SLAM (Simultaneous Localization and Mapping) method in dark light environment

Abstract

The invention discloses an SLAM method in a dark light environment, and the method comprises the steps: obtaining inertial feature information based on a feature extraction system; acquiring an image in a dark light environment, and acquiring a dark light image; processing the dark light image to obtain an exposure image; obtaining feature points in the exposure image to obtain image information; based on the image information, line segment features in the image are obtained; and based on the line segment features in the image and the inertial feature information, carrying out equipment positioning and space map construction. The problems that in the prior art, when an SLAM method in a dark light environment is used, mapping and positioning errors are too large, and the data processing efficiency is low are solved. Therefore, the mapping precision of the environment and the positioning precision of the equipment are comprehensively ensured, and the data processing efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of computer vision and robot positioning, and specifically relates to a SLAM method in a dark light environment. Background Art

[0002] SLAM technology has been widely used in many robotics fields due to its spatial mapping and device positioning capabilities. The core of this technology is to obtain images of the surrounding scene and then map the surrounding space based on accurate image information. However, in current technical applications, the main problems include: 1. Insufficient feature matching rate. Based on measured data, the feature matching rate of the traditional ORB-SLAM3 system dropped to 39.7% in a Lux < 5 environment, resulting in a cumulative positioning error rate of ±2.38m / 10m for the device; 2. Limited dynamic range. When extracting environmental information in a dark environment, the image must be exposed. In the current single-exposure imaging solution, 63.2% of over-exposed / under-exposed areas are generated in alternating light and dark scenes; 3. There is a bottleneck in real-time computing. Currently, a variety of methods have been developed based on SLAM technology. Taking the Mobile SLAM solution as an example, although a feature point screening strategy is used to obtain surrounding image information, the ARM Cortex-A53 platform can only achieve a processing speed of 12.7fps, which is not enough for the technical solutions used in industrial scenarios. The above three problems will cause the SLAM method to have too high an error rate in mapping and positioning, and it has gradually become unable to adapt to high-precision scenes.

[0003] In summary, how to establish a SLAM method in a dark environment to improve mapping and positioning accuracy is a technical problem that technical personnel in this field urgently need to solve. Summary of the Invention

[0004] In order to improve the mapping accuracy and device positioning accuracy in dark environments, this application proposes a SLAM method in dark environments, specifically including: A SLAM method in a dark environment, the method comprising: Based on the feature extraction system, inertial feature information is obtained; Acquire images in a dark environment, acquire dark-light images; Processing the dark light image to obtain an exposure image; Acquire feature points in the exposure image to obtain image information; Based on the image information, obtaining line segment features in the image; Based on the line segment features in the image and the inertial feature information, device positioning and spatial map construction are performed.

[0005] Optionally, the acquiring of inertial feature information based on a feature extraction system includes: Based on the feature extraction system set on the device, obtain the inertial feature information of the device; The inertial characteristic information includes device speed, movement deflection angle, and acceleration.

[0006] Optionally, acquiring an image in a dark light environment, acquiring a dark light image, includes: Set up a camera on the device and acquire continuous images; Based on the continuous images, an image in a dark light environment is acquired to obtain a dark light image.

[0007] Optionally, processing the dark-light image to obtain an exposure image includes: acquiring an illumination map of the dark-light image based on an illumination matrix; acquiring a preset exposure image based on the preprocessed illumination image; The preset exposure image is verified to obtain an exposure image.

[0008] Optionally, acquiring the illumination map of the dark-light image based on the illumination matrix includes: Get the lighting matrix of the space where the device is located. The determination equation of the lighting matrix is: ; in, Represents the lighting matrix; represents the input image; c Indicates contrast; R 、 G 、 B Represents the three primary color components of the pixel values ​​in the input dark-light image; Based on the illumination matrix, an illumination map of the dark light image is obtained. The illumination map of the dark light image is determined by the equation: ; in, represents the lighting diagram; Represents the filter coefficient; Represents the local pixel d Derivative along the dimension in the direction; d Indicates the directional dimension; h and v Indicates the vertical axis direction and the horizontal axis direction; I ( x , y ) represents any pixel on the image.

