High-precision machine vision positioning system

Abstract

The invention, which relates to the technical field of machine vision, discloses a high-precision machine vision positioning system comprising the following components: an adaptive multispectral imaging module, an image data processing module, a positioning calculation module and a storage and control module. The self-adaptive multispectral imaging module is integrated with a visible light imaging subunit, an infrared imaging subunit, an ultraviolet imaging subunit, a self-adaptive optical adjustment subunit and a spectral band self-adaptive switching subunit. Optical parameters are adjusted in real time according to various factors such as ambient light conditions, target distances and target reflection characteristics, so that adaptive switching and cooperative imaging of spectral bands are realized, and the system can acquire comprehensive information of an object in multiple spectral bands; the problems of incomplete information acquisition, image distortion and the like of traditional single visible light wave band imaging in a complex environment are effectively solved.

Description

Technical Field

[0001] The invention relates to the technical field of machine vision, and in particular to a high-precision machine vision positioning system. Background Art

[0002] In many fields such as modern industrial production, logistics, and security, the demand for accurate positioning of objects is growing. As a key technology to achieve this goal, machine vision positioning system can realize automatic recognition and positioning of objects through image acquisition and analysis. With the continuous advancement of technology, machine vision positioning system has been widely used in various complex environments, providing strong support for industrial automation, intelligent manufacturing and other fields.

[0003] Traditional machine vision positioning systems mostly rely on visible light band imaging to analyze and locate objects. However, this single-band imaging method has many limitations. First, traditional machine vision positioning systems can only obtain limited visual information on the surface of objects, such as appearance shape, color, texture, etc., but cannot obtain deeper information. As a result, when facing complex environments and special positioning needs, such as positioning detection of thermal management of electronic components, it is impossible to accurately determine the specific location of components with abnormal heating. Secondly, factors such as ambient light conditions, target distance, and target reflection characteristics will have a significant impact on visible light imaging, causing image distortion and further affecting positioning accuracy. At the same time, changes in target distance and differences in target surface reflection characteristics will also affect imaging quality, causing deviations in positioning results. In addition, although traditional multispectral imaging technology can make up for the shortcomings of visible light imaging to a certain extent, it usually images different spectral bands in a fixed order or preset mode. This imaging method makes the imaging of each band relatively independent, and it is difficult to fully and accurately obtain relevant information about the object.

[0004] In view of the above problems, it is necessary to optimize the existing machine vision positioning system by integrating multi-spectral imaging capabilities, adjusting optical parameters in real time according to environmental and target factors, and realizing adaptive switching and collaborative imaging of spectral bands, so as to achieve accurate positioning of objects. Therefore, it is of great significance to develop a high-precision machine vision positioning system that can comprehensively realize the above characteristics. Summary of the invention

[0005] The purpose of the present invention is to make up for the shortcomings of the prior art and to provide a high-precision machine vision positioning system. It can integrate multi-spectral imaging capabilities such as visible light, infrared, and ultraviolet, adjust optical parameters in real time according to environmental and target factors, realize adaptive switching and collaborative imaging of spectral bands, and obtain comprehensive information of objects in different spectral bands. Combined with image processing and positioning algorithms, it can achieve precise positioning of objects, effectively overcoming the shortcomings of traditional machine vision positioning systems in complex environments and with special positioning requirements.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a high-precision machine vision positioning system, the system comprising the following components: an adaptive multi-spectral imaging module, an image data processing module, a positioning calculation module and a storage and control module;

[0007] The adaptive multi-spectral imaging module integrates visible light, infrared, ultraviolet imaging subunits, an adaptive optical adjustment subunit and a spectral band adaptive switching subunit. The visible light imaging subunit is responsible for collecting image information of the object to be located in the visible light band, and can clearly present the conventional visual features of the object's appearance, shape, color and texture, and provide basic visual information of the object's surface for subsequent positioning analysis. The infrared imaging subunit is used to collect infrared images of the object to be located to obtain the object's heat distribution information. By sensing the infrared radiation emitted by the object itself, the temperature difference in different areas of the object's surface is accurately detected, and then the heat distribution inside or on the surface of the object is presented. The ultraviolet imaging subunit is responsible for collecting images of the object to be located in the ultraviolet band, and can capture information related to the surface material characteristics of the object under ultraviolet light. The spectral band adaptive switching subunit integrates multiple types of sensors for real-time monitoring of the relative humidity of the object to be located. The adaptive optical adjustment subunit uses a high-precision light sensor to monitor the intensity and spectrum distribution parameters of ambient light in real time, and uses laser ranging and triangulation to accurately measure the distance between the object to be located and the imaging system. It emits light of a specific frequency and detects the characteristics of the light reflected by the object, analyzes the reflectivity of the object to be located and the reflection characteristic parameters of the reflection spectrum, and according to the calculated optical system parameter adjustment value, realizes real-time adjustment of the lens curvature and focal length parameters of the optical system through driving motors and piezoelectric elements.

