System and method of enhancing coupling quality of light spots at output end of silicon photonic wafers using artificial intelligence multitask image recognition

Abstract

A method of enhancing coupling quality of light spots at an output end of silicon photonic wafers using artificial intelligence multitask image recognition is proposed. The method includes performing a light spot image capturing step, an image preprocessing step, a light spot segmentation and feature extraction step, an intensity determining model establishing step, a model training and testing step and a practical application step. The intensity determining model establishing step includes establishing a mapping relationship between key features and light intensities of the light spots to establish an intensity determining model. The model training and testing step includes training the intensity determining model based on a training data set and testing the intensity determining model based on a testing data set to obtain a trained and tested intensity determining model with artificial intelligence. The trained and tested intensity determining model is configured to determine a new light spot image. Therefore, the present disclosure can combine an optical imaging technique and an artificial intelligence algorithm to provide a new way of faster, more accurate and non-contact light intensity measurement.

Description

System and methods for strengthening the coupled light quality of light spots at the exit end of silicon photonic wafers using artificial intelligence multitasking image recognition technology The invention relates to a system and a method thereof for strengthening the coupled light quality of the light spot at the exit end of a silicon photonic wafer and, in particular, a system and a method for strengthening the coupled light quality of the light spot at the exit end of a silicon photonic wafer using artificial intelligence (AI) multitasking image recognition technology. Silicon Photonics is an integrated optical path, which is composed of electrons and photons combined. In the testing process of silicon photonic wafers, “end-to-end” testing methods are common practice, in which light is coupled in from one end of the wafer waveguide and then output from the other end. The commonly known test uses photodetectors (e.g., photodiodes, Avalanche Photodiode (APD)) to collect the output light and convert the light power into an electrical signal. By measuring the intensity of the electrical signal, the power of the output light can be calculated. However, the detection time of this well-known silicon photonic wafer testing method is too long, resulting in poor testing efficiency. It can be seen that there is a lack of a system and method on the market that can quickly, accurately and non-contact light intensity measurement using artificial intelligence multitasking image recognition technology to strengthen the coupled light quality of the light spot at the exit end of silicon photonic wafers. The object of the present invention is to provide a system and a method for strengthening the coupled light quality of the light spot at the exit end of a silicon photonic wafer using artificial intelligence multitasking image recognition technology, and a method thereof combining optical imaging and artificial intelligence algorithms for analyzing the light spot at the exit end of a silicon photonic wafer. And through the analysis of the shape, size, symmetry and other features of the light spot, the transmission mode, coupling efficiency and other performance of the waveguide can be evaluated and its strength can be judged. The system of the invention and the method thereof provide a new avenue for fast, accurate and non-contact measurement of light intensity for the performance evaluation of wafer waveguides to solve the well-known problem of excessive detection time consuming by silicon photonic wafer testing methods. According to one embodiment of the present invention, a system for enhancing the light coupling quality of a light spot at the exit of a silicon photonics wafer using artificial intelligence multiplexing image recognition technology is provided. The system includes an image capture device and a processing device. The image capture device captures a light spot at an exit of a silicon photonics wafer to generate a first light spot image and a second light spot image. The processing device is electrically connected to the image capture device and obtains the first light spot image and the second light spot image. The processing device is configured to implement the following operations: performing a preprocessing operation on the first light spot image to generate a preprocessed light spot image; using an image segmentation algorithm to extract a light spot area of ​​the preprocessed light spot image, and extracting a key feature from the light spot area to quantify an intensity distribution of a light spot; establishing a mapping relationship between the key feature and a light intensity of the light spot to establish an intensity interpretation model corresponding to the first light spot image; and performing a model training and testing operation, wherein the model training and testing operation includes training the intensity interpretation model according to a training data set and testing the intensity interpretation model according to a test data set to obtain a trained test intensity interpretation model with artificial intelligence, and the trained test intensity interpretation model is used to interpret the second light spot image. Other embodiments of the aforementioned embodiment are as follows: The aforementioned image capture device includes one of a