[0009] Optionally, acquiring a preset exposure image based on the preprocessed illumination image includes: Obtaining operating parameters of a dark-light image acquisition device, and obtaining a response model and a BTF model of the dark-light image acquisition device; The BTF model equation is: ; in, Indicates the exposure result; represents the basic exposure function; Represents the original image; Indicates exposure rate; β and c Is the exposure rate in the BTF model k Related parameters; in, The equation is: ; in, a and b Represents fixed camera parameters.

[0010] Optionally, verifying the preset exposure image to obtain the exposure image includes: The pixel average value of the regional image in the preset exposure image is obtained. The pixel average value equation of the regional image is: ; in, f ( x )and Mean ( x ) represents the pixel average value of the regional image; x Represents the pixel value of the pixel in the area image; i and j Represents the horizontal and vertical indexes of the region image respectively; N Indicates the total amount of vertical and horizontal indexes of the region image; I i，j ( x ) represents the area image of the dark-light image.

[0011] Based on the pixel average value, a global underexposure image in the dark light image is obtained. The determination equation of the global underexposure image is: ; in, Q ( x ) represents a globally underexposed image, T ( x ) represents the sum of the pixel values ​​of the regional points in the regional image.

[0012] Optionally, obtaining feature points in the exposure image to obtain image information includes: Obtaining exposure rates of all regional images in the exposure image; Based on the exposure rate, the information amount of the regional image is obtained, and the information amount determination equation is: ; in, p i Indicates the i The exposure rate of the image of each region; i The number index representing the region image; N The total number of indexes representing the region images; H ( R ) represents the image entropy of the exposure image; Indicates the maximum exposure rate of the exposed image, k Indicates exposure rate; Based on the information amount of the regional image, obtaining a regional image for feature extraction to obtain a feature extraction regional image; Decompose the feature extraction area image to obtain multiple decomposed images, and determine the decomposed images with illumination invariance based on an illumination invariant filter. The illumination invariant filter equation is: ; in, D iif Represents the Hamming distance between decomposed images; f iif Indicates that the distance between two decomposed images described by illumination invariance at different brightness is determined using the Hamming distance; s 2 Represents the variance between decomposed images; A decomposed image with an information amount higher than a preset information amount is obtained to obtain an image carrying information.

[0013] Optionally, acquiring line segment features in the image based on the image information includes: Acquiring image information from the image carrying information; Based on the EDLines algorithm, line segment features in the image information are obtained.

[0014] Optionally, performing device positioning and spatial map construction based on the line segment features in the image and the inertial feature information includes: Based on all the line segment features, obtaining the spatial relationship between the line segments; constructing a scene image of a dark light environment based on the spatial relationship between the line segments; constructing a spatial scene based on the scene image of the dark light environment; Obtaining the acquisition time of the dark-light image corresponding to the scene image of the dark-light image to obtain the image acquisition time point; Obtaining the extraction time of the inertial feature information to obtain the feature acquisition time point; Aligning the feature extraction system and the corresponding dark-light image on a time axis based on the image acquisition time point and the feature acquisition time point; The spatial position of the device is acquired based on the inertial feature information and a spatial coordinate system set on the spatial scene.

[0015] The beneficial effects of this application include: 1. Improved mapping accuracy. In the technical solution of this application, based on the dual exposure algorithm adopted, after the dark light image is exposed, the density of effective feature points is greatly improved. After actual testing, the density of feature points can be increased to 82.3% of the normal light level. On the basis of ensuring the number of effective feature points, when mapping is performed based on the obtained effective feature points, the accuracy of the obtained mapping results can be guaranteed to meet the application scenarios of high-precision and cutting-edge technologies.

