A table and chair remote management method based on a smart home platform

Abstract

The application provides a table and chair remote management method based on a smart home platform, comprising: establishing a table and chair three-dimensional model, obtaining key structure points and outer contour line coordinate information of the table and chair by analyzing the three-dimensional model; analyzing a home space image by using a computer vision algorithm, extracting main visual elements in the home space, judging spatial relationships and visual coordination between the visual elements, and generating a home space visual feature map; dynamically adjusting the size, shape and color of the hidden contour of the table and chair according to the calculated optimal position and shape parameters, so that the hidden contour is coordinated with the surrounding environment, and an initial design scheme of the hidden contour is generated; and transmitting the optimized hidden contour design scheme parameters to a smart control system, training by a machine learning algorithm in combination with historical design scheme data, and realizing dynamic adaptation of the hidden contour of the table and chair to the surrounding environment.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of information technology, and in particular to a table and chair remote management method based on a smart home platform. BACKGROUND

[0002] In a smart home environment, the core technical problem faced by table and chair design is how to achieve visual integration with the surrounding environment while ensuring functionality. In specific business scenarios, traditional table and chair design often focuses on functionality and structural stability, while ignoring coordination with the smart home environment, resulting in a visual sense of incongruity. In addition, due to the diversity and complexity of the home environment, the outline of the table and chair will differ when hidden in different locations. Therefore, how to dynamically adjust the shape, size, color, and other parameters of the hidden outline according to the specific location of the table and chair to adapt to different environments in real time is a technical difficulty that needs to be overcome. At the same time, the design of the hidden outline also needs to consider the morphological characteristics and structural features of the table and chair, ensuring that the normal use function of the table and chair is not affected while ensuring the hiding effect. SUMMARY

[0003] The present application provides a table and chair remote management method based on a smart home platform, mainly comprising:

[0004] Obtain the surrounding environment information of the location where the table and chair are located, including the color and material attribute data of the wall and floor environment elements, determine the initial color and material parameters of the table and chair hidden outline according to the obtained environment color and material attribute data, including hue, saturation, lightness, and surface texture characteristics;

[0005] Establish a three-dimensional model of the table and chair, obtain the key structural points and outer contour line coordinate information of the table and chair by analyzing the three-dimensional model;

[0006] Analyze the home space image using computer vision algorithms, extract the main visual elements in the home space, judge the spatial relationship and visual coordination between the visual elements, and generate a home space visual feature map;

[0007] Match the key structural points and outer contour line coordinate information of the table and chair three-dimensional model with the home space visual feature map, calculate the best position and shape parameters of the table and chair hidden outline in the home space through a space mapping algorithm;

[0008] According to the calculated best position and shape parameters, dynamically adjust the size, shape, and color of the table and chair hidden outline to make it coordinate with the surrounding environment, and generate an initial design scheme of the hidden outline;

[0009] Adopt the physical simulation method to simulate the movement process of the table and chair, judge whether the hidden contour design affects the normal use function of the table and chair, optimize the hidden contour design according to the simulation result, and generate an optimized hidden contour design scheme;

[0010] The optimized hidden contour design scheme parameters are transmitted to the intelligent control system, combined with historical design scheme data, and trained through a machine learning algorithm to realize dynamic adaptation of the table and chair hidden contour to the surrounding environment.

[0011] The technical scheme provided by the embodiment of the present application can include the following beneficial effects:

[0012] The present application discloses a kind of table and chair remote management method based on smart home platform.The method first obtains the color and material data of the environment around table and chair, establishes table and chair three-dimensional model and extracts key structure information.Then the home space image is analyzed by computer vision algorithm, and visual feature map is generated.The table and chair model is matched with visual feature map, and the best position and shape parameters of hidden contour are calculated.According to the calculation result, the design of hidden contour is dynamically adjusted, and the design scheme is optimized by physical simulation.Finally, the optimized design scheme is transmitted to the intelligent control system by the present application, and the dynamic adaptation of hidden contour and environment is realized by machine learning.The technical effect of the present application is to automatically generate and optimize the hidden contour design of table and chair according to the surrounding environment, so that table and chair can better integrate into home space, improve the visual coordination and aesthetic degree of space.Meanwhile, the normal use function of table and chair is guaranteed, and the dynamic balance of appearance and function of table and chair in smart home environment is realized. BRIEF DESCRIPTION OF DRAWINGS

[0013] Fig. 1 It is a flow chart of the table and chair remote management method based on smart home platform of the present application.

[0014] Fig. 2 It is a schematic diagram of the table and chair remote management method based on smart home platform of the present application.

[0015] Fig. 3 It is another schematic diagram of the table and chair remote management method based on smart home platform of the present application. DETAILED DESCRIPTION

[0016] In order to enable the personnel in the technical field to better understand the technical solutions in the present specification, the technical solutions in the present specification will be described clearly and completely in the present specification, combined with the drawings in the present specification, obviously, the described embodiments are only part of the embodiments of the present specification, not all the embodiments. Based on the embodiments in the present specification, all other embodiments obtained by the personnel in the field without making creative efforts should belong to the protection scope of the present specification.

[0017] As Figs. 1-3 The table and chair remote management method based on the smart home platform can specifically include the following steps.

[0018] In S101, the surrounding environment information of the position where the table and chair are located is acquired, including the color and material attribute data of the wall and floor environment elements. According to the acquired environment color and material attribute data, the initial color and material parameters of the table and chair hidden contour are determined, including hue, saturation, lightness and surface texture characteristics.

[0019] The wall surface area is scanned by using a color sensor array. The wall surface main tone value is obtained from the acquired wall surface color distribution data. The wall surface depth distribution map is obtained by using a material texture analyzer. According to the wall surface main tone value, the wall surface reflection degree is measured by using a ray tracing model. The indoor illumination intensity value is obtained from a luminosity calculation unit. The illumination intensity value is quantitatively processed to obtain an illumination intensity distribution map. The environment overall brightness distribution map is obtained by processing the illumination intensity distribution map and the wall surface depth distribution map through convolution operation. A three-dimensional space brightness model is established for the environment overall brightness distribution map. The surface of the table and chair is three-dimensionally scanned and modeled by using a depth camera. The illumination parameter calculation of the table and chair contour area is performed according to the three-dimensional space brightness model. The table and chair contour area illumination parameter set is obtained.