[0008] The image data processing module receives and processes the multispectral image data from the adaptive multispectral imaging module, uses an adaptive filtering algorithm to perform denoising, and uses relevant technologies to perform enhancement operations according to the characteristics of images in different spectral bands. At the same time, a fusion algorithm based on deep learning is used to fuse the pre-processed image data of different spectral bands, and the fused data is analyzed to extract feature information related to positioning, and transmitted to the positioning calculation module;

[0009] The positioning calculation module performs positioning calculation to determine the precise position of the object to be positioned based on the feature information extracted by the image data processing module, constructs a pre-stored standard feature template by extracting features from sample images collected in advance, and after receiving the feature information of the object to be positioned, uses a positioning algorithm based on feature matching to calculate the similarity with the standard feature template, finds the most similar template, and combines the established unified coordinate system with a coordinate transformation calculation method that introduces correction parameters considering the perspective and posture change factors to obtain the precise position of the object to be positioned in the coordinate system;

[0010] The storage and control module includes a storage submodule and a control submodule. The storage submodule adopts a large-capacity, high-speed storage device to store the original image data collected by the adaptive multi-spectral imaging module, the intermediate data of the image data processing module and the positioning results of the positioning calculation module according to the format classification, and establishes an index to facilitate data management and query. The control submodule can control the imaging parameters of the adaptive multi-spectral imaging module and schedule the work processes of each module to ensure smooth data flow.

[0011] Furthermore, the spectral band adaptive switching subunit in the adaptive multispectral imaging module uses a preset algorithm and rule system to perform comprehensive analysis and judgment, and its algorithm formula is: Among them, S ij represents the i-th spectral band, T ij is the relevant impact value of the surface temperature of the object to be located measured by the temperature sensor in the i-th spectral band under the j-th combination of environment and object characteristics, T max The maximum value of the pre-set temperature-related influence value, I ij I is the influence of the ambient light intensity monitored by the light sensor on the imaging of the i-th spectral band under the j-th combination of environment and object characteristics. max The maximum value of the light intensity influence value is preset, H is the ambient humidity value monitored by the ambient humidity sensor, H max The maximum value of the preset ambient humidity, R ij is the influence value of the reflection characteristics of the object to be located detected by the reflection characteristic detection unit on the imaging of the i-th spectral band under the j-th combination of environment and object characteristics, R max The maximum value of the pre-set reflection characteristic influence value, M ij M is the influence of the material type of the object to be located detected by the material sensor on the imaging of the i-th spectral band under the j-th combination of environment and object characteristics. max Preset maximum value for material type influence.

[0012] Furthermore, the adaptive optical adjustment subunit in the adaptive multi-spectral imaging module realizes real-time adjustment of the lens curvature and focal length parameters of the optical system through a driving motor and a piezoelectric element, and the adjustment formula of the mirror curvature is: Where I is the light intensity value, D is the distance between the object to be located and the imaging system, R is the average reflectivity of the object to be located, and D z is the distance constant related to the propagation characteristics of light, and D1 is the periodic constant related to the distance.

[0013] Furthermore, the adaptive optical adjustment subunit in the adaptive multi-spectral imaging module realizes real-time adjustment of the lens curvature and focal length parameters of the optical system through a driving motor and a piezoelectric element, and the focal length adjustment formula is: Wherein, F represents the focal length of the optical system, I, D, and R have the same meanings as in the lens curvature adjustment formula, representing the ambient light intensity, the distance between the object to be located and the imaging system, and the average reflectivity of the object to be located, respectively, I0 is a reference intensity value related to the light intensity characteristics, and D2 is another distance constant related to the distance.

[0014] Furthermore, the image data processing module uses an adaptive filtering algorithm for denoising, and its calculation formula is: in, Represents the pixel value of the image at coordinate (x, y) after denoising. is the original pixel value of the input image at the coordinate (x, y), k and l define the half width of the filter window in the horizontal and vertical directions respectively, and determine the range of neighborhood pixels involved in the filter calculation, α i,j is the adaptive weight coefficient corresponding to the pixel with coordinates (i, j) in the filter window, and is calculated as follows: in, Represents the local variance of the pixel value at the coordinate (x+i, y+j) of the image, which is used to measure the discrete degree of the pixel value in the local area where the pixel is located, μ x+i,y+j It is the local mean of the pixel value of the image at the coordinate (x+i, y+j), which is used to reflect the average level of pixel values ​​in the local area where the pixel is located. σ0 and μ0 are the preset variance and mean reference values.

[0015] Furthermore, the image data processing module uses a fusion algorithm based on deep learning to fuse the pre-processed image data of different spectral bands. The formula of the fusion algorithm is: Among them, F x,yrepresents the pixel value of the fused image at the coordinate (x, y), which is obtained by weighted fusion of the pixel values ​​of images of different spectral bands at this coordinate and is used to construct a complete fused image. S represents the number of spectral bands, and w s is the weight coefficient corresponding to the sth spectral band, and its value is determined according to the importance of images of different spectral bands in positioning analysis and the characteristics of the current environment and the object to be located. It is the pixel value of the sth spectral band image at the coordinate (x, y), that is, the pixel value at this coordinate after the original image of different spectral bands collected from the adaptive multispectral imaging module is preprocessed, which serves as the basic data for fusion calculation.