[0016] 2. Improve the performance of illumination robustness feature extraction. In the technical solution of this application, a new exposure algorithm is used to expose the dark-light image collected in a dark-light environment. At the same time, based on the obtained exposure processing results, the feature points in the exposed image are determined. At this time, the extracted features can be considered similar to those obtained under normal lighting conditions, which can fully improve the illumination robustness features.

[0017] 3. Improve processing performance. In the technical solution of this application, after obtaining feature points during image processing, the EDLines algorithm is used to identify line segments in the image to improve scene construction efficiency. At the same time, based on the obtained construction results and combined with the inertial characteristics of the device, mapping and positioning are achieved simultaneously, improving the information processing performance of the entire system. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following is a brief introduction to the drawings required for the embodiments of the present application or the prior art. Obviously, the following description is only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. The drawings are used to provide a further understanding of the present disclosure and constitute part of the specification. Together with the following specific embodiments, they are used to explain the present disclosure, but do not constitute a limitation of the present disclosure. In the drawings: Figure 1A flow chart of a SLAM method in a dark environment provided by an embodiment of the present application; Figure 2 A schematic diagram of a SLAM system in a dark environment provided in an embodiment of the present application. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application. In addition, in the embodiments of the present application, "first", "second", etc. are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0020] Current SLAM methods have been widely used in industrial systems, but they still face some application challenges. First, information processing is weak in low-light environments, often leading to information loss. Once this problem occurs, accurate mapping becomes difficult to achieve, and the positioning performance of related equipment is also reduced, making it difficult to meet the requirements of industrial applications in all weather and all environments. Second, information processing efficiency is low. Both mapping and positioning functions require a large amount of information support and rapid response to various types of information. However, in current technical solutions, the image processing and feature extraction processes are inefficient, making it difficult to ensure that the obtained image information can be processed quickly, which is particularly contrary to the guarantee of device positioning accuracy. Finally, the feature extraction process is not robust enough. In industrial production, lighting changes are likely to occur. This requires that SLAM methods can obtain environmental and device location information regardless of the lighting environment. However, in current methods, it is easy to encounter problems when the lighting environment suddenly changes to low light, making it difficult to quickly identify and extract low-light environment information, resulting in insufficient robustness.

[0021] In order to solve the problems existing in the prior art, this application discloses a SLAM method in a dark environment, such as Figure 1 As shown in FIG, a flow chart of a SLAM method in a dark environment provided by an embodiment of the present application, specifically including: S110 : Acquire inertial feature information based on a feature extraction system.

[0022] S120: Acquire an image in a dark light environment, that is, acquire a dark light image.

[0023] S130: Process the dark light image to obtain an exposure image.

[0024] S140: Acquire feature points in the exposure image to obtain image information.

[0025] S150: Based on the image information, obtain line segment features in the image.

[0026] S160: Perform device positioning and spatial map construction based on the line segment features in the image and the inertial feature information.

[0027] The purpose of all the above steps is to construct a SLAM method and ensure that the obtained method can accurately and efficiently extract features from dark light patterns to achieve precise mapping and device positioning.

[0028] The following are the specific technical solutions for all the above steps: As described in step S110, the specific technical solution of this step includes: Based on the feature extraction system set on the device, obtain the inertial feature information of the device; The inertial characteristic information includes device speed, movement deflection angle, and acceleration.

[0029] Among them, for the feature extraction process, the entire SLAM algorithm framework runs on a movable single-camera device, which can be expanded to a device equipped with an inertial measurement unit.

[0030] Among them, based on the inertial measurement unit carried out, the inertial characteristic information of the current algorithm framework can be obtained.

[0031] As described in step S120, the specific technical solution of this step includes: Set up a camera on the device and acquire continuous images; Based on the continuous images, an image in a dark light environment is acquired to obtain a dark light image.

[0032] A camera is set on the mobile device, usually a monocular camera.