[0020] Specifically, for the wall environment around the position where the table and chair are located, a color sensor array is used to scan the wall area, and the main tone value of the wall is obtained from the color distribution data of the wall environment by color space quantization processing. The surface layer of the wall is scanned by a material texture analyzer to obtain a depth distribution map. According to the main tone value of the wall, the light reflection degree of the wall is measured by a ray tracing model, and the indoor light intensity value is obtained from a luminosity calculation unit to quantitatively process the environmental illumination and obtain a light intensity distribution map. According to the light intensity distribution map and the wall depth distribution map, the overall brightness distribution map of the environment is obtained by convolution operation, and a three-dimensional space brightness model is established for the overall brightness distribution map of the environment. The surface of the table and chair is scanned by a depth camera to establish a three-dimensional model, and the reflected light data of the table and chair surface is obtained by a ray tracing model, and the light parameters of the contour area of the table and chair are calculated by combining the three-dimensional space brightness model. For the light parameters of the contour area of the table and chair, a color space mapping algorithm is used for hue quantization calculation, and the highest matching hue parameter is extracted from the preset color database. The surface of the table and chair is scanned by a material texture analyzer to obtain a surface texture feature vector, and the main texture feature value is obtained by principal component analysis of the texture feature vector. According to the hue parameter and the main texture feature value, the highest similarity texture parameter combination is retrieved from the preset material database to generate an initial color and material parameter set for the hidden contour of the table and chair. In the process of obtaining the environment information of the table and chair, the color sensor array scans the wall through photoelectric conversion principle, and each sensor unit collects spectral data. The spectral data is converted into digital signals through color space quantization processing. For example, a certain sensor unit collects visible light spectrum with wavelength range of 380-780 nanometers, and converts to get RGB three-channel values of (200, 150, 100). Through color space conversion, the hue value in HSV color space is 25 degrees, the saturation is 0.5, and the brightness is 0.78. The material texture analyzer uses structured light projection method to project specific stripe pattern to the wall, and obtains the surface depth distribution by analyzing the deformation degree of the stripes. Assuming that 20 equally spaced parallel stripes are projected on a 1 square meter wall area, each stripe width is 2 millimeters, and the center line position offset of the stripes is extracted by a high-resolution camera to calculate the depth value in the range of 0-255. The ray tracing model is based on the principle of geometric optics, and uses the Monte Carlo method to simulate the light propagation path. In a 5m x 4m x 3m indoor space, 500 light emitting points are set, each emitting 1000 rays, and the intersection coordinates and incident angles of the rays with the wall are recorded. The luminosity calculation unit measures the indoor light intensity between 200 and 800 lux, and calculates the illumination value of each point on the wall according to the incident angle of the light. The three-dimensional point cloud data of the table and chair collected by the depth camera is registered to establish a table and chair surface grid model. The grid model contains 10000 vertices, and each vertex records the spatial coordinates and normal vector information.The environmental light illumination data is projected to the surface of the grid model, and the illumination parameters at each vertex, including the diffuse reflection coefficient and the specular reflection coefficient, are calculated. The texture feature extraction is performed on the surface of the table and chair, and the gray level co-occurrence matrix method is used to calculate the texture statistical features. In a 5x5 pixel window, the gray value relationship of pixel pairs is counted, and the feature values such as contrast, correlation, energy and homogeneity are calculated. Through principal component analysis, the 16-dimensional texture feature vector is reduced to a 4-dimensional principal component space, retaining more than 90% of the feature information. Finally, the best matching item in the preset database is retrieved according to the hue parameter and the texture feature value, and the database contains the parameter templates of 1000 common materials, each template containing three color parameters of hue, saturation and lightness, and three texture parameters of contrast, directionality and roughness. The similarity matching is performed by calculating the Euclidean distance, and the parameter combination with the smallest distance is selected as the initial parameter of the hidden contour of the table and chair.

[0021] S102, a three-dimensional model of the table and chair is established, and by analyzing the three-dimensional model, the key structural points and the outer contour line coordinate information of the table and chair are obtained.

[0022] A laser scanner is used to obtain a set of point cloud data of the surface of the table and chair, and the point cloud data set is fused by a point cloud registration algorithm to obtain a complete point cloud model. According to the complete point cloud model, a triangular mesh is divided by a Poisson reconstruction algorithm, and the vertex coordinates of the mesh are optimized by a Laplace smoothing method to obtain a smooth mesh model. The curvature values of the vertexes of the smooth mesh model are extracted by using a Gaussian curvature calculation formula, and a set of feature points is identified by setting a curvature threshold. According to the set of feature points, an edge detection is performed by using a sobel operator to obtain a binary image, and a sequence of key corner coordinates is extracted from the binary image.