[0016] Furthermore, the image data processing module uses a fusion algorithm based on deep learning to fuse the pre-processed image data of different spectral bands. In the fusion formula, when determining the weight coefficient w s When , the algorithm formula for dynamic adjustment is: Among them, α s , β s , γ s , δ s : is the adjustment parameter related to the sth spectral band, which is used to adjust the weight coefficient according to the characteristics of different spectral bands and their importance in different scenarios. s represents the environmental factor impact value related to the imaging of the sth spectral band, D0 is a reference value related to the light intensity, T s is the temperature influence value of the object to be located related to the imaging of the sth spectral band, T0 is a reference value related to temperature, R s is the reflection characteristic influence value of the object to be located related to the imaging of the sth spectral band, and R0 is the reference value related to the reflection characteristic.

[0017] Furthermore, the positioning calculation module uses a positioning algorithm based on feature matching to calculate the similarity with the standard feature template and find the most similar template. The algorithm formula is: Among them, S ij represents the similarity score between the i-th group of features of the object to be located and the j-th pre-stored standard feature template, F ik is the kth feature value in the i-th set of features of the object to be located, T jk is the kth feature value in the jth standard symbol template stored in advance, K represents the number of features used for matching, Δ is a preset feature difference tolerance value, which is used to limit the calculation range of the symbol difference to avoid excessive influence on the similarity score due to excessive differences in individual feature values. ik -T jk | When it exceeds Δ, take Δ as the calculated value, and based on the similarity score calculated above, use the formula Determine the standard feature template that best matches the features of the object to be located, where j match Represents the index value of the standard feature template that best matches the feature of the object to be located. By finding the similarity score S ij The largest j value determines the best matching template, argmax j is to make the following expression S ij An operator that takes the value of j at its maximum value.

[0018] Furthermore, the positioning calculation module combines the established unified coordinate system and the coordinate transformation calculation method that introduces correction parameters considering the viewing angle and posture change factors to obtain the precise position of the object to be positioned in the coordinate system. The transformation calculation formula is: Among them, x object and object They represent the precise x-coordinate and y-coordinate of the object to be located in the unified coordinate system, respectively. tenplate and lernplate is the known x-coordinate and y-coordinate of the standard feature template that best matches the object to be located in the unified coordinate system. and They represent the feature offset of the object to be located relative to the most matching standard feature template in the horizontal and vertical directions, and θ represents the rotation angle of the object to be located relative to the most matching standard feature template.

[0019] Compared with the existing technology, this high-precision machine vision positioning system has the following beneficial effects:

[0020] 1. The present invention integrates multi-spectral imaging capabilities such as visible light, infrared, and ultraviolet, and adjusts optical parameters in real time according to multiple factors such as ambient light conditions, target distance, and target reflection characteristics, thereby realizing adaptive switching and collaborative imaging of spectral bands, enabling the system to obtain comprehensive information about objects in multiple spectral bands, effectively overcoming the problems of incomplete information acquisition and image distortion in traditional single visible light band imaging in complex environments.

[0021] 2. The present invention uses adaptive filtering algorithms for denoising and image enhancement technology to output high-quality, high signal-to-noise ratio image data, providing a solid foundation for subsequent positioning calculations. At the same time, the fusion algorithm based on deep learning can automatically learn the intrinsic connections and feature representations between image data in different spectral bands, achieve accurate fusion of image data, and further extract feature information related to positioning. In the positioning calculation module, the system uses a positioning algorithm based on feature matching combined with calculation methods such as coordinate transformation, which can accurately determine the precise position of the object to be located, ensuring a high level of positioning accuracy.

[0022] Other advantages, objectives and features of the present invention will be set forth in part in the following description and, in part, will be apparent to those skilled in the art based on an examination of the following or may be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0024] Figure 1 It is a structural diagram of a high-precision machine vision positioning system;

[0025] Figure 2 The figure is a flow chart of a high-precision machine vision positioning system. DETAILED DESCRIPTION

[0026] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] Example

[0028] This embodiment describes in detail the positioning of components on an electronic component production line using a high-precision machine vision positioning system. Various tiny electronic components can be accurately positioned through the present invention.