[0033] As described in step S130, the specific technical solution of this step includes: acquiring an illumination map of the dark-light image based on an illumination matrix; acquiring a preset exposure image based on the preprocessed illumination image; The preset exposure image is verified to obtain an exposure image.

[0034] The obtained image is acquired based on a monocular camera set on the device to obtain a dark light image.

[0035] The acquired dark light image is exposed to obtain an exposure image.

[0036] Among them, for the acquired exposure image, it is very likely that although it has been exposed, there are still bright areas and dark areas on the image. Among them, the dark areas usually carry less image information. Therefore, in the judgment of the scene, the dark areas are no longer identified, but are identified as non-information carrying areas. Then, only the information in the bright areas is obtained, thereby realizing the verification of the exposure pattern.

[0037] For all steps in step S130, specifically: S131, acquiring an illumination map of the dark-light image based on an illumination matrix, includes: Get the lighting matrix of the space where the device is located. The determination equation of the lighting matrix is: ; in, B ( x ) represents the lighting matrix; I c ( x ) represents the input image; c Indicates contrast; R 、 G 、 B Represents the three primary color components of the pixel values ​​in the input dark-light image; Based on the illumination matrix, an illumination map of the dark light image is obtained. The illumination map of the dark light image is determined by the equation: ; in, represents the lighting diagram; Represents the filter coefficient; Represents the local pixel d Derivative along the dimension in the direction; d Indicates the directional dimension; h and v Indicates the vertical axis direction and the horizontal axis direction; I ( x , y ) represents any pixel on the image.

[0038] The purpose of this step is to process the obtained dark light image. In obtaining the result, it is actually a dual exposure algorithm, so as to obtain the result based on the algorithm.

[0039] The dual exposure algorithm is used to perform dual exposure on low-light images captured by a monocular camera to restore them to their original normal state. This algorithm module simplifies the existing multiple-exposure framework and incorporates biological inspiration to propose a real-time dual exposure algorithm. The existing multiple-exposure framework consists of a sample generator, a sample sampler, and a fuser. The proposed dual exposure algorithm simplifies the existing sample generator by leveraging the fact that the human eye automatically adjusts image brightness by assigning weights to different brightness ranges. The existing sample generator generates multiple samples, which are then sampled by the sample sampler, discarding any that do not conform to the specified range. Finally, the sample fuser fuses these samples to produce the final image. This approach is complex in terms of its implementation and cannot meet the requirements for fast image processing. By leveraging the human biological tendency to automatically calculate weights for different brightness ranges, this algorithm simplifies the aforementioned process. The dual exposure algorithm no longer generates multiple samples of varying brightness, but instead directly calculates the brightness and uses this brightness to generate a single exposure sample. Finally, the fuser fuses these samples to produce the final image.

[0040] The specific formula of the dual exposure algorithm is: Overall formula: ; in, and Output and input images respectively. is the image fusion weight, is the exposure function.

[0041] From the overall formula, three parts determine the final result of the algorithm: (1) Grayscale dual exposure evaluator; (2) Grayscale dual exposure generator; (3) Grayscale dual sampler; The formula for the grayscale dual exposure estimator (1) is as follows: ; in T is the scene lighting evaluation map, m is a parameter that controls the degree of enhancement. m = 0, the output image is equal to the input image, that is, no enhancement is performed. m = 1, both underexposed and well-exposed pixels are enhanced. m When > 1, the pixels may be saturated and the output image may lose details. Therefore, the parameter can be adjusted m To obtain an enhanced image for a specific environment.