charge-coupled device (CCD) image sensor and a complementary metal oxide semiconductor (CMOS) image sensor. The image capture device has a dynamic range, a linear response, a magnification, and a field of view to capture changes in light intensity, and covers an output end face of the silicon photonics wafer. The output end face corresponds to the exit end. Other embodiments of the aforementioned embodiment are as follows: the aforementioned preprocessing operation includes at least one of a background correction operation and a denoising operation. The image segmentation algorithm includes one of a threshold segmentation method and an edge detection method. The key feature includes one of a brightness mean, a peak, and a standard deviation. Other examples of the aforementioned embodiment are as follows: The aforementioned intensity interpretation model includes one of a physical-based analytical model and a machine-learning-based data-driven model. The machine-learning-based data-driven model includes one of a regression model and a neural network model. The regression model includes one of a simplified swarm optimization (SSO), a linear regression, and a support vector machine. The neural network model includes a convolutional neural network. Model training and testing operations are repeatedly performed to obtain an optimized trained and tested intensity interpretation model. Other embodiments of the aforementioned embodiment are as follows: the aforementioned image capture device captures the light spot at a first time to generate a first light spot image, and captures the light spot at a second time to generate a second light spot image. The first time and the second time are different, and the light coupling quality of the light spot corresponds to the light intensity. According to one embodiment of the method sample of the present invention, a method for enhancing the light coupling quality of the light spot at the exit end of a silicon photonic wafer using artificial intelligence multiplexed image recognition technology is provided, which includes: using an image capture device to capture a light spot at an exit end of a silicon photonic wafer to generate a first light spot image and a second light spot image; using a processing device to perform a preprocessing operation on the first light spot image to generate a preprocessed light spot image; using an image segmentation algorithm by the processing device to extract a light spot area of ​​the preprocessed light spot image, and extracting a key feature from the light spot area to quantify an intensity distribution of a light spot; using the processing device to establish a mapping relationship between the key feature and a light intensity of the light spot to establish an intensity judgment model corresponding to the first light spot image; and using the processing device to perform a model training and testing operation, wherein the model training and testing operation includes training the intensity judgment model based on a training data set and testing the intensity judgment model based on a test data set to obtain a trained test intensity judgment model with artificial intelligence, and the trained test intensity judgment model is used to judge the second light spot image. Other embodiments of the aforementioned embodiment are as follows: The aforementioned image capture device includes one of a charge-coupled device (CCD) image sensor and a complementary metal oxide semiconductor (CMOS) image sensor. The image capture device has a dynamic range, a linear response, a magnification, and a field of view to capture changes in light intensity, and covers an output end face of the silicon photonics wafer. The output end face corresponds to the exit end. Other embodiments of the aforementioned embodiment are as follows: the aforementioned preprocessing operation includes at least one of a background correction operation and a denoising operation. The image segmentation algorithm includes one of a threshold segmentation method and an edge detection method. The key feature includes one of a brightness mean, a peak, and a standard deviation. Other examples of the aforementioned embodiment are as follows: The aforementioned intensity interpretation model includes one of a physical-based analytical model and a machine-learning-based data-driven model. The machine-learning-based data-driven model includes one of a regression model and a neural network model. The regression model includes one of a simplified swarm optimization (SSO), a linear regression, and a support vector machine. The neural network model includes a convolutional neural network. Model training and testing operations are repeatedly performed to obtain an optimized trained and tested intensity interpretation model. Other embodiments of the aforementioned embodiment are as follows: the aforementioned image capture device captures the light spot at a first time to generate a first light spot image, and captures the light spot at a second time to generate a second light spot image. The first time and the second time are different, and the light coupling quality of the light spot corresponds to the light intensity. According to another embodiment of the method of the present invention, a method for enhancing the light coupling quality of the light spot at the exit end of a silicon photonic wafer by using artificial intelligence multiplexing image recognition technology is provided, which comprises: using an image capture device to capture multiple light spots at multiple exit ends of a silicon photonic wafer to generate a first light spot image and a second light spot image; using a processing device to perform a preprocessing