[0029] As the front-end image acquisition part of the entire positioning system, the adaptive multispectral imaging module is responsible for collecting comprehensive and clear multispectral image data according to the complex environment of the electronic component production line and the characteristics of the electronic components themselves. It includes visible light, infrared, and ultraviolet imaging subunits, adaptive optical adjustment subunits, and spectral band adaptive switching subunits. The visible light imaging subunit is equipped with a high-resolution visible light camera and an adaptive optical lens system. Based on the reflection principle of visible light by electronic components, it collects images of components in the visible light band and presents conventional visual features such as the appearance shape, color, and texture of the components. In the production line environment, these feature information helps to preliminarily identify the basic form and surface condition of electronic components, and provide basic visual data for subsequent more accurate positioning analysis. In addition, during operations such as collaborative imaging, the imaging parameters such as exposure time and gain can be dynamically adjusted according to instructions from the control submodule to ensure accurate positioning. In order to ensure effective coordination with imaging in other bands and obtain better quality images, the infrared imaging subunit is equipped with a highly sensitive infrared detector and infrared optical lens. It collects infrared images of electronic components based on the principle of thermal radiation, thereby obtaining the heat distribution information of the components. In the production process of electronic components, some components have temperature differences due to their working conditions or their own characteristics. Infrared imaging is very critical for detecting these abnormal heating parts or identifying components with different thermal characteristics, and can provide an important basis for precise positioning. The ultraviolet imaging subunit uses specially designed ultraviolet optical lenses and image sensors sensitive to ultraviolet light to collect images of electronic components in the ultraviolet band and capture information related to the material characteristics of the component surface. Different electronic component materials may exhibit unique reflection or absorption characteristics in the ultraviolet band. Ultraviolet imaging can obtain these material-related feature information, further enriching the image data content required for positioning.

[0030] The adaptive optical adjustment subunit uses high-precision sensors to monitor the ambient light intensity, spectral distribution and other parameters of the electronic component production workshop in real time, and quickly transmits this information to the optical parameter adjustment unit. At the same time, it uses appropriate technologies such as laser ranging and triangulation to accurately measure the distance between the electronic component to be located and the imaging system, and transmits the measurement results to the optical parameter adjustment unit in the form of digital signals. In addition, by emitting light of a specific frequency and detecting the characteristics of the light reflected back by the electronic component, the average reflectivity, reflection spectrum and other reflection characteristic parameters of the electronic component to be located are analyzed and obtained, and these detection data are transmitted to the optical parameter adjustment unit in a timely manner. The surface materials and process treatments of different electronic components may cause their reflection characteristics to vary. Understanding these reflection characteristics helps to adjust the optical system in a targeted manner to optimize the imaging quality. Based on the light information received from the light sensor, the distance data of the distance measurement unit and the reflection characteristic parameters of the reflection characteristic detection unit, the preset optical parameter adjustment algorithm is used for calculation to adjust the lens curvature and focal length and other parameters of the optical system of each imaging subunit in real time. The lens curvature C adjustment formula is: Where I is the light intensity value. The difference in light intensity in different areas of the workshop will affect the adjustment of the lens curvature. D is the distance between the electronic component to be located and the imaging system. The farther the distance, the more serious the light divergence. R is the average reflectivity of the electronic component to be located. Electronic components of different materials and surface treatments have different reflectivities. Components with high reflectivity may require special lens curvature settings to optimize the imaging effect. D z is a distance constant related to the light propagation characteristics, which is predetermined based on the characteristics of the optical system and the typical light propagation conditions in the electronic component production workshop. D1 is a periodic constant related to the distance, which is also predetermined based on the characteristics of the optical system and the production workshop environment. The focal length (F) adjustment formula is: Among them, F represents the focal length of the optical system. Accurately setting the focal length is crucial for clearly presenting the electronic components to be located on the imaging plane. The meanings of I, D, and R are the same as those in the lens curvature adjustment formula, representing the ambient light intensity, the distance between the electronic components to be located and the imaging system, and the average reflectivity of the electronic components to be located, respectively. I0 is a reference intensity value related to the light intensity characteristics, which is predetermined according to the sensitivity of the optical system to light intensity and the common light intensity range in electronic component production workshops. D2 is another distance constant related to the distance.

[0031] The spectral band adaptive switching subunit integrates multiple sensors to comprehensively monitor the environment of the electronic component production line and the characteristics of the electronic components themselves. Among them, the material sensor is used to detect the material type of the electronic components, the temperature sensor is used to measure the surface temperature of the electronic components, and the surface roughness sensor is used to obtain the roughness information of the surface of the electronic components, etc., to monitor the characteristics of the electronic components. At the same time, a light sensor is set to sense the ambient light intensity in the workshop, and a humidity sensor is used to monitor the ambient humidity in the workshop. In combination with the image analysis algorithm, the image data of the visible light imaging subunit is used to analyze the background interference in real time, etc., to monitor environmental factors, and use the pre-set complex algorithm and rule system to comprehensively analyze and judge the monitoring data. The specific judgment is made through the following formula: Among them, S ij represents the imaging suitability score of the i-th spectral band (i=1 represents the visible light band, i=2 represents the infrared band, and i=3 represents the ultraviolet band) under the j-th combination of environment and object characteristics. The higher the score, the more suitable the spectral band is for imaging under the current situation. The most suitable spectral band combination and collaborative imaging strategy are determined by comparing the scores of different bands. ij is the impact value of the surface temperature of the electronic component measured by the temperature sensor in the i-th spectral band under the j-th combination of environment and object characteristics, Tmax The maximum value of the pre-set temperature-related influence value, I ij I is the influence of the ambient light intensity monitored by the light sensor on the imaging of the i-th spectral band under the j-th combination of environment and object characteristics. max The maximum value of the light intensity influence value is preset, H is the ambient humidity value monitored by the ambient humidity sensor, H maπ The maximum value of the preset ambient humidity, R ij is the i-th spectral band under the j-th combination of environment and object characteristics, R max The maximum value of the pre-set reflection characteristic influence value, M ij is the influence of the material type of the electronic component detected by the material sensor on the imaging of the i-th spectral band under the j-th combination of environment and object characteristics, M max The maximum value of the influence value of the material type is preset, and according to the decision result, precise control instructions are issued to the visible light imaging subunit, infrared imaging subunit and ultraviolet imaging subunit, including starting or stopping the imaging operation of a certain imaging subunit, dynamically adjusting the imaging parameters of each imaging subunit (such as exposure time, gain, detector sensitivity, ultraviolet light excitation intensity, etc.), and controlling the timing and method of collaborative imaging of each imaging subunit at the same time, so as to realize adaptive switching of spectral bands and collaborative imaging of different bands at the same time, so as to obtain comprehensive and targeted multispectral image data of electronic components.