[0042] To obtain the illumination map of an image, it is usually necessary to solve the following optimization problem: ; in, B ( x ) represents the lighting matrix; I c ( x ) represents the input image; c Indicates contrast; R 、 G 、 B Represents the three primary color components of the pixel value in the input dark light image. The illuminated image needs to maintain the texture structure and remove the noise information in the image, so it is optimized by solving the following method T : ; First-order derivative filter Include (horizontal) and (vertical). M is the weight matrix, l is the coefficient. The first term of the equation makes the initial image L and refined images T The difference between the two is the smallest, while the second term remains T However, this dual exposure algorithm improves the original algorithm based on the characteristics of grayscale images. The specific formula is as follows: ; in B It is a matrix based on grayscale characteristics. This algorithm uses 0.5 as the standard for no difference brightness in [0,1]. is the input image after passing through the filter. To reduce the complexity, we use the following formula to approximate the above formula: ; Finally, this algorithm assumes that we have T Infinitely close to standard lighting. We get the lighting diagram T as follows: ; From the above formula, we can see that efficiency can be improved by changing the image size. Since changing the image scale does not change the brightness of the original image area, this application interpolates and resizes the image to 32*32.

[0043] The second step in step S130, namely the technical solution described in S132, includes: Obtaining operating parameters of a dark-light image acquisition device, and obtaining a response model and a BTF model of the dark-light image acquisition device; The BTF model equation is: ; in, Indicates the exposure result; represents the basic exposure function; Represents the original image; Indicates exposure rate; β and c is the exposure rate in the BTF model k Related parameters; in, The equation is: ; in, a and b Represents fixed camera parameters.

[0044] The third step in S130, namely, step S133, specifically includes: The pixel average value of the regional image in the preset exposure image is obtained. The pixel average value equation of the regional image is: ; in, f ( x )and Mean ( x ) represents the pixel average value of the regional image; x Represents the pixel value of the pixel in the area image; i and j Represents the horizontal and vertical indexes of the region image respectively; N Indicates the total amount of vertical and horizontal indexes of the region image; I i，j ( x ) represents the area image of the dark-light image.

[0045] Based on the pixel average value, a global underexposure image in the dark light image is obtained. The determination equation of the global underexposure image is: ; in, Q ( x ) represents a globally underexposed image, T ( x ) represents the sum of the pixel values ​​of the regional points in the regional image.

[0046] This section implements a dual exposure generator using a camera response model. The camera response model consists of two parts: a camera response function (CRF) model and a BTF model. While the parameters of the CRF model are determined solely by the camera, the parameters of the BTF model are determined by both the camera and the exposure. If pixel values ​​are linearly scaled before traditional gamma correction, the resulting image closely approximates a well-exposed true image. Therefore, the BTF model can be described using a two-parameter function: ; in β and c is the exposure rate in the BTF model k Since some camera manufacturers design the exposure as a gamma curve, it can perfectly fit these cameras. c = 1, the CRF model is a two-parameter function and the BTF model is a nonlinear function. Since BTF is nonlinear for most cameras, we mainly consider c = 1. In BTF g The solution is: ; Assume that no information about the camera is provided and use fixed camera parameters (a = -0.3293, b = 1.1258) that are suitable for most cameras.

[0047] The algorithm eliminates well-exposed pixels and obtains a globally underexposed image. Although a standard value of 0.5 for grayscale images can be used to determine whether the image is well-exposed, the histogram distribution of low-light images is mostly concentrated at the bottom, so using a hard threshold will lead to overexposure and slow down the program runtime. Therefore, this algorithm uses the pixel average of the regional image to determine whether the pixels in the region are well-exposed, as shown in the following formula: ; ; Among them, Q(x) is the evaluation function, f(x) is the average function of the pixel indications in the obtained area. After obtaining the underexposed area, the algorithm can provide more information to humans based on the fact that the visibility of well-exposed images is higher than that of underexposed images. Therefore, the optimal k should provide the maximum amount of information. To measure the amount of information and obtain results quickly, the algorithm uses one-dimensional image entropy to measure the information contained in the exposed image, which is defined as: ;

[0048] Since the image entropy first increases and then decreases with the increase of exposure rate, It can be solved by a one-dimensional minimizer. In addition, considering the real-time performance, Resize the image to 32*32.