operation on the first light spot image to generate a preprocessed light spot image; using an image segmentation algorithm by the processing device to extract multiple light spot areas of the preprocessed light spot image, and extracting complex light spot images from these light spot areas. The invention relates to a method for determining a plurality of key features to quantify the complex intensity distribution of these light spots; establishing a complex mapping relationship between these key features and the complex light intensities of these light spots by a processing device to establish an intensity interpretation model corresponding to the first light spot image; and performing a model training and testing operation by the processing device, wherein the model training and testing operation includes training the intensity interpretation model according to a training data set and testing the intensity interpretation model according to a test data set to obtain a trained test intensity interpretation model with artificial intelligence, and the trained test intensity interpretation model is used to interpret the second light spot image. Other embodiments of the aforementioned embodiment are as follows: The image capture device includes one of a charge-coupled device (CCD) image sensor and a complementary metal oxide semiconductor (CMOS) image sensor. The image capture device has a dynamic range, a linear response, a magnification, and a field of view to capture these light intensity variations, and covers an output end face of the silicon photonics wafer. The output end face corresponds to the exit end. Other embodiments of the aforementioned embodiment are as follows: the aforementioned preprocessing operation includes at least one of a background correction operation and a denoising operation. The image segmentation algorithm includes one of a threshold segmentation method and an edge detection method, and each key feature includes one of a brightness mean, a peak value, and a standard deviation. Other examples of the aforementioned embodiment are as follows: The aforementioned intensity interpretation model includes one of a physical-based analytical model and a machine-learning-based data-driven model; the machine-learning-based data-driven model includes one of a regression model and a neural network model; the regression model includes one of a simplified swarm optimization (SSO) algorithm, a linear regression algorithm, and a support vector machine; and the neural network model includes a convolutional neural network. Model training and testing operations are repeatedly performed to obtain an optimized trained and tested intensity interpretation model. Other embodiments of the aforementioned implementation are as follows: the aforementioned image capture device captures these light spots at a first time to generate a first light spot image, and captures these light spots at a second time to generate a second light spot image. The first time and the second time are different, and the coupling qualities of these light spots correspond to these light intensities respectively. Thus, the system and method of the present invention for enhancing the light coupling quality of the light spot at the exit end of a silicon photonic wafer by utilizing artificial intelligence multiplexing image recognition technology collects image data of the light spot at the exit end of a silicon photonic wafer, and then, after artificial intelligence and deep learning, can be used to judge the light coupling intensity and quality, thereby shortening the inspection time of the silicon photonic wafer. The following describes several embodiments of the present invention with reference to the accompanying drawings. For clarity, many practical details are included in the following description. However, it should be understood that these practical details are not intended to limit the present invention. In other words, in some embodiments of the present invention, these practical details are not essential. Furthermore, to simplify the drawings, some commonly used structures and components are depicted in a simplified schematic manner; and repeated components may be represented using the same reference numerals. In addition, in this article, when a certain component (or unit or module, etc.) is "connected" to another component, it may mean that the component is directly connected to the other component, or it may mean that the component is indirectly connected to the other component, that is, there are other components between the component and the other component. When it is explicitly stated that a certain component is "directly connected" to another component, it means that there are no other components between the component and the other component. The terms "first", "second", "third", etc. are only used to describe different components and do not limit the components themselves. Therefore, the first component can also be renamed as the second component. Moreover, the combination of components / units / circuits in this article is not a generally known, conventional or conventional combination in this field. Whether the components / units / circuits themselves are conventional cannot be used to determine whether their combination relationship can be easily completed by ordinary knowledge in the technical field. This invention proposes an innovative system and method combining optical imaging and artificial intelligence algorithms for analyzing the light spot at the exit of a silicon photonics wafer. By analyzing the shape, size, symmetry, and other characteristics of the light spot, the waveguide's transmission mode, coupling efficiency, and other performance characteristics