[0032] The image data processing module uses an adaptive filtering algorithm for denoising. The core algorithm formula is: in, Represents the pixel value of the image at the coordinate (x; y) after denoising. is the original pixel value of the input image (i.e., the original image of each spectral band collected by the adaptive multispectral imaging module) at the coordinate (x, y), k and l define the half width of the filter window in the horizontal and vertical directions, respectively, and determine the range of neighborhood pixels involved in the filter calculation, α i,j is the adaptive weight coefficient corresponding to the pixel with coordinates (i, j) in the filter window, and is calculated as: in, Represents the local variance of the pixel value at the coordinate (x+i, y+j) of the image, which is used to measure the discrete degree of the pixel value in the local area where the pixel is located, μ x+i,y+j is the local mean of the pixel value of the image at the coordinate (x+i, y+j), σ0 and μ0 are the pre-set variance and mean reference values ​​for normalization, so that α i,jThe value range of is within the appropriate interval, so as to better adjust the weight coefficient according to the characteristics of different areas of the image, achieve adaptive denoising effect, and use the backbone multi-spectral feature weighted fusion algorithm to fuse image data of different spectral bands. The formula is Among them, F x,y represents the pixel value of the fused image at the coordinate (x:y), which is obtained by weighted fusion of the pixel values ​​of images of different spectral bands at the coordinate and is used to construct a complete fused image. S represents the number of spectral bands. In the present invention, considering the three spectral bands of visible light, infrared, and ultraviolet, S=3, w s is the weight coefficient corresponding to the sth spectral band, and its value is determined according to the importance of images of different spectral bands in positioning analysis and the characteristics of the current environment and the electronic components to be located. is the pixel value of the sth spectral band image at the coordinate (x; y), that is, the pixel value of the original image of different spectral bands collected from the adaptive multispectral imaging module at this coordinate after preprocessing, which is used as the basic data for fusion calculation and used to determine the weight coefficient w s When , we can further use the formula To dynamically adjust according to environmental factors and object characteristics, where α s ,β s , γ s , δ s : is the adjustment parameter related to the sth spectral band, which is used to adjust the weight coefficient according to the characteristics of different spectral bands and their importance in different scenarios. s represents the environmental factor impact value related to the imaging of the sth spectral band, D0 is the reference value related to the light intensity, T s is the temperature influence value of the electronic component to be located related to the imaging of the sth spectral band, T0: is a reference value related to temperature, R s : is the reflection characteristic influence value of the electronic component to be located related to the imaging of the sth spectral band. The reflection characteristics of different electronic components are different in different spectral bands. This value reflects the influence of the reflection characteristics on the imaging quality of this band, which in turn affects its weight in the fusion. R0 is a reference value related to the reflection characteristics. Based on the fused image data, further analysis is performed to extract feature information related to positioning, such as shape features of electronic components (such as the curvature of the contour, the ratio of the side length, etc.), texture features (such as the direction and frequency of the texture), and heat distribution features (if it involves heat information obtained by infrared band imaging). These feature information will be transmitted to the positioning calculation module to provide key data support for subsequent positioning calculations.

