CN-121357765-B - Illumination brightness self-adaptive adjusting system based on visual detection

CN121357765BCN 121357765 BCN121357765 BCN 121357765BCN-121357765-B

Abstract

The invention discloses an illumination brightness self-adaptive adjusting system based on visual detection, which comprises an image acquisition and preprocessing module, an illumination characteristic extraction and modeling module, a space illumination distribution generation module, a brightness deviation calculation module and a driving execution module, wherein the image acquisition and preprocessing module is used for acquiring an environment image and preprocessing the environment image, the illumination characteristic extraction and modeling module is used for inputting the preprocessed environment image into an improved Neural ODE network to obtain illumination characteristic data, the space illumination distribution generation module is used for forming space illumination distribution data according to the illumination characteristic data, the brightness deviation calculation module is used for calculating illumination brightness deviation according to the space illumination distribution data, the self-adaptive control module is used for inputting the illumination brightness deviation into an MRAC control model to generate an illumination brightness adjustment instruction, and the driving execution module is used for driving an illumination driving circuit according to the illumination brightness adjustment instruction to modulate driving current, so that the illumination brightness of an illumination lamp is self-adaptively adjusted, and real-time perception and intelligent adjustment of illumination environment brightness are realized.

Inventors

CHEN SHOUJIAN
GAO GAOYUAN

Assignees

广州市凌锐工业科技有限公司

Dates

Publication Date: 20260508
Application Date: 20251022

Claims (6)