can be assessed, and its intensity can be determined. This system and method provides a new, fast, accurate, and non-contact method for measuring light intensity for wafer waveguide performance evaluation. Spot analysis provides rich information about waveguide transmission characteristics, supplementing simple power measurements. For example, the shape of the spot can reveal the waveguide's mode structure, such as single-mode or multimode. The spot's symmetry can indicate the waveguide's manufacturing quality, such as defects and stress. The spot's intensity distribution can reveal waveguide loss mechanisms, such as scattering and absorption. The following describes each embodiment in detail. Please refer to Figures 1 and 2 together, wherein Figure 1 is a schematic diagram of a system 100 for enhancing the light coupling quality of a light spot at an exit end of a silicon photonics wafer using artificial intelligence multiplexing image recognition technology according to a first embodiment of the present invention; and Figure 2 is a block diagram of the system 100 for enhancing the light coupling quality of a light spot at an exit end of a silicon photonics wafer using artificial intelligence multiplexing image recognition technology shown in Figure 1. The system 100 for enhancing the light coupling quality of a light spot at an exit end of a silicon photonics wafer using artificial intelligence multiplexing image recognition technology includes an image capture device 200 and a processing device 300. The image capture device 200 is used to capture a light spot at an exit end OE of a silicon photonics wafer 110 to generate a first light spot image 210 and a second light spot image 220. The processing device 300 is electrically connected to the image capture device 200 and obtains the first light spot image 210 and the second light spot image 220. The processing device 300 performs the following operations: an image preprocessing operation 310, a light spot segmentation and feature extraction operation 320, an intensity interpretation model establishment operation 330, a model training and testing operation 340, and an actual application operation 350. The image preprocessing operation 310 includes performing a preprocessing operation on the first light spot image 210 to generate a preprocessed light spot image. The light spot segmentation and feature extraction operation 320 includes using an image segmentation algorithm to extract a light spot region of the preprocessed light spot image and extracting a key feature from the light spot region to quantify an intensity distribution of a light spot. The intensity interpretation model establishment operation 330 includes establishing a mapping relationship between the key feature and a light intensity of the light spot to establish an intensity interpretation model corresponding to the first light spot image 210. The model training and testing operation 340 includes training the intensity interpretation model according to a training data set and testing the intensity interpretation model according to a test data set to obtain a trained test intensity interpretation model with artificial intelligence. The trained test intensity interpretation model is used to interpret the second light spot image 220. The actual application operation 350 includes using the trained test intensity interpretation model to perform real-time analysis and intensity interpretation on the new second light spot image 220. Specifically, the image capture device 200 may include one of a charge-coupled device (CCD) image sensor and a complementary metal oxide semiconductor (CMOS) image sensor. The image capture device 200 has a dynamic range, a linear response, a magnification, and a field of view to capture changes in light intensity, and covers an output end face 112 of the silicon photonics wafer 110. The output end face 112 corresponds to the output end OE (the output end OE is located at the output end face 112). The image capture device 200 captures a light spot at a first time to generate a first light spot image 210, and captures a light spot at a second time to generate a second light spot image 220. The first time and the second time are different (the first time is before the second time), and the light coupling quality of the light spot corresponds to the light intensity. To this end, the present invention utilizes a high-resolution image capture device 200 (imaging device) to capture light spot images (e.g., first light spot image 210 and second light spot image 220) at the output end (i.e., exit end OE) of the wafer waveguide. The image capture device 200 possesses sufficient dynamic range and linear response to accurately capture variations in light intensity. Furthermore, the image capture device 200 possesses an appropriate magnification and field of view to cover the entire output end face 112. In the image pre-processing operation 310, the pre-processing operation includes at least one of a background correction operation and a noise removal operation, thereby improving image quality and eliminating interference from external factors. In the spot segmentation and feature extraction operation 320, the image segmentation algorithm includes one of a threshold segmentation method and an edge detection method, and the key feature includes one of a brightness mean, a peak value, and a standard deviation. This allows for a quantitative description of the intensity distribution of the spot. In other embodiments, the key feature may be any two or all of the brightness mean, peak value, and standard deviation (i.e., the number of key features may be multiple), but the present invention is not limited to this. In the