[0033] The positioning calculation module collects a large number of sample images for different types of electronic components, and extracts feature information from these sample images through the image data processing module, such as shape features of electronic components (such as the curvature of the contour, the ratio of the side length, etc.), texture features (such as the direction and frequency of the texture, etc.), heat distribution features (if it involves heat information obtained by infrared band imaging), etc., and organizes and normalizes these feature information to construct pre-stored standard feature templates, which are stored in the database inside the module. These standard feature templates will be used as a reference for subsequent matching with the features of the electronic components to be located. After receiving the feature information of the electronic components to be located transmitted from the image data processing module, it is matched with the pre-stored standard feature templates, and the following positioning algorithm formula based on feature matching is used to calculate the similarity between the features of the electronic components to be located and the standard feature templates: Among them, S ij It represents the similarity score between the i-th group of features of the electronic component to be located and the j-th pre-stored standard feature template. The higher the score, the more similar the electronic component to be located is to the standard feature template. By comparing the similarity scores between the electronic component to be located and each standard feature template, the best matching template can be found, and then the position of the electronic component can be determined. ik The kth eigenvalue in the i-th group of features of the electronic component to be located is extracted by processing the collected multispectral image data through the image data processing module, which depends on the imaging band and the characteristics of the electronic component. jk The kth eigenvalue in the jth pre-stored standard feature template is constructed by collecting a large number of sample images for different types of electronic components in the early stage, extracting features through the image data processing module, and then arranging and normalizing them. It is stored in the database inside the module and is used to match the features of the electronic components to be located. K represents the number of features used for matching, that is, the number of eigenvalues ​​contained in each set of features. Δ is a preset feature difference tolerance value, which is used to limit the calculation range of feature differences to avoid excessive influence on the similarity score due to excessive differences in individual eigenvalues. When |F ik -T jk | When it exceeds Δ, take Δ as the calculation value, and based on the similarity score calculated above, use the formula Determine the standard feature template that best matches the features of the object to be located, where j match Represents the index value of the standard feature template that best matches the feature of the electronic component to be located. By finding the similarity score S ij The largest j value determines the best matching template, argmax j is a method to make the following expression (here S ij) obtains the maximum value of j, that is, among all possible standard feature templates (index is j), find the index of the template with the highest similarity to the feature of the electronic component to be located. During the imaging process, a unified coordinate system is established. The coordinate system covers the imaging area and the possible position range of the electronic component to be located. The origin of the coordinate system, the direction of the coordinate axis and other parameters are set when the system is initialized and kept consistent. During the imaging process, a unified coordinate system is established. The coordinate system covers the imaging area and the possible position range of the electronic component to be located. The origin of the coordinate system, the direction of the coordinate axis and other parameters are set when the system is initialized and kept consistent. According to the matching results and the coordinate system, the following coordinate transformation calculation formula is used to determine the precise position of the electronic component to be located in the unified coordinate system: Among them, x object and object They represent the precise x-coordinate and y-coordinate of the electronic component to be located in the unified coordinate system, which are the final position coordinates of the electronic component to be located. tenplate and lernplate It is the known x-coordinate and y-coordinate of the standard feature template that best matches the electronic component to be located in the unified coordinate system. These coordinates are determined and stored when the standard feature template is constructed. Because the standard feature template has its fixed position information in the coordinate system, the position of the electronic component to be located is inferred through the position information of this template that matches the electronic component to be located. and They respectively represent the feature offsets of the electronic component to be located in the horizontal and vertical directions relative to the most matching standard feature template (the units are consistent with the units of the coordinate system, such as meters, pixels, etc.). These offsets are determined by analyzing the feature differences between the electronic component to be located and the matching template. For example, the shape features of the electronic component may have certain changes such as scaling, translation or rotation. The offsets in the horizontal and vertical directions are obtained by quantitative analysis of these feature changes. θ represents the rotation angle of the electronic component to be located relative to the most matching standard feature template (in radians). Similarly, by analyzing the feature differences between the electronic component to be located and the matching template, especially when it comes to changes in the posture of the electronic component, the rotation of the electronic component relative to the template is determined, so that the rotation factor can be accurately considered in the coordinate transformation calculation, thereby more accurately calculating the position of the electronic component to be located.

[0034] The storage and control module consists of a storage submodule and a control submodule, which is responsible for the data storage, operation control and coordination of the entire system. The storage submodule uses large-capacity, high-speed storage devices, such as solid-state drives (SSDs) or large-capacity hard disks (HDDs). The stored data types include the original image data collected by the adaptive multispectral imaging module, the intermediate data in the image data processing module, and the positioning results calculated by the positioning calculation module. For different types of data, appropriate file formats are used for storage. For example, the original image data can use common image file formats (such as JPEG, PNG, etc.), and the intermediate data can use custom data formats for subsequent analysis and processing. The positioning results can use text file formats to record the location coordinates of electronic components and other information. In order to facilitate data query, call and management, a complete data management system is established, and each piece of stored data is numbered and an index table is established. The required data can be quickly located through the index table. At the same time, the data is classified and stored according to the data type, acquisition time, processing stage and other attributes to improve the efficiency of data retrieval. The control submodule can control the imaging parameters of the adaptive multispectral imaging module. Such as exposure time, gain, aperture size, etc. According to different imaging requirements and environmental conditions, these parameters are adjusted by sending corresponding control instructions to the adaptive multispectral imaging module to ensure the acquisition of high-quality image data. At the same time, it is responsible for scheduling the workflow of each module of the system. When the system starts, each module is started in turn according to the predetermined order, and the data flow between the modules is ensured to be smooth. For example, the adaptive multispectral imaging module is triggered to perform imaging first, and then the positioning calculation module is started to perform positioning calculation after the image data processing module has processed the image data. During the operation of the system, the working order and working mode of each module can be flexibly adjusted according to the real-time situation (such as abnormal conditions or the need to re-collect images, etc.). In addition, the operating status of the system is monitored in real time, including the working status of each module (such as whether it is working normally, whether there is a fault, etc.), the usage of system resources (such as CPU usage, memory occupancy, etc.) and the status of data transmission (such as whether the data is successfully transmitted, the transmission speed, etc.). By monitoring these status information, problems can be discovered in time and corresponding measures can be taken to deal with them, such as restarting the faulty module and adjusting the allocation of system resources, so as to ensure that the system can run continuously and stably.