[0049] As described in step S140, the purpose of this step is to further verify the acquired exposed image to find an image that carries more scene information, including: Obtaining exposure rates of all regional images in the exposure image; Based on the exposure rate, the information amount of the regional image is obtained, and the information amount determination equation is: ; in, p i Indicates the i The exposure rate of the image of each region; i The number index representing the region image; N The total number of indexes representing the region images; H ( R ) represents the image entropy of the exposure image; Indicates the maximum exposure rate of the exposed image. k Indicates exposure rate; Based on the information amount of the regional image, obtaining a regional image for feature extraction to obtain a feature extraction regional image; Decompose the feature extraction area image to obtain multiple decomposed images, and determine the decomposed images with illumination invariance based on an illumination invariant filter. The illumination invariant filter equation is: ; in, D iif Represents the Hamming distance between decomposed images; f iif Indicates that the distance between two decomposed images described by illumination invariance at different brightness is determined using the Hamming distance; s 2 Represents the variance between decomposed images; A decomposed image with an information amount higher than a preset information amount is obtained to obtain an image carrying information.

[0050] This filter algorithm is used to solve the problem of feature point tracking. The currently popular optical flow method uses the invariance of local brightness to track feature points, but the image repaired by the exposure algorithm will have brightness inconsistencies. The illumination invariance filter enhances the robustness of the optical flow method by filtering out points with varying brightness. The filter's working algorithm is as follows: it primarily calculates the feature point's descriptor to obtain the feature's illumination characteristics. If the feature point has the illumination invariance property, it is retained; if not, it is removed and subsequently not tracked using the optical flow method.

[0051] The mathematical expression and proof process of the illumination invariant filter are as follows: ; ; in, yes Variance of the region (here we choose the size of 8*8). The Hamming distance is used to determine the distance between two identical regions described by brightness invariance at different brightness.

[0052] Proof: D iif Marked as D and .Just when and It can be proved that ; for D iif , the above formula can be written as: ; is a constant, as shown below: ; because , ; As described in step S150, after obtaining the verified exposure image, key information in the surrounding scene image can be obtained based on the exposure image. Therefore, it is necessary to obtain feature information based on this method, including: Acquiring image information from the image carrying information; Based on the EDLines algorithm, line segment features in the image information are obtained.

[0053] This section utilizes the fuzzy-controlled EDLines algorithm, which addresses the difficulty of detecting line features in low-light conditions. The EDLines algorithm has a large number of adjustable hyperparameters that can be adjusted to alter its performance in specific environments. However, these adjustments often rely on empirical knowledge and are not adaptive. This algorithm abstracts specific phenomena into fuzzy laws and uses these laws to adjust hyperparameters to achieve adaptability. The specific workflow involves extracting line features from an input 2D image using the fuzzy-controlled EDLines algorithm.

[0054] The regulation library of fuzzy control is as follows: ; The threshold formula is as follows: ; Using the detected line numbers, the FIS rule base is defined in Table 1. The EDLines edge detector threshold is determined using the number of feature points detected in the feature line segment detection module. When the number of detected lines is higher / lower / closer to the expected value, the EDLines edge detector threshold is adjusted accordingly.

[0055] As described in step S160, after acquiring an image of the spatial environment in which the device is located and extracting feature information from the image, it is necessary to perform mapping and position determination based on the acquired image and the inertial information of the device, including: Based on all the line segment features, obtaining the spatial relationship between the line segments; constructing a scene image of a dark light environment based on the spatial relationship between the line segments; constructing a spatial scene based on the scene image of the dark light environment; Obtaining the acquisition time of the dark-light image corresponding to the scene image of the dark-light image to obtain the image acquisition time point; Obtaining the extraction time of the inertial feature information to obtain the feature acquisition time point; Aligning the feature extraction system and the corresponding dark-light image on a time axis based on the image acquisition time point and the feature acquisition time point; The spatial position of the device is acquired based on the inertial feature information and a spatial coordinate system set on the spatial scene.

