CN-122016871-A - Lens defect detection method and system based on photoelectric multi-fusion perception technology

CN122016871ACN 122016871 ACN122016871 ACN 122016871ACN-122016871-A

Abstract

The invention discloses a lens defect detection method and a system based on photoelectric multi-fusion perception technology, which relate to the field of product defect detection and comprise a capturing module, a detection module and a detection module, wherein the capturing module is used for acquiring visible light, infrared and laser scattering signals on the surface and in the lens and converting the acquired signals into processable electric signals; the invention acquires and processes the multi-mode signal cooperatively, captures the surface and internal defect information of the lens comprehensively, and improves the defect recognition accuracy and classification degree greatly based on the comparison of the dynamic adaptation judgment standard and the high-efficiency characteristic, locks the three-dimensional position of the defect accurately, outputs reliable detection data and feeds back quickly, and improves the quality and production efficiency of the lens effectively.

Inventors

HUAN WEIWEI
ZHAO GANG
CHEN LANG
CAO LONGCHEN
YANG FENG

Assignees

湖北五方光电股份有限公司

Dates

Publication Date: 20260512
Application Date: 20260317

Claims (10)

1. A lens defect detection system based on photoelectric multi-fusion perception technology is characterized by comprising: The capturing module is used for collecting visible light, infrared and laser scattering signals on the surface and in the lens and converting the collected signals into processable electric signals; The registration module is used for receiving the electric signals output by the capture module, denoising, normalizing and spatially registering the electric signals, and establishing signal mapping according to the spatial correspondence of the same detection area of the lens so as to finish the preliminary association of the multi-mode signals; the extraction module is used for mining gray scale, texture and geometric features corresponding to the defects of the lens based on the associated information of the multi-mode signals, and constructing a feature vector set containing defect differentiation information; The identification module is used for carrying out hierarchical analysis on the feature vector set based on the dynamic threshold value and feature matching, identifying the lens defects and marking the preliminary category attributes of the defects; the integration module is used for acquiring the space coordinate information of the multi-mode signals, identifying the positions of the defects on the lenses, integrating the defect types, the positions and the signal intensity information and establishing defect detection data packets; and the feedback module is used for receiving the defect detection data packet output in the integration module and feeding back the defect detection data packet to the preset production management terminal.
2. The lens defect detection system based on the photoelectric multi-fusion perception technology according to claim 1, wherein the capturing module is integrated by a visible light sensor array, an infrared focal plane detector and a laser scattering detection unit; The sampling frequency of the visible light sensor array is synchronous with the frame frequency of the infrared focal plane detector, and the emission wavelength of the laser scattering detection unit and the characteristic absorption wavelength of the lens material are in a complementary relation; The capture module synchronously calibrates the collected multi-source original signals by the following formula: ; wherein: the synthesized electric signal after the calibration at the time t is obtained; 、、 the original electric signals corresponding to the visible light, infrared and laser scattering at the moment t; 、、 Is a visible light signal weight coefficient, an infrared signal weight coefficient and a laser scattering signal weight coefficient.
3. The lens defect detection system based on the photoelectric multi-fusion perception technology according to claim 1, wherein the electrical signal denoising obeys the following formula in the registration module: ; wherein: is the electrical signal after denoising at the coordinates (x, y); is the original electrical signal at coordinates (x, y); half-size for the filter window; M and p are respectively the transverse offset and the longitudinal offset relative to the central coordinates (x, y) in the filtering window; the average value of the original electric signal in the filtering window; the normalization processing of the electric signals adopts a min-max normalization mode to map the amplitude of the denoised electric signals to a preset interval, the spatial registration is based on the three-dimensional coordinates of preset reference mark points of lenses, the spatial alignment of the multi-mode signals is carried out through rigid transformation, and the signal mapping is completed by establishing a gray scale response correlation model of the multi-mode signals under the same spatial coordinates.
4. The lens defect detection system based on the photoelectric multi-fusion perception technology according to claim 1, wherein the geometrical characteristics are extracted by the extraction module, and the geometrical characteristics comprise three-dimensional profile parameters, curvature distribution and connected domain morphological parameters of the defects; The three-dimensional contour parameters are obtained through gray gradient inverse mapping calculation of the multi-mode signals; The construction process of the feature vector set comprises the steps of respectively carrying out normalization processing on gray scale, texture and geometric features, calculating defect recognition contribution degree of each feature based on an information gain criterion, sequencing, selecting the first X key features to form a basic feature vector, and mapping the basic feature vector to a high-dimensional feature space through kernel function mapping to form a final feature vector set containing defect differentiation information.