[0035] In summary, through the close collaboration of the above four modules, the high-precision machine vision positioning system of the present invention can accurately locate various tiny electronic components on the electronic component production line, provide precise position information for subsequent assembly, detection and other processes, and effectively improve production efficiency and product quality.

[0036] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A high-precision machine vision positioning system, characterized in that: The system includes the following components: adaptive multi-spectral imaging module, image data processing module, positioning calculation module and storage and control module; The adaptive multi-spectral imaging module integrates visible light, infrared, ultraviolet imaging subunits, an adaptive optical adjustment subunit and a spectral band adaptive switching subunit. The visible light imaging subunit is responsible for collecting image information of the object to be located in the visible light band, and can clearly present the conventional visual features of the object's appearance, shape, color and texture, and provide basic visual information of the object's surface for subsequent positioning analysis. The infrared imaging subunit is used to collect infrared images of the object to be located to obtain the object's heat distribution information. By sensing the infrared radiation emitted by the object itself, the temperature difference in different areas of the object's surface is accurately detected, and then the heat distribution inside or on the surface of the object is presented. The ultraviolet imaging subunit is responsible for collecting images of the object to be located in the ultraviolet band, and can capture information related to the surface material characteristics of the object under ultraviolet light. The spectral band adaptive switching subunit integrates multiple types of sensors for real-time monitoring of the relative humidity of the object to be located. The adaptive optical adjustment subunit uses a high-precision light sensor to monitor the intensity and spectrum distribution parameters of ambient light in real time, and uses laser ranging and triangulation to accurately measure the distance between the object to be located and the imaging system. It emits light of a specific frequency and detects the characteristics of the light reflected by the object, analyzes the reflectivity of the object to be located and the reflection characteristic parameters of the reflection spectrum, and according to the calculated optical system parameter adjustment value, realizes real-time adjustment of the lens curvature and focal length parameters of the optical system through driving motors and piezoelectric elements. The image data processing module receives and processes the multispectral image data from the adaptive multispectral imaging module, uses an adaptive filtering algorithm to perform denoising, and uses relevant technologies to perform enhancement operations according to the characteristics of images in different spectral bands. At the same time, a fusion algorithm based on deep learning is used to fuse the pre-processed image data of different spectral bands, and the fused data is analyzed to extract feature information related to positioning, and transmitted to the positioning calculation module; The positioning calculation module performs positioning calculation to determine the precise position of the object to be positioned based on the feature information extracted by the image data processing module, constructs a pre-stored standard feature template by extracting features from sample images collected in advance, and after receiving the feature information of the object to be positioned, uses a positioning algorithm based on feature matching to calculate the similarity with the standard feature template, finds the most similar template, and combines the established unified coordinate system with a coordinate transformation calculation method that introduces correction parameters considering the perspective and posture change factors to obtain the precise position of the object to be positioned in the coordinate system; The storage and control module includes a storage submodule and a control submodule. The storage submodule adopts a large-capacity, high-speed storage device to store the original image data collected by the adaptive multi-spectral imaging module, the intermediate data of the image data processing module and the positioning results of the positioning calculation module according to the format classification, and establishes an index to facilitate data management and query. The control submodule can control the imaging parameters of the adaptive multi-spectral imaging module and schedule the work processes of each module to ensure smooth data flow.

2. A high-precision machine vision positioning system according to claim 1, characterized in that: The spectral band adaptive switching subunit in the adaptive multispectral imaging module uses a preset algorithm and rule system to perform comprehensive analysis and judgment, and its algorithm formula is: Among them, S ij represents the i-th spectral band, T ij is the relevant impact value of the surface temperature of the object to be located measured by the temperature sensor in the i-th spectral band under the j-th combination of environment and object characteristics, T max The maximum value of the pre-set temperature-related influence value, I ij I is the influence of the ambient light intensity monitored by the light sensor on the imaging of the i-th spectral band under the j-th combination of environment and object characteristics. max The maximum value of the light intensity influence value is preset, H is the ambient humidity value monitored by the ambient humidity sensor, H max The maximum value of the preset ambient humidity, R ij is the influence value of the reflection characteristics of the object to be located detected by the reflection characteristic detection unit on the imaging of the i-th spectral band under the j-th combination of environment and object characteristics, R max The maximum value of the pre-set reflection characteristic influence value, M ij M is the influence of the material type of the object to be located detected by the material sensor on the imaging of the i-th spectral band under the j-th combination of environment and object characteristics. max Preset maximum value for material type influence.

3. The high-precision machine vision positioning system according to claim 1, characterized in that: The adaptive optical adjustment subunit in the adaptive multi-spectral imaging module realizes real-time adjustment of the lens curvature and focal length parameters of the optical system through the driving motor and the piezoelectric element. The adjustment formula of the mirror curvature is: Where I is the light intensity value, D is the distance between the object to be located and the imaging system, R is the average reflectivity of the object to be located, and D z is the distance constant related to the propagation characteristics of light, and D1 is the periodic constant related to the distance.