[0056] Here, on the basis of having acquired the line segment features, the association relationship between each line segment is acquired, and based on the association relationship, the environmental space information is acquired and the mapping process is performed.

[0057] Among them, for the positioning operation of the device, based on the inertial information of the device, the inertial characteristic parameters generated when it moves at different positions in the entire scene are determined, and then the position information is obtained based on the inertial characteristic parameters to perform device positioning.

[0058] For all the above technical solutions, such as Figure 2 , which is a schematic diagram of a SLAM system in a dark environment provided by an embodiment of the present application.

[0059] The beneficial effects of this application include: 1. Improved mapping accuracy. In the technical solution of this application, based on the dual exposure algorithm adopted, after the dark light image is exposed, the density of effective feature points is greatly improved. After actual testing, the density of feature points can be increased to 82.3% of the normal light level. On the basis of ensuring the number of effective feature points, when mapping is performed based on the obtained effective feature points, the accuracy of the obtained mapping results can be guaranteed to meet the application scenarios of high-precision and cutting-edge technologies.

[0060] 2. Improve the performance of illumination robustness feature extraction. In the technical solution of this application, a new exposure algorithm is used to expose the dark-light image collected in a dark-light environment. At the same time, based on the obtained exposure processing results, the feature points in the exposed image are determined. At this time, the extracted features can be considered similar to those obtained under normal lighting conditions, which can fully improve the illumination robustness features.

[0061] 3. Improve processing performance. In the technical solution of this application, after obtaining feature points during image processing, the EDLines algorithm is used to identify line segments in the image to improve scene construction efficiency. At the same time, based on the obtained construction results and combined with the inertial characteristics of the device, mapping and positioning are achieved simultaneously, improving the information processing performance of the entire system.

[0062] Those skilled in the art will understand that all or part of the steps of the above-mentioned method embodiments can be implemented by hardware related to computer program instructions, and the aforementioned computer program can be stored in a non-volatile storage medium. When the computer program is executed, it executes the steps of the above-mentioned method embodiments. Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a non-volatile storage medium. Based on this understanding, the technical solution of the embodiment of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a non-volatile storage medium and includes a number of instructions for enabling an electronic device (which can be a personal computer, server, network device, etc.) to execute all or part of the methods described in each embodiment of the present invention.

[0063] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed by the present invention, which should be covered by the scope of protection of the present invention.

Claims

1. A SLAM method in a dark environment, the method comprising: Based on the feature extraction system, inertial feature information is obtained; Acquire images in a dark environment, acquire dark-light images; Processing the dark light image to obtain an exposure image; Acquire feature points in the exposure image to obtain image information; Based on the image information, obtaining line segment features in the image; Based on the line segment features in the image and the inertial feature information, device positioning and spatial map construction are performed.

2. The SLAM method in a dark environment according to claim 1, wherein: The inertial feature information is obtained based on the feature extraction system, including: Based on the feature extraction system set on the device, obtain the inertial feature information of the device; The inertial characteristic information includes device speed, movement deflection angle, and acceleration.

3. The SLAM method in a dark environment according to claim 1, wherein: The acquiring of an image in a dark light environment comprises: Set up a camera on the device and acquire continuous images; Based on the continuous images, an image in a dark light environment is acquired to obtain a dark light image.

4. The SLAM method in a dark environment according to claim 1, wherein: The step of processing the dark light image to obtain an exposure image includes: acquiring an illumination map of the dark-light image based on an illumination matrix; acquiring a preset exposure image based on the preprocessed illumination image; The preset exposure image is verified to obtain an exposure image.