5. The lens defect detection system based on the photoelectric multi-fusion perception technology according to claim 1, wherein the dynamic threshold of the identification module is updated in real time by the following formula: ; Wherein; A dynamic threshold value for the kth detection; A dynamic threshold value for the k-1 th detection; the threshold value is an up-regulation coefficient and a threshold value is a down-regulation coefficient; a signal complexity factor for the kth detection region; Detecting a dispersion factor of the feature vector set for the kth time; a background uniformity factor for the kth detection region; detecting the aggregation factor of the feature vector set for the kth time; And in the feature matching stage, feature matching is performed by calculating the fusion matching degree of the feature vector to be detected and the standard defect feature vector: ; wherein: Is a feature vector And (3) with Is a fusion matching degree of (1); Is cosine similarity weight; Is that And (3) with Cosine similarity of (c); Is that And (3) with Mahalanobis distance; And (5) regulating parameters for the mahalanobis distance.
6. The lens defect detection system based on the photoelectric multi-fusion perception technology according to claim 5, wherein the hierarchical analysis operation in the identification module comprises: A primary screening layer for screening each dimension characteristic value and dynamic threshold value in the characteristic vector Comparing, eliminating all dimension characteristic values lower than Preserving at least one dimension feature value higher than Is a candidate defect feature vector for the defect; A fine matching layer for fusion matching the candidate defect feature vector with each class of feature vector in the preset standard defect feature vector library and calculating fusion matching degree Screening out feature vectors with fusion matching degree higher than a preset matching threshold and corresponding standard defect categories; The category confirmation layer is used for screening candidate categories from the fine matching layer, calculating the fusion matching degree average value of the feature vector to be detected and all standard feature vectors under the category, and marking the category as a preliminary category attribute of the defect if the average value is higher than a category confirmation threshold value, selecting the category with the highest fusion matching degree average value as the preliminary category attribute if the non-unique candidate category meets the condition, and marking the category as a composite category attribute if the difference value between the highest average value and the next highest average value is lower than a preset difference threshold value; the process of identifying the lens defects comprises the steps of obtaining candidate defect feature vectors through a preliminary screening layer, completing feature comparison through a fine matching layer, combining the judging result of a category confirmation layer, and outputting defect identification results and corresponding preliminary category attributes.
7. The lens defect detection system based on the photoelectric multi-fusion perception technology according to claim 1, wherein the integration module determines a three-dimensional accurate position of the defect through a spatial coordinate calibration model of the multi-mode signal, and the spatial coordinate calibration model is established based on a coordinate conversion relation between a CAD three-dimensional model of the lens and an actual detection scene: ; wherein [ X, Y, Z ] is the three-dimensional accurate coordinate of the defect; a rotation matrix parameter for coordinate conversion; Translation vector parameters for coordinate transformation; initial three-dimensional coordinates of the defect detected by the visible light signal; Initial three-dimensional coordinates of the defect detected for the infrared signal; Initial three-dimensional coordinates of the defect detected by the laser scattering signal; For the weight of the visible light coordinates, For the weight of the infrared coordinates, Is the laser scattering coordinate weight.
8. The lens defect detection system of claim 7, wherein the defect detection data packet further comprises a signal reliability parameter for the defect.
9. The lens defect detection system based on the photoelectric multi-fusion perception technology according to claim 1, wherein the capturing module is interactively connected with the registration module through a wireless network, the configuration module is interactively connected with the extraction module and the identification module through the wireless network, the identification module is interactively connected with the integration module through the wireless network, and the integration module is interactively connected with the feedback module through the wireless network.
10. A method for detecting a lens defect based on a photoelectric multi-fusion sensing technology, which is an implementation method for a lens defect detection system based on a photoelectric multi-fusion sensing technology as set forth in any one of claims 1 to 9, and is characterized by comprising the following steps: Collecting visible light, infrared and laser scattering original signals on the surface and inside of the lens, converting the signals into processable electric signals, distributing signal weights according to the material characteristics of the lens, the surface reflectivity and the heat conduction coefficient, and completing multi-source signal synchronous calibration; Performing adaptive denoising and min-max standardization processing on the calibrated electric signals, realizing multi-mode signal space alignment through rigid transformation based on three-dimensional coordinates of the lens reference mark points, and establishing signal mapping association of the same detection area; Mining gray scale, texture and three-dimensional geometric features corresponding to defects, screening key features according to an information gain criterion after normalization processing, and constructing a high-dimensional feature vector set containing defect differentiation information through kernel function mapping; Updating dynamic threshold values in real time according to the signal complexity factors, the feature vector set dispersion factors, the background uniformity factors and the feature vector set concentration factors, calculating fusion matching degree by means of hierarchical analysis of preliminary screening, fine matching and category confirmation and combining cosine similarity and mahalanobis distance, identifying defects and marking preliminary category attributes; Determining the three-dimensional accurate position of the defect based on the coordinate conversion relation between the CAD three-dimensional model of the lens and the multi-mode signal, integrating the defect type, position, signal intensity and credibility parameters, and generating a defect detection data packet; and feeding back the defect detection data packet to a preset production management terminal.

