updated in real time according to the real-time position of the AR glasses and the full-color images collected by the AR glasses. After a new spatial model appears, it is necessary to use the construction method of the environmental model to update and construct the combined model with the full-color images collected by the AR glasses within the range of the spatial model. Then, the acquisition relationship between the AR glasses and the spatial features in the updated combined model is established. The combined model with the established acquisition relationship between the AR glasses and the spatial features is used as an image model to obtain the night vision image display effect of the AR glasses, so that the image can be displayed more clearly.

[0073] In one embodiment, the step S21 of acquiring the real-time position of the AR glasses in real time, configuring a new space model according to the real-time position of the AR glasses, and combining the new space model with the environment model as a combined model includes:

[0074] S211, obtaining in real time the horizontal displacement between the real-time position of the AR glasses and the center point of the previous environment model, and matching a new spatial model when the horizontal displacement exceeds a preset displacement;

[0075] S212, merging the overlapping parts of the new space model and the environment model, and reconstructing and combining the non-overlapping parts of the new space model and the environment model to obtain a combined model;

[0076] As described in the above steps S211 and S212, it is equivalent to configuring a new spatial model corresponding to each preset displacement representation, and the environmental model needs to be updated. Multiple spatial models need to be merged or overlapped with the previous environmental model. The specific combination method can be achieved through software such as Blender, Maya, 3ds Max, etc. The user can select the models to be merged and use the merging tools provided by the software to operate. For example, in Blender, multiple models can be merged into a single object. These software usually provide a wealth of adjustment and optimization options to help users obtain the ideal merging effect, so that a combined model can be obtained, which can facilitate the subsequent image integration processing of AR glasses and facilitate the display of night vision images;

[0077] In one embodiment, the step S23 of supplementing the combined model with spatial features and spatially locating the spatial features includes:

[0078] S231, obtaining a full-color image of the non-overlapping part of the new space model and the environment model, extracting multiple features in the full-color image and marking the corresponding spatial coordinate system in the combined model;

[0079] S232, spatially constructing the multiple features added in the combined model to obtain the spatial features added in the combined model, and placing the spatial features added in the combined model in place of the corresponding features in the spatial model;

[0080] S233, obtaining the center points of multiple spatial models in the combined model, respectively mapping the center points of the multiple spatial models downward to obtain root points, and establishing spatial positioning between the root points and corresponding spatial features;

[0081] As described in the above steps S231-S233, a full-color image of the non-overlapping part between the new spatial model and the previous environmental model is obtained here. The full-color image here is a full-color image of a newly added AR glasses acquisition range. Since the previous environmental model is acquired after the acquisition, only the full-color image of the newly added position needs to be acquired here, which can quickly improve the processing efficiency of the night vision imaging of the AR glasses without affecting the imaging quality. The construction of multiple features in the full-color image in the spatial model and the placement of the spatial features of the features in the spatial model are the same as those described in steps S111-S113. The only difference here is that one is to process the entire spatial model to obtain the night vision image, and the other is to process the spatial model of the non-overlapping part; a combination of the previous environmental model and the current environmental model In the operations of the combined model after combination, the image features in the combined model all exist in the form of spatial features. The spatial features represent the spatial model and spatial position of the features in the combined model. The spatial model is a three-dimensional model of the features, and the spatial position is the position of the spatial model in the combined model. The spatial position here corresponds to the position of the features actually collected. The projection transformation (such as perspective projection or orthogonal projection) is used to determine the downward mapping of the center point of the three-dimensional object model to obtain the root point, and then the spatial positioning between the root point and the corresponding spatial feature is established, which is equivalent to binding the information between the root point and the corresponding spatial feature. Here, as spatial positioning, it can facilitate the subsequent rapid and accurate presentation of the image in the image model collected by the AR glasses according to the viewing angle of the AR glasses, which can improve the imaging processing efficiency and imaging quality of the AR glasses.