[0023] Specifically, the laser scanner is used to obtain the point cloud data of the table and chair surface, the spatial coordinate information is extracted from the original point cloud data, the multi-angle scanning data is registered and fused by the point cloud registration algorithm, and the complete point cloud data set of the table and chair is obtained. According to the point cloud data set, a grid topology structure is constructed, a triangular mesh is divided for the point cloud by using the Poisson reconstruction algorithm, and the vertex coordinates of the mesh are optimized by the Laplace smoothing method to obtain a smooth and continuous surface mesh model of the table and chair. For the vertex of the mesh model, the Gaussian curvature calculation formula is used to extract the curvature value, the feature point set is identified by setting the curvature threshold, and the distribution map of the feature area of the table and chair surface is obtained. According to the feature area distribution map, the sobel operator is used for edge detection, the detection result is binarized, and the continuous edge line segment is extracted by morphological operation. For the edge line segment, the douglas-peucker algorithm is used to extract the key inflection point coordinates, and the least square method is used to fit the straight line segment for the inflection point sequence. According to the fitted straight line segment, the cubic spline interpolation method is used to generate a continuous and smooth contour line, and the contour line coordinate sequence is obtained by uniform sampling. The spatial coordinate transformation is performed on the contour line coordinate sequence, the corresponding relationship of the local coordinate system is established, and the coordinate data is stored in the structured database. In the three-dimensional modeling of the table and chair, the laser scanner emits a laser beam with a wavelength of 905 nanometers, and the distance information of the space point is obtained by measuring the reflection time of the laser. A standard office chair is scanned at 360 degrees, and sampling is performed every 15 degrees, obtaining 24 groups of point cloud data, each containing about 50000 spatial point coordinates. These original point cloud data are registered by the iterative closest point algorithm, the registration error is controlled within 0.5 millimeters, and finally a complete point cloud data set containing about 1 million points is formed. In the point cloud reconstruction process, the Poisson reconstruction algorithm first calculates the normal vector field of the point cloud, and estimates the local tangent plane for each point. By solving the Poisson equation ▽2χ=▽·V, where V is the vector field and χ is the indicator function, the implicit function expression of the object surface is obtained. The algorithm sets the octree depth to 8, and the reconstructed mesh model contains about 200,000 triangular facets. The mesh is processed by Laplace smoothing, and the vertex coordinates are updated according to the weighted average of the surrounding vertices in each iteration, and the weight is inversely proportional to the edge length. The surface curvature analysis uses the Gaussian curvature calculation formula K=κ1κ2, where κ1 and κ2 are the principal curvatures. For each vertex on the mesh model, a quadratic surface z=ax2+by2+cxy is fitted in a 3-ring neighborhood, the coefficients a, b, and c are obtained by solving the least squares equation system, and the Gaussian curvature value of the point is calculated. Set the curvature threshold to 0.05, extract the points with absolute curvature greater than the threshold as feature points, and obtain a set of about 5000 feature points. The edge detection uses a 3x3 sobel operator to calculate the gradient in the horizontal and vertical directions. Set the gradient amplitude threshold to 128, mark the points greater than the threshold as edge points, and remove noise by morphological opening operation to obtain continuous edge line segments.The edge line segment is simplified by using the douglas-peucker algorithm, a distance threshold of 2 millimeters is set, and about 200 key inflection points are reserved. The inflection point sequence is segmented, and when the distance between adjacent inflection points is greater than 10 millimeters, a new line segment is divided. The least square method is used to fit the straight line equation ax+by+c=0, and the fitting error is controlled within 1 millimeter. A smooth contour line is generated by cubic spline interpolation, the interpolation point spacing is set to 5 millimeters, and the first and second derivatives of the curve are ensured to be continuous at the nodes. Finally, a local coordinate system with the seat center as the origin is established, and the spatial coordinates of the contour line sampling points are stored in the structured database after coordinate transformation.

[0024] S103, using computer vision algorithm to analyze the home space image, extracting the main visual elements in the home space, judging the spatial relationship and visual coordination between the visual elements, generating the home space visual feature map.

[0025] The original image of the home space is obtained by using a depth camera, the original image is processed by a Gaussian filter and gamma correction to obtain an enhanced image; the gray level co-occurrence matrix texture feature, color moment color feature and gradient operator edge feature are extracted from the enhanced image to obtain a feature vector; the region growing segmentation is performed on the feature vector to obtain a visual element label map; the visual element centroid coordinates are calculated for the visual element label map, the Euclidean distance method is used to construct a spatial distance matrix, the color histogram intersection is used to calculate the color similarity of the visual elements to obtain a visual coordination degree matrix; the spectral clustering method is used to hierarchically divide the visual elements according to the visual coordination degree matrix, the spatial division structure is constructed by the quadtree method, and the brightness weight calculation is used to generate a visual attention distribution map to obtain the home space visual feature map.

[0026] Specifically, a depth camera is used to acquire images of a home space, a Gaussian filter is used to denoise the original images, gamma correction and histogram equalization are performed on the denoised images to obtain enhanced images. Feature extraction is performed on the enhanced images, texture features are calculated using a gray level co-occurrence matrix, color features are extracted using a color moment method, and edge features are calculated using a gradient operator to obtain an image feature vector. For the image feature vector, a similarity threshold is set for region growing segmentation, a convolutional neural network is used to recognize visual elements such as tables, chairs, cabinets, sofas, and walls in the home space, and a visual element label map is obtained. The visual element centroid coordinates are calculated based on the visual element label map, and a spatial distance matrix is constructed using the Euclidean distance method, and a spatial adjacency relationship is established for points with a distance less than a predetermined threshold. For the visual element region, color features are extracted using a color histogram, color similarity is calculated by calculating the intersection of the histograms, and a visual coordination degree matrix is obtained by combining the spatial adjacency relationship. Based on the visual coordination degree matrix, a spectral clustering method is used to hierarchically divide the visual elements, and the spatial arrangement rule is determined by principal direction calculation. For the hierarchical division result of the visual elements, a quadtree method is used to construct a spatial division structure, a brightness weight calculation is performed to generate a visual attention distribution map, and a home space visual feature map is obtained. The depth camera uses a structured light projection principle to acquire RGB images and depth maps of the home scene, the image resolution is 1920x1080 pixels, and the depth accuracy is ±1mm. The RGB image is filtered by a 5x5 Gaussian kernel, the standard deviation is set to 1.5, and the random noise is removed. The filtered image is gamma corrected, the gamma value is set to 0.8, and the dark details are improved. The image gray value is remapped to the range of 0-255 using histogram equalization to enhance the image contrast. In the feature extraction stage, the displacement vector of the gray level co-occurrence matrix is set to (0,1), (1,0), and (1,1), the co-occurrence frequency of pixel pairs in the 0-255 gray level is calculated, and four texture features including contrast, correlation, energy, and homogeneity are calculated. The color moment is calculated for the image in the RGB space, including the mean moment, standard deviation moment, and skewness moment, to obtain a 9-dimensional color feature vector. The sobel operator is used to calculate the horizontal and vertical direction gradients of the image, and an 8-dimensional edge feature vector is constructed by the gradient amplitude and direction. The region growing segmentation takes the center region of the image as the seed point, and the growth criterion is that the Euclidean distance of the feature vector is less than the threshold value 10, and the adjacent pixels are iteratively merged. The convolutional neural network uses a pre-trained model to classify the segmented regions, and the confidence threshold is set to 0.85 to obtain visual element regions labeled with class labels. The centroid coordinates of each visual element region are calculated, and an N*N distance matrix is constructed, where N is the number of visual elements. The spatial adjacency threshold is set to 0.2 times the diagonal line length of the image, and an adjacency relationship is established when the centroid distance of two visual elements is less than the threshold. A 32-dimensional color histogram is calculated for each visual element region, the color similarity between each other is calculated by histogram intersection, and a visual coordination degree matrix is obtained by combining the adjacency relationship.The spectral clustering method is used to decompose the coordination matrix, the first three eigenvectors are taken to construct the feature space, and the visual elements are divided into three levels by the k-means algorithm. The main direction of the visual elements in each level is calculated, the Hough transform is used to detect the straight line features, and the spatial arrangement rules are extracted. Finally, the image is recursively divided into 16x16 grids according to the quadtree method, the average brightness value of each grid is calculated as the weight, a 256x256 resolution visual attention distribution map is generated, and the final visual feature map is obtained by superimposing the hierarchical division results.