4. A high-precision machine vision positioning system according to claim 1, characterized in that: The adaptive optical adjustment subunit in the adaptive multi-spectral imaging module realizes real-time adjustment of the lens curvature and focal length parameters of the optical system through the driving motor and the piezoelectric element. The focal length adjustment formula is: Wherein, F represents the focal length of the optical system, I, D, and R have the same meanings as in the lens curvature adjustment formula, representing the ambient light intensity, the distance between the object to be located and the imaging system, and the average reflectivity of the object to be located, respectively, I0 is a reference intensity value related to the light intensity characteristics, and D2 is another distance constant related to the distance.

5. The high-precision machine vision positioning system according to claim 1, characterized in that: The image data processing module uses an adaptive filtering algorithm for denoising, and its calculation formula is: in, Represents the pixel value of the image at coordinate (x, y) after denoising. is the original pixel value of the input image at the coordinate (x, y), k and l define the half width of the filter window in the horizontal and vertical directions respectively, and determine the range of neighborhood pixels involved in the filter calculation, α i,j is the adaptive weight coefficient corresponding to the pixel with coordinates (i, j) in the filter window, and is calculated as follows: in, Represents the local variance of the pixel value at the coordinate (x+i, y+j) of the image, which is used to measure the discrete degree of the pixel value in the local area where the pixel is located, μ x+i,y+j It is the local mean of the pixel value of the image at the coordinate (x+i, y+j), which is used to reflect the average level of pixel values ​​in the local area where the pixel is located. σ0 and μ0 are the preset variance and mean reference values.

6. The high-precision machine vision positioning system according to claim 1, characterized in that: The image data processing module uses a fusion algorithm based on deep learning to fuse the pre-processed image data of different spectral bands. The formula of the fusion algorithm is: Among them, F x,y represents the pixel value of the fused image at the coordinate (x, y), which is obtained by weighted fusion of the pixel values ​​of images of different spectral bands at this coordinate and is used to construct a complete fused image. S represents the number of spectral bands, and w s is the weight coefficient corresponding to the sth spectral band, and its value is determined according to the importance of images of different spectral bands in positioning analysis and the characteristics of the current environment and the object to be located. It is the pixel value of the sth spectral band image at the coordinate (x, y), that is, the pixel value at this coordinate after the original image of different spectral bands collected from the adaptive multispectral imaging module is preprocessed, which serves as the basic data for fusion calculation.

7. A high-precision machine vision positioning system according to claim 6, characterized in that: The image data processing module uses a fusion algorithm based on deep learning to fuse the pre-processed image data of different spectral bands. In the fusion formula, the weight coefficient w is determined. s When , the algorithm formula for dynamic adjustment is: Among them, α s , β s , γ s , δ s : is the adjustment parameter related to the sth spectral band, which is used to adjust the weight coefficient according to the characteristics of different spectral bands and their importance in different scenarios. s represents the environmental factor impact value related to the imaging of the sth spectral band, D0 is a reference value related to the light intensity, T s is the temperature influence value of the object to be located related to the imaging of the sth spectral band, T0 is a reference value related to temperature, R s is the reflection characteristic influence value of the object to be located related to the imaging of the sth spectral band, and R0 is the reference value related to the reflection characteristic.

8. The high-precision machine vision positioning system according to claim 1, characterized in that: The positioning calculation module uses a positioning algorithm based on feature matching to calculate the similarity with the standard feature template and find the most similar template. The algorithm formula is: Among them, S ij represents the similarity score between the i-th group of features of the object to be located and the j-th pre-stored standard feature template, F ik is the kth feature value in the i-th set of features of the object to be located, T jk is the kth feature value in the jth standard symbol template stored in advance, K represents the number of features used for matching, Δ is a preset feature difference tolerance value, which is used to limit the calculation range of the symbol difference to avoid excessive influence on the similarity score due to excessive differences in individual feature values. ik -T jk | When it exceeds Δ, take Δ as the calculated value, and based on the similarity score calculated above, use the formula Determine the standard feature template that best matches the features of the object to be located, where j match Represents the index value of the standard feature template that best matches the feature of the object to be located. By finding the similarity score S ij The largest j value determines the best matching template, argmax j is to make the following expression S ij An operator that takes the value of j at its maximum value.

9. The high-precision machine vision positioning system according to claim 1, characterized in that: The positioning calculation module combines the established unified coordinate system and the coordinate transformation calculation method that introduces correction parameters considering the viewing angle and posture change factors to obtain the precise position of the object to be positioned in the coordinate system. The transformation calculation formula is: and object They represent the precise x-coordinate and y-coordinate of the object to be located in the unified coordinate system, respectively. tenplate and lernplate is the known x-coordinate and y-coordinate of the standard feature template that best matches the object to be located in the unified coordinate system. and They represent the feature offset of the object to be located relative to the most matching standard feature template in the horizontal and vertical directions, and θ represents the rotation angle of the object to be located relative to the most matching standard feature template.

Images