[0082] In one embodiment, step S24 of updating the acquisition relationship between the AR glasses and the spatial features in the combined model in real time according to the real-time position of the AR glasses and the full-color image acquired by the AR glasses includes:

[0083] S241, configuring the root point with a corresponding sensing point, and connecting the root point with the AR glasses based on the sensing point;

[0084] S242, updating the connection line between the AR glasses and the spatial feature sensing point according to the real-time position of the AR glasses to obtain a collection relationship between the AR glasses and the spatial feature;

[0085] As described in the above steps S241 and S242, the root point is the point mapped downward from the center point of the spatial model, and the corresponding sensing point is configured corresponding to the root point. A wireless communication module matching the sensing point is integrated on the sensing point, such as Bluetooth, Wi-Fi, RFID reader, etc., and a positioning algorithm is developed or integrated to process the signal from the sensing point and calculate the position information of the device relative to the sensing point. Here, the connection is made through the communication connection between the AR glasses and the sensing point, and the connection existence is obtained through the connection between the AR glasses and the sensing point, wherein the sensing point is a temporarily configured position to determine the sensing point, which is used to record the position of the root point and can establish a connection with the AR glasses. The connection is displayed through the connection line. The AR glasses are generally changeable, so in the process of changing the AR glasses, the connection line between the AR glasses and the sensing point will also change. Since the image model is constructed in advance, the angle relationship between the connection line and the AR glasses can facilitate the subsequent rapid and accurate provision of the image in the image model collected by the AR glasses according to the viewing angle of the AR glasses, which can improve the imaging processing efficiency and imaging quality of the AR glasses;

[0086] In step S3, the image model is mapped to the viewing angle of the AR glasses to obtain a night vision image:

[0087] According to the angular relationship between the connecting line and the spatial feature, a plane display image of the spatial feature from the perspective of the connecting line is given;

[0088] Based on the perspective of the AR glasses (representing the angle of the connection line), a plane display image of the spatial features of the corresponding connection line can be obtained, and the plane display images of multiple spatial features under the perspective of the AR glasses are integrated within the acquisition range under the perspective of the AR glasses as a night vision image;

[0089] After determining the viewing angle of the AR glasses, a plane display image of the spatial features corresponding to the connecting lines under the viewing angle of the AR glasses is obtained. After acquiring the plane display image, the rendered two-dimensional image is further adjusted and optimized, such as color correction, contrast adjustment, sharpening, etc. to obtain the plane display image after processing. After determining the viewing angle of the AR glasses, it has a collection range under this viewing angle. Therefore, the plane display images corresponding to multiple spatial features under the collection range are integrated within the collection range according to the positions of the corresponding spatial features to obtain a night vision image.

[0090] Example 2, please refer to Figure 2 As shown, the wearable night vision system of AR glasses described in this embodiment includes:

[0091] An acquisition module, used for acquiring night vision images, extracting spatial features based on the night vision images, wherein the spatial features include feature information in the night vision images, and constructing an environment model based on the spatial features;

[0092] A construction module is connected to the acquisition module and is used to obtain the real-time position of the AR glasses, perform image understanding according to the real-time position corresponding to the environment model to obtain image information, and combine the image information with the environment model to obtain an image model;

[0093] The imaging module is connected to the building module and is used to match the image model to the viewing angle of the AR glasses to obtain a night vision image.

[0094] The acquisition module comprises:

[0095] A construction unit, used for constructing a spatial model and setting a spatial coordinate system on the spatial model;

[0096] The marking unit is used to restore the dark light image into a full-color image based on the image collected by the AR glasses through the AI-ISP technology, and extract multiple features in the full-color image and mark the corresponding spatial coordinate system in the spatial model;

[0097] A management unit is used for taking the state of a plurality of features in a spatial model as a night vision image.

[0098] The acquisition module also includes:

[0099] An extraction unit, used to extract features in the night vision image and construct feature information in the night vision image as a spatial feature, wherein the feature information is a spatial model and a spatial position of the feature in the night vision image;

[0100] The setting unit is used to replace the features in the night vision image with the corresponding spatial features and place them in the spatial model to obtain the environment model.