[0027] In S104, the key structure points and the outer contour line coordinate information of the table and chair three-dimensional model are matched with the home space visual feature map, and the best position and shape parameters of the hidden contour of the table and chair in the home space are calculated through a space mapping algorithm.

[0028] The key structure points of the table and chair three-dimensional model are projected to a two-dimensional plane by a perspective projection method, and a set of normalized plane coordinate points is calculated through a homography matrix. Feature descriptors are obtained by a surf feature point extraction algorithm according to the set of normalized plane coordinate points, and a set of corresponding relationships of feature points is obtained by a nearest neighbor matching method. A least squares optimization objective function is established for the set of corresponding relationships of feature points, and the initial placement position of the table and chair is obtained by iterative optimization of the position coordinates of the table and chair using a gradient descent method. According to the initial placement position of the table and chair, contour fitting is performed through a cubic Bezier curve, the interval ratio of the curve control points and the tangent direction constraint are set, and a contour shape feature vector is obtained by calculating the curvature value of the sampling points.

[0029] Specifically, the key structure points and the outer contour line coordinates of the table and chair three-dimensional model are converted in the coordinate system, the three-dimensional space coordinates are projected to a two-dimensional plane by using the perspective projection method, the projected coordinates are affine transformed by using the homography matrix, and the standardized plane coordinate point set is obtained. According to the home space visual feature map, the surf feature point extraction algorithm is used to obtain the feature descriptor, the contour points of the table and chair are matched with the space feature points by using the nearest neighbor matching method, and the feature point corresponding relationship set is obtained. According to the feature point corresponding relationship set, the least square optimization objective function is established, the gradient descent method is used to iteratively optimize the position coordinates and size ratio of the table and chair, and the initial placement position of the table and chair is obtained. For the initial placement position of the table and chair, the contour is fitted by using the cubic Bezier curve, the interval ratio of the curve control points and the tangent direction constraint are set, and the continuous and smooth contour curve is obtained. According to the smooth contour curve, the discrete sampling point sequence is calculated, the sampling point curvature value is obtained by using the curvature calculation formula, and the contour shape feature vector is obtained by normalizing the curvature value. The position and direction of the contour curve in the actual space are calculated by using the homogeneous coordinate transformation method, the actual size parameters of the table and chair are determined by using the scale transformation, and the space mapping result is obtained. The space mapping result is data encapsulated, the position coordinates, size ratio and curve parameters are organized according to the preset data structure, the standard data package is generated, and is transmitted to the smart home data center. In the table and chair space mapping process, the projection of the three-dimensional coordinates to the two-dimensional plane adopts the perspective projection matrix, and the matrix elements include the camera intrinsic parameters and extrinsic parameters. For a camera with a focal length of 50 mm and a principal point coordinate (320, 240), the projection matrix P can be represented as a 3x4 matrix, wherein the intrinsic matrix K includes the focal length and the principal point information, and the extrinsic matrix [R|t] describes the camera pose. The affine correction of the projected coordinates is performed by using the homography matrix H, the perspective distortion is eliminated, and the parallel lines remain parallel after projection. The surf algorithm is used to extract the feature descriptor in the feature point matching, and each feature point includes a 64-dimensional vector describing its local features. In the home space visual feature map, 1000 feature points are extracted, and the gradient direction histogram of each feature point is calculated. The nearest neighbor matching method is used, the distance ratio threshold is set to 0.7, and when the ratio of the nearest neighbor distance to the second nearest neighbor distance is less than the threshold, the matching is considered to be valid. Usually, 200-300 pairs of valid matching points can be obtained. In the least square optimization, the objective function J is composed of the sum of the squares of the Euclidean distances between the matching point pairs, J = å ||pi-qi||2, wherein pi and qi are the coordinates of the corresponding feature point pairs. The gradient descent method is used to optimize the position parameters x and y and the size ratio s, the learning rate is set to 0.01, the iteration is stopped when the objective function value decreases by less than 0.001, and usually 50-100 iterations are needed to reach convergence. The contour curve is described by using the cubic Bezier curve, the curve equation is P(t) = (1-t)3P0+3t(1-t)2P1+3t2(1-t)P2+t3P3, and the parameter t varies in the interval [0, 1].The selection of the control points P0, P1, P2 and P3 is based on the positions of the contour inflection points, the distance between adjacent control points is set to 0.25-0.3 times the total length of the contour line, and the tangent direction is consistent with the local trend of the contour line. 100 sampling points on the contour curve are obtained by uniform sampling, and the curvature κ of each point is calculated as κ = |x'y''-y'x''| / (x'2+y'2)^(3 / 2), wherein x', y' are the first-order derivatives of the parameter t, and x'', y'' are the second-order derivatives. The curvature values are normalized and mapped to the interval [0, 1] to obtain a shape feature vector. Actual space mapping uses a 4x4 homogeneous transformation matrix, which includes a rotation matrix R and a translation vector t. According to the size of the room, the scale is determined, and the normalized coordinates are converted into actual sizes. Finally, the calculation results are packaged in JSON format, and the data includes position coordinates (x, y, z), direction angles (α, β, γ), size scale s and 100 contour curve sampling point coordinates. The calculation process takes about 200 milliseconds on a general computer, and the positioning accuracy is better than 5 centimeters.