1. An illumination brightness self-adaptive adjustment system based on visual detection, comprising: The image acquisition and preprocessing module is used for acquiring an environment image and preprocessing the environment image; the illumination characteristic extraction and modeling module is used for inputting the preprocessed environment image into an improved Neural ODE network, and obtaining illumination characteristic data through input coding, continuous time dynamic modeling and numerical integration; The space illumination distribution generation module is used for acquiring global brightness information and local brightness information according to the illumination characteristic data to form space illumination distribution data; the brightness deviation calculation module is used for comparing the space illumination distribution data with a preset brightness reference value and calculating illumination brightness deviation; the self-adaptive control module is used for inputting the illumination brightness deviation into the MRAC control model to generate an illumination brightness adjustment instruction; The driving execution module is used for driving the illumination driving circuit according to the illumination brightness adjustment instruction, modulating driving current and realizing self-adaptive adjustment of illumination brightness through a photoelectric conversion relation; The generation of the spatial illumination distribution data specifically comprises the following steps: reconstructing the brightness vectors in the illumination characteristic data into an illumination brightness matrix according to preset row numbers and preset column numbers, wherein the reconstruction process is to sequentially fill the brightness vectors into a two-dimensional matrix according to a row priority mode according to the preset row numbers and the preset column numbers to form the illumination brightness matrix; Carrying out average calculation on brightness values of all pixels in the light illumination matrix to obtain global average brightness, and calculating square average of difference values between the brightness values of all pixels and the global average brightness to obtain global brightness variance to form global brightness information; Setting a local window size, taking each pixel brightness value as a center in an illumination brightness matrix, calculating the average of the pixel brightness values of the neighborhood in the local window size to obtain local average brightness, calculating the weighted square average of the difference value between the pixel brightness values of the neighborhood and the local average brightness to obtain local brightness variance, and forming local brightness information; Combining the illumination brightness matrix, the global brightness information and the local brightness information to generate space illumination distribution data; the calculating of the illumination brightness deviation amount specifically includes: Setting a preset brightness reference value, wherein the preset brightness reference value comprises a global brightness reference and a local brightness reference; Calculating a global luminance deviation based on a global luminance reference, the global luminance deviation being a difference between global luminance information and the global luminance reference; Constructing a local brightness deviation matrix based on the local brightness reference, wherein the local brightness deviation matrix construction refers to element-by-element subtraction of local brightness information and the local brightness reference at corresponding positions; calculating a local brightness deviation scalar based on the local brightness deviation matrix, wherein the local brightness deviation scalar is obtained by averaging squares of elements in the local brightness deviation matrix; And carrying out non-negative weight coefficient weighting on the square of the global brightness deviation and the local brightness deviation scalar to obtain the illumination brightness deviation quantity.
2. The illumination brightness self-adaptive adjusting system based on visual detection according to claim 1, wherein the modules are realized by the following method: Acquiring an environment image through an image acquisition device, and preprocessing; Dynamically processing the preprocessed environment image by using an improved Neural ODE network, and obtaining illumination characteristic data through characteristic extraction and continuous illumination change modeling; Obtaining global brightness information and local brightness information according to the illumination characteristic data, and generating space illumination distribution data; comparing the space illumination distribution data with a preset brightness reference value, and calculating the illumination brightness deviation amount; Inputting the illumination brightness deviation amount into an MRAC control model, and generating an illumination brightness adjustment instruction through the operation of the MRAC control model; based on the illumination brightness adjusting instruction, the illumination driving circuit is driven to adaptively adjust the brightness of the illumination lamp.
3. The illumination intensity adaptive adjustment system based on visual inspection of claim 2, wherein the preprocessing includes image denoising, image enhancement, illumination correction, scene segmentation, and feature extraction.
4. The illumination brightness self-adaptive adjustment system based on visual detection according to claim 2, wherein the obtaining of the illumination characteristic data specifically comprises: Inputting the preprocessed environment image into an improved Neural ODE network, wherein the improved Neural ODE network comprises an input coding module, a dynamic modeling module, a numerical integration module and a characteristic output module, the input coding module is used for converting the preprocessed environment image into an initial illumination characteristic vector, the dynamic modeling module is used for modeling the initial illumination characteristic vector in a continuous time, the numerical integration module is used for integrating the output of the dynamic modeling module in a time interval, and the characteristic output module is used for converting a final-state illumination characteristic vector into illumination characteristic data with illumination brightness self-adaptive adjustment; The method comprises the steps of constructing a preprocessed environment image into an environment image data matrix, converting the environment image data matrix into an initial illumination feature vector through an input coding module, extracting brightness information through weighted RGB linear combination, carrying out normalization processing on extracted brightness values to obtain pixel intensity which is limited in a [0,1] interval, carrying out spatial resampling on the normalized brightness image, and mapping the original image into a size by bilinear interpolation The method comprises the steps of constructing an environment image data matrix, flattening the environment image data matrix into a vector form according to rows in the conversion process, performing linear mapping on the flattened vector by using a weight matrix and a bias vector, and outputting an initial illumination feature vector by combining activation function calculation; inputting the initial illumination feature vector into a dynamic modeling module, wherein the dynamic modeling module introduces an image feature vector, a time sine and cosine basis function and a nonlinear mapping function to perform continuous modeling to obtain a dynamic equation; performing numerical integration on the dynamic equation in a time interval through a numerical integration module to obtain a target illumination characteristic vector; And inputting the target illumination characteristic vector to a characteristic output module, and outputting illumination characteristic data.
5. The illumination brightness self-adaptive adjustment system based on visual detection according to claim 2, wherein the generation of the illumination brightness adjustment command specifically comprises: Defining the illumination brightness deviation as an error signal, and setting a state variable of a reference model and outputting the reference model by using the reference model in the MRAC control model; Constructing a regression vector based on a state variable of a reference model, a reference model output and an error signal, and presetting an adaptive parameter vector, wherein the construction process is to carry out scale unification on the state variable of the reference model, the reference model output and the error signal, generate a cross characteristic through the product of the reference model output and the error signal, and cascade the state variable of the reference model, the reference model output, the error signal, the cross characteristic and a constant item after the scale unification according to a fixed order to form the regression vector; generating a control law by using the linear combination of the regression vector and the self-adaptive parameter vector, and calculating a control quantity, namely, arranging the constructed regression vector into column vectors according to a fixed sequence, multiplying the self-adaptive parameter vector and the components corresponding to the regression vector one by one, and summing all products to obtain an output result of the control law; Associating the control quantity with the error signal, updating the self-adaptive parameter vector according to the product of the control quantity and the error signal, and correcting the control law by utilizing the updated self-adaptive parameter vector to obtain the updated control quantity; Applying amplitude constraint to the updated control quantity, and limiting the control quantity between a preset lower limit and an upper limit to obtain a constraint control quantity; And converting the constraint control quantity into an illumination brightness adjustment instruction and outputting the illumination brightness adjustment instruction, wherein the conversion process is to convert the constraint control quantity into a voltage signal through level mapping after normalization processing, so as to form the illumination brightness adjustment instruction.
6. The illumination brightness adaptive adjustment system based on visual detection according to claim 2, wherein the illumination brightness adaptive adjustment specifically comprises: Inputting an illumination brightness adjustment instruction to an illumination driving circuit in the form of a duty ratio signal; Generating a driving current reference value according to the duty ratio signal, wherein the driving current reference value is obtained by carrying out proportional mapping between a preset current lower limit value and a preset current upper limit value; Generating an actual driving current in the illumination driving circuit, and calculating a difference value between a driving current reference value and the actual driving current to obtain a current error; Generating a driving control amount by means of proportional-integral operation based on the current error; The driving control quantity is acted on the illumination driving circuit to modulate the actual driving current, and the illumination brightness is obtained through the photoelectric conversion relation, and the illumination brightness is adaptively adjusted, wherein the photoelectric conversion relation is a quantitative relation that the actual driving current is mapped into the illumination brightness.