[0101] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. An AI-based image processing method, characterized in that: The following steps are involved: Acquire a night vision image, extract spatial features based on the night vision image, wherein the spatial features include feature information in the night vision image, and construct an environment model based on the spatial features; The step of acquiring night vision images comprises: Construct a spatial model and set a spatial coordinate system on the spatial model; Based on the images collected by AR glasses, the dark light images are restored to full-color images through AI-ISP technology, and multiple features in the full-color image are extracted and marked in the spatial coordinate system in the spatial model. The features are the feature subjects in the full-color image, and a feature subject represents an object image in the full-color image. The state of multiple features in the spatial model is taken as a night vision image; The step of constructing an environment model based on spatial features includes: Extracting features from night vision images to construct feature information in the night vision images as spatial features, wherein the feature information is a spatial model and a spatial position of the features in the night vision images, and the spatial model of the features is constructed by searching for corresponding objects as target objects according to the image of the features, and obtaining and using the model corresponding to the target object; The features in the night vision image are replaced by the corresponding spatial features and placed in the spatial model to obtain the environmental model; Obtain the real-time position of the AR glasses, perform image understanding based on the real-time position corresponding to the environmental model to obtain image information, and combine the image information with the environmental model to obtain an image model; The step of combining the image information with the environment model to obtain the image model comprises: The real-time position of the AR glasses is obtained in real time, a new space model is configured according to the real-time position of the AR glasses, and the new space model is combined with the environment model as a combined model; The step of updating the acquisition relationship between the AR glasses and the spatial features in the combined model in real time according to the real-time position of the AR glasses and the full-color image acquired by the AR glasses includes: Configure the root point with the corresponding sensing point, and connect the root point to the AR glasses based on the sensing point; According to the real-time position of the AR glasses, the connection line between the AR glasses and the spatial feature sensing point is updated to obtain the collection relationship between the AR glasses and the spatial feature; Overlaying the combined model based on the spatial coordinate system in the environment model; Supplement the spatial features of the combined model and spatially locate the spatial features; The step of supplementing the spatial features of the combined model and spatially locating the spatial features includes: Obtain a full-color image of the non-overlapping part of the new space model and the environment model, extract multiple features in the full-color image and mark the corresponding spatial coordinate system in the combined model; The multiple features added in the combined model are spatially constructed to obtain the spatial features added in the combined model, and the spatial features added in the combined model are placed in the spatial model instead of the corresponding features; Obtain the center points of multiple spatial models in the combined model, respectively map the center points of the multiple spatial models downward to obtain root points, and establish spatial positioning between the root points and corresponding spatial features; According to the real-time position of the AR glasses and the full-color images collected by the AR glasses, the acquisition relationship between the AR glasses and the spatial features in the combined model is updated in real time; The combined model of the acquisition relationship between the AR glasses and the spatial features is used as the image model; The image model is matched to the viewing angle of the AR glasses to obtain a night vision image.

2. The AI-based image processing method according to claim 1, characterized in that: The step of acquiring the real-time position of the AR glasses in real time, configuring a new space model according to the real-time position of the AR glasses, and combining the new space model with the environment model as a combined model includes: Obtain the horizontal displacement between the real-time position of the AR glasses and the center point of the previous environment model in real time, and match the new space model when the horizontal displacement exceeds the preset displacement; The overlapping parts of the new spatial model and the environmental model are merged, and the non-overlapping parts of the new spatial model and the environmental model are reconstructed and combined to obtain a combined model.

3. A wearable night vision system of AR glasses, used to implement an AI-based image processing method as described in any one of claims 1 and 2, characterized in that: include: An acquisition module, used for acquiring night vision images, extracting spatial features based on the night vision images, wherein the spatial features include feature information in the night vision images, and constructing an environment model based on the spatial features; A construction module is connected to the acquisition module and is used to obtain the real-time position of the AR glasses, perform image understanding according to the real-time position corresponding to the environment model to obtain image information, and combine the image information with the environment model to obtain an image model; The imaging module is connected to the building module and is used to match the image model to the viewing angle of the AR glasses to obtain a night vision image.

4. The wearable night vision system for AR glasses according to claim 3, characterized in that: The acquisition module comprises: A construction unit, used for constructing a spatial model and setting a spatial coordinate system on the spatial model; The marking unit is used to restore the dark light image into a full-color image based on the image collected by the AR glasses through the AI-ISP technology, extract multiple features in the full-color image and mark the corresponding spatial coordinate system in the spatial model; A management unit is used for taking the state of a plurality of features in a spatial model as a night vision image.

5. The wearable night vision system for AR glasses according to claim 3, characterized in that: The acquisition module also includes: An extraction unit, used to extract features in the night vision image and construct feature information in the night vision image as a spatial feature, wherein the feature information is a spatial model and a spatial position of the feature in the night vision image; The setting unit is used to replace the features in the night vision image with the corresponding spatial features and place them in the spatial model to obtain the environment model.

Images