intensity interpretation model building operation 330 , the intensity interpretation model includes one of a physical-based analytical model and a machine-learning-based data-driven model. The machine-learning-based data-driven model includes one of a regression model and a neural network model, wherein the regression model includes one of a simplification swarm optimization (SSO), a linear regression, and a support vector machine, and the neural network model includes a convolutional neural network. In model training and testing operation 340, to train and test the intensity interpretation model, the processing device 300 prepares a labeled dataset containing spot images and their corresponding real-world light intensity values. The labeled dataset is then divided into a training dataset and a test dataset. The training dataset is used to learn model parameters, and the test dataset is used to evaluate model performance. Through repeated training and optimization, an accurate and robust intensity interpretation model can be obtained. Furthermore, model training and testing operation 340 can be repeated to obtain an optimized trained and tested intensity interpretation model. The number of repetitions can be determined based on user needs. In the actual application operation 350, the processing device 300 integrates the trained test intensity judgment model into the entire test system to perform real-time analysis and intensity judgment on the new second light spot image 220. The above-mentioned test system may include an optical probe system for multi-point high-dimensional synchronous alignment light coupling of silicon photons and a system 100 for enhancing the light coupling quality of the light spot at the exit end of the silicon photonic wafer using artificial intelligence multiplexing image recognition technology. The optical probe system for multi-point high-dimensional synchronous alignment light coupling of silicon photons corresponds to the entry end IE of the silicon photonic wafer 110, and the system 100 for enhancing the light coupling quality of the light spot at the exit end of the silicon photonic wafer using artificial intelligence multiplexing image recognition technology corresponds to the exit end OE. The optical probe-to-silicon photonic multi-point high-dimensional synchronous alignment light coupling system is used to perform alignment light coupling on the silicon photonic wafer 110. The optical probe-to-silicon photonic multi-point high-dimensional synchronous alignment light coupling system includes an optical probe array and a moving device. The optical probe array includes a plurality of optical probes, and these optical probes are set toward the silicon photonic wafer 110. The moving device is connected to the optical probe array and is used to move the optical probe array. These optical probes synchronously emit a plurality of incident lights to the plurality of entry ports IE (coupling points) of the silicon photonic wafer 110 at a time point to perform multi-point alignment light coupling. In this way, the optical probe-to-silicon photonic multi-point high-dimensional synchronous alignment light coupling system can use the optical probe array and the moving device to realize a highly adaptive and intelligent multi-point synchronous coupling light alignment solution, thereby being able to quickly and accurately find the optimal entry port IE and maintain stable coupling performance under environmental interference and system drift. Furthermore, the optical probe system for multi-point, high-dimensional, synchronous alignment of optical coupling for silicon photonics has multi-point parallel processing capabilities, alleviating the conflict between coupling efficiency and throughput, effectively improving the production efficiency of silicon photonic devices. The optical probe system for multi-point, high-dimensional, synchronous alignment of optical coupling for silicon photonics, combined with a system 100 that utilizes artificial intelligence multiplexing image recognition technology to enhance the coupling quality of the light spot at the exit of the silicon photonic wafer, enables multi-point synchronous alignment of optical coupling at the entry point (IE) and multiplexing image recognition at the exit point (OE). The processing device 300 may include a memory and a processor. The memory may include a random access memory (RAM) or other type of dynamic storage device that can store information and instructions for execution by the processor. The training dataset and test dataset of the aforementioned labeled dataset are stored in the memory. The processor may be a cloud processor, a digital signal processor (DSP), a microprocessor (MPU), a central processing unit (CPU), or other electronic processor, but the present invention is not limited to the above. Thus, the present invention's system 100 for enhancing the optical coupling quality of the light spot at the exit of a silicon photonics wafer using artificial intelligence multiplexed image recognition technology combines optical imaging and artificial intelligence algorithms to analyze the light spot at the exit of a silicon photonics wafer 110. By analyzing characteristics such as the shape, size, and symmetry of the light spot, the waveguide's transmission mode, coupling efficiency, and other properties can be evaluated, and its intensity can be determined. This system provides a new, fast, accurate, and non-contact method for measuring light intensity for wafer waveguide performance evaluation, addressing the lengthy testing time associated with conventional silicon photonics wafer testing methods. Please refer to Figures 1, 2, and 3. Figure 3 illustrates