5. The SLAM method in a dark environment according to claim 4, characterized in that: The step of obtaining an illumination map of the dark-light image based on the illumination matrix includes: Get the lighting matrix of the space where the device is located. The determination equation of the lighting matrix is: ； in, Represents the lighting matrix; represents the input image; c Indicates contrast; R 、 G 、 B Represents the three primary color components of the pixel values ​​in the input dark-light image; Based on the illumination matrix, an illumination map of the dark light image is obtained. The illumination map of the dark light image is determined by the equation: ； in, represents the lighting diagram; Represents the filter coefficient; Represents the local pixel d Derivative along the dimension in the direction; d Indicates the directional dimension; h and v Indicates the vertical axis direction and the horizontal axis direction; I ( x , y ) represents any pixel on the image.

6. The SLAM method in a dark environment according to claim 4, characterized in that: The step of acquiring a preset exposure image based on the preprocessed illumination image includes: Obtaining operating parameters of a dark-light image acquisition device, and obtaining a response model and a BTF model of the dark-light image acquisition device; The BTF model equation is: ； in, Indicates the exposure result; represents the basic exposure function; Represents the original image; Indicates exposure rate; β and γ Is the exposure rate in the BTF model k Related parameters; in, The equation is: ； in, a and b Represents fixed camera parameters.

7. The SLAM method in a dark environment according to claim 4, characterized in that: The verifying the preset exposure image to obtain the exposure image includes: The pixel average value of the regional image in the preset exposure image is obtained. The pixel average value equation of the regional image is: ； in, f ( x )and Mean ( x ) represents the pixel average value of the regional image; x Represents the pixel value of the pixel in the area image; i and j Represents the horizontal and vertical indexes of the region image respectively; N Indicates the total amount of vertical and horizontal indexes of the region image; I i，j ( x ) represents the area image of the dark light image; Based on the pixel average value, a global underexposure image in the dark light image is obtained. The determination equation of the global underexposure image is: ； in, Q ( x ) represents a globally underexposed image, T ( x ) represents the sum of the pixel values ​​of the regional points in the regional image.

8. The SLAM method in a dark environment according to claim 1, characterized in that: The acquiring feature points in the exposure image to obtain image information includes: Obtaining exposure rates of all regional images in the exposure image; Based on the exposure rate, the information amount of the regional image is obtained, and the information amount determination equation is: ； in, p i Indicates the i The exposure rate of the image of each region; i The number index representing the region image; N The total number of indexes representing the region images; H ( R ) represents the image entropy of the exposure image; Indicates the maximum exposure rate of the exposed image, k Indicates exposure rate; Based on the information amount of the regional image, obtaining a regional image for feature extraction to obtain a feature extraction regional image; Decompose the feature extraction area image to obtain multiple decomposed images, and determine the decomposed images with illumination invariance based on an illumination invariant filter. The illumination invariant filter equation is: ； in, D iif Represents the Hamming distance between decomposed images; f iif Indicates that the distance between two decomposed images described by illumination invariance at different brightness is determined using the Hamming distance; σ 2 Represents the variance between decomposed images; A decomposed image with an information amount higher than a preset information amount is obtained to obtain an image carrying information.

9. The SLAM method in a dark environment according to claim 1, characterized in that: The acquiring line segment features in the image based on the image information includes: Acquiring image information from the image carrying information; Based on the EDLines algorithm, line segment features in the image information are obtained.

10. The SLAM method in a dark environment according to claim 1, characterized in that: The positioning of the device and construction of a spatial map based on the line segment features in the image and the inertial feature information includes: Based on all the line segment features, obtaining the spatial relationship between the line segments; constructing a scene image of a dark light environment based on the spatial relationship between the line segments; constructing a spatial scene based on the scene image of the dark light environment; Obtaining the acquisition time of the dark-light image corresponding to the scene image of the dark-light image to obtain the image acquisition time point; Obtaining the extraction time of the inertial feature information to obtain the feature acquisition time point; Aligning the feature extraction system and the corresponding dark-light image on a time axis based on the image acquisition time point and the feature acquisition time point; The spatial position of the device is acquired based on the inertial feature information and a spatial coordinate system set on the spatial scene.

Images