[0030] S105, according to the calculated optimal position and shape parameters, dynamically adjusting the size, shape and color of the hidden contour of the table and chair to coordinate with the surrounding environment, generating an initial design scheme of the hidden contour.

[0031] According to the affine transformation method, the size of the table and chair contour is scaled, and the table and chair contour is obtained by cubic spline interpolation to obtain a smooth boundary curve; a color clustering method is used to obtain a main tone from the smooth boundary curve, and a linear gradient color band is obtained by hue saturation mapping; according to the linear gradient color band, a polynomial radial basis function is used for spatial interpolation, and a transition area color distribution map is obtained by color difference threshold constraint; for the transition area color distribution map, a weight fusion method is used to superimpose the environment texture, and an edge transparency decreasing hidden contour is obtained by an alpha blending function.

[0032] Specifically, according to the optimal position parameters, the affine transformation method is used to scale the size and deform the shape of the table and chair hidden contour, the cubic spline interpolation is used to smooth the contour boundary, and the boundary point sequence is extracted from the smooth boundary curve according to the fixed arc length interval. For the color distribution of the environment space, the color clustering method is used to extract the main color tone, the reference color of the hidden contour is calculated through the hue saturation mapping, and the linear gradient color band is generated according to the brightness gradient. According to the gradient color band, the spatial interpolation of the regional color is carried out by using the polynomial radial basis function, the color distribution of the transition area is optimized by color difference threshold constraint, and the color transition map is obtained. The color transition map is processed by Gaussian smoothing, the brightness histogram equalization method is used to adjust the color light and dark distribution, and the smooth color map is obtained. For the texture characteristics of the environment, the Markov random field method is used to generate similar texture with the environment, and the texture coordination degree is calculated by matching the texture feature vector. According to the texture coordination degree distribution, the texture pattern and the smooth color map are superimposed by using the weight fusion method, and the edge area is decreased in transparency by using the alpha blending function. The fusion result is adjusted in contrast, the material boundary is smoothly transitioned by using the bilateral filtering method, and the initial design scheme of the hidden contour is obtained. In the process of dynamic adjustment of the table and chair contour, the affine transformation uses a 3*3 transformation matrix, which contains translation, rotation and scaling parameters. For the adjustment of the office chair contour, the scaling coefficient is set to 0.8-1.2, and the rotation angle changes within ±15 degrees. The cubic spline interpolation uses the basis function B(t)=(1-t)3 / 6, and the control point interval is set to 20mm to uniformly sample 200 boundary points on the contour curve. In color processing, the k-means clustering is used to extract the main color tone of the environment, and the number of clustering centers is set to 5. The pixels in the RGB color space are clustered to obtain the color value and proportion of each clustering center. The clustering center with the highest proportion is selected as the reference color, and the gradient color band is generated in the HSV space, with a hue offset range of ±10 degrees and a saturation variation range of 0.2. The radial basis function uses a polynomial form φ(r)=(r2+c2)^(1 / 2), where r is the spatial distance and c is the shape parameter, which is set to 0.5. A 16*16 control point grid is set in the contour area, and the interpolation coefficients are obtained by solving the linear equation system. The color difference threshold is set to 3 units, and local adjustment is performed when the color difference between adjacent regions exceeds the threshold. Color smoothing uses a 5*5 Gaussian kernel for filtering, with a standard deviation of 1.2. The brightness value is remapped to the range of 40-220 through histogram equalization, maintaining details while avoiding excessive darkness or brightness. The texture generation is based on a 32*32 pixel sample block, and the Markov random field is used to establish the conditional probability relationship between pixels. The texture feature vector contains three components: contrast, directionality and roughness, which are calculated by the gray level co-occurrence matrix. The feature vectors of the environment texture and the generated texture are calculated, the texture coordination degree is measured by cosine similarity, and the coordination threshold is set to 0.85. In weight fusion, the weight of the texture pattern decreases linearly from the center to the edge, with a decrease range of 50 pixels.The edge processing adopts an alpha blending function f(x) = 1-x2, x being a normalized distance to the edge. The spatial standard deviation of the bilateral filtering is set to 3, and the value range standard deviation is set to 0.1, to achieve a smooth effect of edge preservation. The contrast gain is adjusted to 1.2 to improve the material detail performance. The image resolution generated in the processing process is 1024*1024 pixels, the color depth is 24 bits, the processing time is about 300 milliseconds, and the generated design scheme contains four key elements of position, shape, color and texture, and the data amount is about 2MB.

[0033] In S106, a physical simulation method is used to simulate the movement process of the table and chair, to determine whether the hidden contour design affects the normal use function of the table and chair, and to optimize the hidden contour design according to the simulation result to generate an optimized hidden contour design scheme.