Description

Lens defect detection method and system based on photoelectric multi-fusion perception technology Technical Field The invention relates to the technical field of product defect detection, in particular to a lens defect detection method and system based on a photoelectric multi-fusion perception technology. Background Product defect detection is usually performed by collecting product images and extracting features to quickly identify defects such as surface flaws, dimensional deviations and the like. The method has the advantages of non-contact, high precision and high speed, is widely applied to quality control in manufacturing industries such as electronics, automobiles, 3C and the like, and realizes defect real-time screening and efficiency improvement. In the field of lens manufacturing such as optical filters, defects such as scratches, bubbles, uneven film layers and internal microcracks on the surface of the lens can directly influence optical indexes such as light transmittance and cut-off band performance, and higher requirements are put on the precision and dimension of defect detection. The invention patent application with the application number of 202111488051.X discloses a product defect detection method and device, and aims to solve the problems that in the training process of detecting the surface defects of a mobile phone, the unbalance of different defect sample types is considered, meanwhile, the generalization capability of a model is considered, different mobile phone projects generally need to construct a new detection model and retrain, so that the generalization capability of the model is weaker, in addition, the surface textures and colors of different mobile phones are different, and the model with stronger generalization capability is built. However, the related art relying on image recognition alone cannot satisfy the technical diversity of the lens defect detection scene. Especially for lens such as optical filter, the material is transparent, the surface is plated with multilayer optical film, and the hidden defects such as internal bubbles, film defects and microcracks are difficult to be effectively detected by a single visible light imaging means, so that the cooperative application of the multi-mode sensing technology is needed. Disclosure of Invention Aiming at the defects in the prior art, the invention provides a lens defect detection method and a lens defect detection system based on a photoelectric multi-fusion perception technology, which can effectively solve the problems in the prior art. In order to achieve the above object, the present invention is achieved by the following technical scheme; The invention discloses a lens defect detection system based on photoelectric multi-fusion perception technology, which comprises: The system comprises a lens, a capturing module, a registering module, a feedback module, an integrating module, a characteristic vector set, an identification module, a display module and a display module, wherein the capturing module is used for acquiring visible light, infrared and laser scattering signals on the surface and inside of the lens, converting the acquired signals into processable electric signals, the registering module is used for receiving the electric signals output by the capturing module, denoising, normalizing and spatially registering the electric signals, establishing signal mapping according to the spatial correspondence of the same detection area of the lens so as to finish preliminary association of multi-mode signals; The capture module is interactively connected with the registration module through a wireless network, the configuration module is interactively connected with the extraction module and the identification module through the wireless network, the identification module is interactively connected with the integration module through the wireless network, and the integration module is interactively connected with the feedback module through the wireless network. Further, the capturing module is integrated by a visible light sensor array, an infrared focal plane detector and a laser scattering detection unit; The sampling frequency of the visible light sensor array is synchronous with the frame frequency of the infrared focal plane detector, and the emission wavelength of the laser scattering detection unit and the characteristic absorption wavelength of the lens material are in a complementary relation; The capture module synchronously calibrates the collected multi-source original signals by the following formula: ; wherein: the synthesized electric signal after the calibration at the time t is obtained; 、、 the original electric signals corresponding to the visible light, infrared and laser scattering at the moment t; 、、 Is a visible light signal weight coefficient, an infrared signal weight coefficient and a laser scattering signal weight coefficient. Still further, the electrical signal denoising compliance