Description

Illumination brightness self-adaptive adjusting system based on visual detection Technical Field The invention relates to the field of visual detection and intelligent illumination, in particular to an illumination brightness self-adaptive adjusting system based on visual detection. Background With the development of intelligent buildings, intelligent houses and intelligent cities, lighting systems have gradually evolved from traditional manual switch control to intelligent and self-adaptive control modes. The existing lighting control method mainly comprises timer-based control, sensor-based control and man-machine interaction-based control. For example, the ambient brightness and personnel activities are detected by using a photosensitive sensor or an infrared sensor, and then the light is subjected to switch control or brightness adjustment through simple logic judgment. Although the method can realize energy conservation and convenience to a certain extent, the method often has the defects of insufficient perception of illumination distribution in a complex space, low adjustment precision, poor self-adaptability and the like because the method relies on a single-point sensor to acquire limited environmental information. In the prior art, the illumination control mode based on the sensor can only reflect the brightness condition of local point positions, and cannot comprehensively reflect the illumination distribution of the whole space. When there are multiple light sources or occluded areas in the environment, the information output by a single point sensor may be significantly different from the actual visual perception, resulting in an adjustment result that deviates from the actual demand. In addition, the sensor is high in cost and complex in installation and arrangement, and the difficulty in system deployment is increased. On the other hand, partial researches attempt to introduce an illumination analysis method based on image processing, and acquire an environment image through a camera and extract brightness information so as to realize auxiliary judgment on space illumination. However, the traditional image processing method is mostly dependent on static feature extraction and threshold segmentation, is easily affected by noise, reflection and environmental dynamic change, has poor robustness, and cannot realize high-precision and real-time illumination control. Disclosure of Invention The invention aims to provide an illumination brightness self-adaptive adjustment system based on visual detection, which fully utilizes the technologies of environment image acquisition, an improved Neural ODE network, space illumination distribution analysis, an MRAC control model and the like, and describes the overall method for intelligent self-adaptive adjustment of illumination brightness by extracting illumination characteristics through image sensing, calculating global and local brightness deviation and generating an adjustment instruction. According to an embodiment of the invention, an illumination brightness self-adaptive adjusting system based on visual detection comprises: The image acquisition and preprocessing module is used for acquiring an environment image and preprocessing the environment image; the illumination characteristic extraction and modeling module is used for inputting the preprocessed environment image into an improved Neural ODE network, and obtaining illumination characteristic data through input coding, continuous time dynamic modeling and numerical integration; The space illumination distribution generation module is used for acquiring global brightness information and local brightness information according to the illumination characteristic data to form space illumination distribution data; the brightness deviation calculation module is used for comparing the space illumination distribution data with a preset brightness reference value and calculating illumination brightness deviation; the self-adaptive control module is used for inputting the illumination brightness deviation into the MRAC control model to generate an illumination brightness adjustment instruction; And the driving execution module is used for driving the illumination driving circuit according to the illumination brightness adjustment instruction, modulating driving current and realizing self-adaptive adjustment of illumination brightness through a photoelectric conversion relation. Optionally, the modules are realized by the following method: Acquiring an environment image through an image acquisition device, and preprocessing; Dynamically processing the preprocessed environment image by using an improved Neural ODE network, and obtaining illumination characteristic data through characteristic extraction and continuous illumination change modeling; Obtaining global brightness information and local brightness information according to the illumination characteristic data, and generating space illumination dist