[0034] The table and chair structure is meshed by tetrahedral elements, the grid node positions are obtained by the element mass criterion, and the elastic modulus and Poisson's ratio parameters are extracted from a material database; boundary conditions are established according to the grid node positions and material parameters, a stiffness matrix is constructed by a node displacement interpolation function, and a node displacement field is obtained by solving the static equilibrium equation for the stiffness matrix; the element strain is calculated according to the node displacement field, the element stress is solved by a constitutive equation, and a stress distribution cloud map is obtained by the von-mises equivalent stress criterion for the element stress; the structure shape is topologically optimized according to the stress distribution cloud map, the natural frequency and mode shape are calculated by harmonic response analysis, and the optimized structure parameters are generated for the natural frequency and mode shape.

[0035] Specifically, tetrahedral elements were used to mesh the table and chair structure using the finite element method. Mesh quality was controlled by an element quality criterion, and fixed and displacement constraints were applied to the mesh nodes. Elastic modulus and Poisson's ratio parameters were extracted from a material database. Based on the structural mesh model, boundary conditions were established using gravity and external loads. A stiffness matrix was constructed using nodal displacement interpolation functions, yielding the static equilibrium equations. For the static equilibrium equations, the Lagrange equations were used to establish the kinematic differential equations. The Runge-Kutta method was used to numerically solve the kinematic equations, obtaining the nodal displacement field. Based on the nodal displacement field, the strain-displacement relationship was used to calculate the element strain, and the constitutive equations were used to solve for the element stress, obtaining the structural stress field distribution. For the stress field distribution, the von Mises equivalent stress criterion was used for strength verification. The stress gradient method was used to refine the mesh in areas exceeding the strength limit, obtaining a stress distribution contour map. Based on the stress distribution contour map, a topology optimization method was used to adjust the structural shape. Sensitivity analysis was used to determine the optimization direction, resulting in an improved structural scheme. The improved structural design was dynamically verified. Harmonic response analysis was used to calculate natural frequencies and mode shapes, and modal superposition was employed to verify the structure's dynamic characteristics. Based on the verification results, optimized hidden profile curve equations were generated. Structural parameters were encapsulated in JSON format and encrypted using AES encryption before being transmitted to the smart home platform. In the finite element analysis of the table and chair structure, tetrahedral elements were used to mesh the structure, with element sizes set to 5-10 mm. Mesh quality was controlled using a Jacobian coefficient greater than 0.6. For a standard office chair structure, approximately 50,000 tetrahedral elements were generated, resulting in approximately 15,000 mesh nodes. Steel was selected as the material, with an elastic modulus of 210 GPa, a Poisson's ratio of 0.3, and a density of 7850 kg / m³. The static equilibrium equations were expressed as a linear system of equations Kx = F, where K is the stiffness matrix, x is the nodal displacement vector, and F is the external load vector. The element stiffness matrix ke = ∫BTDBdV is constructed by interpolating the displacements within the element using shape functions N(ξ,η,ζ), where B is the strain matrix and D is the elasticity matrix. Gravity loads are considered based on the chair's self-weight, and external loads are set as a uniformly distributed load of 1000N on the seat surface. The equations of motion are in Lagrange form. M is the mass matrix, and C is the damping matrix. The fourth-order Runge-Kutta method is used for solving, and the time step is set to 0.001 seconds, and the total calculation time is 1 second. For the seat moving working condition, a horizontal thrust of 100 N is applied, and the maximum displacement is about 20 mm. The stress calculation is based on Hooke's law σ=Dε, the element strain ε=Bu is calculated by the node displacement, and then the stress σ is solved. The material yield strength is set to 235 MPa, and the von-mises criterion σvm=(σ12-σ1σ2+σ22)^(1 / 2)≤[σ] is used to judge the strength. The stress concentration occurs at the connection of the backrest, and the maximum equivalent stress reaches 180 MPa. The grid size is refined to 2 mm to improve the calculation accuracy. The structure optimization uses the density method, and the design variable is the relative density of the element ρ, and the objective function is the minimization of the structure flexibility. The SIMP model is used to establish the relationship between density and elastic modulus E(ρ)=E0ρ^p, where p=3. Sensitivity analysis shows that the material utilization rate is low at the backrest support, and the thickness of this area is reduced through iterative optimization. The first-order natural frequency is 8.5 Hz, and the mode shape is front and rear swing. The Rayleigh damping C=αM+βK is used, and the coefficients are α=0.5 and β=0.002. The harmonic response analysis is performed in the frequency range of 0-50 Hz, and the resonance peak displacement amplitude is less than 5 mm. Finally, the optimization results are packaged in json format, including node coordinates, element connection relationship, material parameters, etc. The data packet size is about 2 MB, and the aes-256 encryption algorithm is used for encryption before uploading to the platform.

[0036] In S107, the optimized hidden contour design scheme parameters are transmitted to the intelligent control system, combined with historical design scheme data, and trained through a machine learning algorithm to realize dynamic adaptation of the table and chair hidden contour to the surrounding environment.

[0037] According to the hidden contour design scheme data packet sent by the smart home platform, the data packet is decrypted by a symmetric encryption algorithm, and valid data sequences are obtained by a data integrity verification method; for the valid data sequences, a sliding window method is used to preprocess the parameter time series, and smooth parameter sequences are obtained by wavelet transform to filter out noise data; according to the smooth parameter sequence, a recurrent neural network is used to construct a time series predictor, and time series feature vectors are obtained by feature extraction of the smooth parameter sequence through a long short-term memory unit; for the time series feature vectors, a radial basis function support vector regressor is used for training and fitting, and if the prediction error of the validation set is less than a preset threshold, online incremental learning is performed by collecting environmental data through a sliding time window.

[0038] Specifically, the optimized hidden contour design scheme data packet is obtained from the smart home platform data center, decrypted by symmetric encryption algorithm, and verified for data validity by data integrity check method. According to the data packet content, the parameter time series is preprocessed by sliding window method, and the noise data is filtered by wavelet transform to obtain the smooth parameter sequence. For the smooth parameter sequence, a recurrent neural network is constructed to build a time series predictor, and the historical data is feature extracted by long short-term memory unit to obtain the time series feature vector. According to the time series feature vector, the feature data is normalized by min-max normalization method, and the data subsets are divided according to the time proportion by stratified sampling method. The data subsets are verified by five-fold cross-validation, the training data is fitted by radial basis function support vector regressor, and the kernel function parameters are optimized by grid search method. According to the verification set, the prediction error is calculated, the network parameters are updated by adaptive moment estimation method, and the parameter convergence is controlled by learning rate annealing method. The trained network is incrementally learned online, the environmental data is continuously collected by sliding time window, and the dynamic update is realized by parameter adaptive method. The updated network parameters and control commands are packaged, the data is transmitted to the intelligent control unit by encrypted transmission protocol, and the real-time dynamic adaptive control is realized. The data transmission adopts AES-256 encryption algorithm to encrypt the design scheme data packet, the key length is 256 bits, and the packet length is 128 bits. The data packet contains contour parameters, material parameters and environmental parameters, and the total size is about 2MB. The data integrity is verified by MD5 checksum to ensure that the data is not tampered during transmission. When preprocessing the parameter sequence, a sliding window with a length of 24 is used, and the window step is set to 8. The data is decomposed by db4 wavelet basis for 3 layers, the high frequency coefficients are denoised by soft threshold method, and the threshold is set to 2.5 times the standard deviation. The signal-to-noise ratio of the processed data is improved by about 6dB, and the data smoothness is significantly improved. The recurrent neural network adopts a bidirectional LSTM structure, containing 2 layers of hidden layers, each layer with 128 neurons. The input features include position coordinates, shape parameters, environmental lighting, etc. 10 dimensions. The forgetting gate controls the historical information retention ratio, and the input gate controls the new information update ratio. The data standardization maps the feature values to the [0, 1] interval by the formula x'=(x-xmin) / (xmax-xmin). The hierarchical sampling divides the data set into 5 time periods in chronological order, and randomly selects 80% of the samples in each time period as the training set and 20% as the test set. The positive and negative sample ratio in the training set is 1:1. The support vector regressor adopts Gaussian radial basis kernel function K(x,x')=exp(-γ||x-x'||2), where γ is the kernel parameter. The γ value is optimized in the range of [0.001, 1] by grid search method, and the penalty factor C is searched in the range of [1, 1000]. The cross-validation is performed in five folds, and the average prediction error is less than 5%.The parameter update adopts an adaptive moment estimation algorithm, the initial value of the learning rate is set to 0.001, and is attenuated to 0.95 times of the original value after 1000 iterations. The momentum term is set to 0.9, which effectively prevents parameter oscillation. The root mean square error of the network on the validation set is reduced from the initial 0.15 to 0.03. Online incremental learning adopts a 60-second sliding window to continuously collect environmental data, and the new and old data are fused by an exponential weighting average method, and the weight attenuation coefficient is 0.95. The adaptive algorithm dynamically adjusts the network parameters according to the prediction error size, and triggers parameter update when the error exceeds the threshold value 0.1. The updated control command is transmitted to the execution unit at a frequency of 50Hz, realizing millisecond-level response.

[0039] The above is only an example and description of the structure of the present application. Those skilled in the art can make various modifications or supplements to the described specific embodiments or use similar methods to replace them, as long as they do not deviate from the structure of the application or exceed the scope defined by the present claims.

Claims

1. A smart home platform-based table and chair remote management method, characterized in that, The method comprises: Obtaining the environment information around the position where the table and chair are located, including the color and material attribute data of the wall and floor environment elements, determining the initial color and material parameters of the hidden contour of the table and chair according to the obtained environment color and material attribute data, including hue, saturation, lightness and surface texture characteristics, wherein the hidden contour of the table and chair refers to a visual adaptive contour obtained by combining the visual features of the home space and taking into account the normal use function and environmental visual coordination of the table and chair, and the physical outer contour line of the table and chair refers to an original contour reflecting the inherent shape and spatial coordinates of the table and chair obtained by laser scanning the physical structure of the table and chair and extracting key inflection point coordinates; Establishing a three-dimensional model of the table and chair, and obtaining the key structure point and outer contour line coordinate information of the table and chair by analyzing the three-dimensional model; Analyzing the home space image by using a computer vision algorithm, extracting the main visual elements in the home space, judging the spatial relationship and visual coordination between the visual elements, and generating a home space visual feature map; Matching the key structure point and outer contour line coordinate information of the three-dimensional model of the table and chair with the home space visual feature map, and calculating the best position and shape parameters of the hidden contour of the table and chair in the home space by using a space mapping algorithm; According to the calculated best position and shape parameters, dynamically adjusting the size, shape and color of the hidden contour of the table and chair to make it coordinate with the surrounding environment, and generating an initial design scheme of the hidden contour, wherein the dynamic adjustment of the color refers to obtaining the main color tone from the smooth boundary curve of the hidden contour of the table and chair by using a color clustering method, generating a linear gradient color band by hue saturation mapping, performing spatial interpolation on the color by using a polynomial radial basis function, and finally obtaining a transition area color distribution map by color difference threshold constraint to realize the natural coordination of the color with the surrounding environment; Simulating the motion process of the table and chair by using a physical simulation method, judging whether the hidden contour design affects the normal use function of the table and chair, optimizing the hidden contour design according to the simulation result, and generating an optimized hidden contour design scheme; Transferring the parameters of the optimized hidden contour design scheme to an intelligent control system, training by using a machine learning algorithm in combination with historical design scheme data, and realizing the dynamic adaptation of the hidden contour of the table and chair to the surrounding environment.

2. The method of claim 1, wherein, The method comprises: Scanning the wall area by using a color sensor array, obtaining the wall main color tone value from the wall color distribution data, and obtaining the wall lightness distribution map by using a material texture analyzer; Measuring the wall reflection degree by using a ray tracing model according to the wall main color tone value, obtaining the indoor illumination intensity value from a luminosity calculation unit, and obtaining the illumination intensity distribution map by quantitatively processing the illumination intensity value; The light intensity distribution map and the wall depth distribution map are processed through convolution operation to obtain an overall ambient brightness distribution map, and a three-dimensional space brightness model is established according to the overall ambient brightness distribution map; A depth camera is used to perform three-dimensional scanning modeling on the surfaces of the table and chairs, light parameter calculation is performed on the contour regions of the table and chairs according to the three-dimensional space brightness model, and a set of light parameters of the contour regions of the table and chairs is obtained.

3. The method of claim 1, wherein, The three-dimensional model of the table and chairs is established, the key structural points and the outer contour line coordinate information of the table and chairs are obtained by analyzing the three-dimensional model, including: A laser scanner is used to obtain a point cloud data set of the surfaces of the table and chairs, the point cloud data set is fused through a point cloud registration algorithm to obtain a complete point cloud model; According to the complete point cloud model, a triangular mesh is divided by using a Poisson reconstruction algorithm, and a smooth mesh model is obtained by optimizing the grid vertex coordinates through a Laplace smoothing method; According to the smooth mesh model, a grid vertex curvature value is extracted by using a Gaussian curvature calculation formula, and a feature point set is obtained by setting a curvature threshold; According to the feature point set, a binary image is obtained by performing edge detection on the feature point set by using a Sobel operator, and a key corner coordinate sequence is extracted from the binary image.

4. The method of claim 1, wherein, The computer vision algorithm is used to analyze the home space image, extract the main visual elements in the home space, judge the spatial relationship and visual coordination between the visual elements, and generate a home space visual feature map, including: An original image of the home space is obtained by using a depth camera, and the original image is processed by a Gaussian filter and a gamma correction to obtain an enhanced image; According to the enhanced image, a gray level co-occurrence matrix texture feature, a color moment color feature, and a gradient operator edge feature are extracted to obtain a feature vector, and a visual element label map is obtained by region growing segmentation of the feature vector; The visual element centroid coordinates are calculated according to the visual element label map, a spatial distance matrix is constructed by using the Euclidean distance method, the color similarity of the visual elements is calculated by the intersection of the color histogram, and a visual coordination degree matrix is obtained; According to the visual coordination degree matrix, the visual elements are hierarchically divided by using a spectral clustering method, a space division structure is constructed by using a quadtree method, a visual attention distribution map is generated by using a brightness weight calculation, and a home space visual feature map is obtained.

5. The method of claim 1, wherein, The key structural points and the outer contour line coordinate information of the three-dimensional model of the table and chairs are matched with the home space visual feature map, the best position and shape parameters of the hidden contour of the table and chairs in the home space are calculated by using a space mapping algorithm, including: The key structural points of the three-dimensional model of the table and chairs are projected onto a two-dimensional plane by using a perspective projection method, and a standardized plane coordinate point set is obtained by using a homography matrix calculation; According to the standardized plane coordinate point set, a feature descriptor is obtained by using a surf feature point extraction algorithm, and a feature point corresponding relationship set is obtained by using a nearest neighbor matching method; A least squares optimization objective function is established according to the feature point corresponding relationship set, and the position coordinates of the table and chairs are iteratively optimized by using a gradient descent method to obtain an initial placement position of the table and chairs. According to the initial placement of the table and chair, a contour is fitted by a cubic Bezier curve, a curve control point interval ratio and a tangent direction constraint are set, and a contour shape feature vector is calculated by a sampling point curvature value.

6. The method of claim 1, wherein, According to the calculated optimal position and shape parameters, the size, shape and color of the hidden contour of the table and chair are dynamically adjusted to be coordinated with the surrounding environment, and an initial design scheme of the hidden contour is generated, including: According to the table and chair contour, an affine transformation method is used for size scaling, and the table and chair contour is subjected to cubic spline interpolation to obtain a smooth boundary curve; A color clustering method is used to obtain a main tone from the smooth boundary curve, and a linear gradient color band is obtained by hue saturation mapping; According to the linear gradient color band, a polynomial radial basis function is used for spatial interpolation, and a transition area color distribution map is obtained by color difference threshold constraint; For the transition area color distribution map, a weight fusion method is used to superimpose the environment texture, and a hidden contour with decreasing edge transparency is obtained by an alpha blending function.

7. The method of claim 1, wherein, The physical simulation method is used to simulate the movement process of the table and chair, to judge whether the hidden contour design affects the normal use function of the table and chair, and to optimize the hidden contour design according to the simulation result to generate an optimized hidden contour design scheme, including: The table and chair structure is meshed by tetrahedral elements, the grid node positions are obtained by element mass criterion, and the material database is used to extract the elastic modulus and Poisson's ratio parameters; Boundary conditions are established according to the grid node positions and material parameters, a stiffness matrix is constructed by node displacement interpolation function, and node displacement field is obtained by solving static equilibrium equation of the stiffness matrix; According to the node displacement field, the element strain is calculated, the element stress is solved by constitutive equation, and the stress distribution cloud map is obtained by von-mises equivalent stress criterion; According to the stress distribution cloud map, the structure shape is topologically optimized, the natural frequency and mode shape are calculated by harmonic response analysis, and the optimized structure parameters are generated according to the natural frequency and mode shape.

8. The method of claim 1, wherein, The optimized hidden contour design scheme parameters are transmitted to the intelligent control system, combined with the historical design scheme data, and trained by machine learning algorithm to realize the dynamic adaptation of the table and chair hidden contour to the surrounding environment, including: According to the hidden contour design scheme data packet sent by the smart home platform, the data packet is decrypted by symmetric encryption algorithm, and the effective data sequence is obtained by data integrity verification method; For the effective data sequence, the parameter time series is preprocessed by sliding window method, and the smooth parameter sequence is obtained by wavelet transform to filter out noise data; According to the smooth parameter sequence, a time series predictor is constructed by recurrent neural network, and the time series feature vector is obtained by feature extraction of the smooth parameter sequence through long short-term memory unit; For the time series feature vector, a radial basis kernel function support vector regressor is trained and fitted, and if the prediction error of the validation set is less than the preset threshold, the environment data is collected by sliding time window for online incremental learning.

Images