a flow chart of method S0 for enhancing the light coupling quality of a light spot at the exit of a silicon photonics wafer using artificial intelligence multiplexing image recognition technology, according to a second embodiment of the present invention. Method S0 for enhancing the light coupling quality of a light spot at the exit of a silicon photonics wafer using artificial intelligence multiplexing image recognition technology includes steps S01, S02, S03, S04, S05, and S06. Steps S01, S02, S03, S04, S05, and S06 are performed sequentially. Step S01 includes capturing a light spot at the exit end (OE) of the silicon photonics wafer 110 by the image capture device 200 to generate a first light spot image 210 and a second light spot image 220. Step S02 includes performing a pre-processing operation on the first light spot image 210 by the processing device 300 to generate a pre-processed light spot image. Step S03 includes extracting a light spot region from the pre-processed light spot image using an image segmentation algorithm by the processing device 300, and extracting key features from the light spot region to quantify the intensity distribution of the light spot. Step S04 includes establishing a mapping relationship between the key features and the light intensity of the light spot by the processing device 300 to establish an intensity interpretation model corresponding to the first light spot image 210. Step S05 includes executing a model training and testing operation 340 by the processing device 300, wherein the model training and testing operation 340 includes training an intensity judgment model based on a training data set and testing the intensity judgment model based on a test data set to obtain a trained test intensity judgment model with artificial intelligence, and the trained test intensity judgment model is used to judge the second light spot image 220. Step S06 includes using the trained test intensity judgment model by the processing device 300 to perform real-time analysis and intensity judgment on the new second light spot image 220. Thus, the method S0 of the present invention for enhancing the light coupling quality of the light spot at the exit end of the silicon photonic wafer using artificial intelligence multiplexed image recognition technology combines optical imaging and artificial intelligence algorithms to analyze the light spot at the exit end OE of the silicon photonic wafer 110. By analyzing the characteristics of the light spot, such as shape, size, and symmetry, the transmission mode, coupling efficiency, and other performance of the waveguide can be evaluated, and its intensity can be determined. This method provides a new, fast, accurate and non-contact way to measure light intensity for wafer waveguide performance evaluation, solving the problem of long detection time in conventional silicon photonic wafer testing methods. Please refer to Figures 1, 2, 3, and 4. Figure 4 is a flow chart illustrating a method S2 for enhancing the light coupling quality of a light spot at the exit of a silicon photonics wafer using artificial intelligence multiplexing image recognition technology, according to a third embodiment of the present invention. Method S2 for enhancing the light coupling quality of a light spot at the exit of a silicon photonics wafer using artificial intelligence multiplexing image recognition technology includes steps S21, S22, S23, S24, S25, and S26. Steps S21, S22, S23, S24, S25, and S26 are performed sequentially. Step S21 includes capturing a plurality of light spots at a plurality of exit ends OE of the silicon photonics wafer 110 by the image capture device 200 to generate a first light spot image 210 and a second light spot image 220. Step S22 includes performing a pre-processing operation on the first light spot image 210 by the processing device 300 to generate a pre-processed light spot image. Step S23 includes extracting a plurality of light spot regions from the pre-processed light spot image using an image segmentation algorithm by the processing device 300, and extracting a plurality of key features from these light spot regions to quantify the complex intensity distribution of these light spots. Step S24 includes establishing a complex mapping relationship between these key features and the complex light intensities of these light spots by the processing device 300 to establish an intensity interpretation model corresponding to the first light spot image 210. Step S25 includes executing a model training and testing operation 340 by the processing device 300, wherein the model training and testing operation 340 includes training an intensity judgment model based on a training data set and testing the intensity judgment model based on a test data set to obtain a trained test intensity judgment model with artificial intelligence, and the trained test intensity judgment model is used to judge the second light spot image 220. Step S26 includes using the trained test intensity judgment model by the processing device 300 to perform real-time analysis and intensity judgment on the new second light spot image 220. Thus, the method S2 of the present invention for enhancing the light coupling quality of the light spot at the exit end of the silicon photonic wafer using artificial intelligence multiplexed image recognition technology combines optical imaging and artificial intelligence algorithms to analyze the light spot at the exit end OE of the silicon photonic wafer 110. By analyzing the characteristics of the light spot, such as shape, size, and symmetry, the transmission mode, coupling efficiency, and other performance of the waveguide can be evaluated, and its intensity can be determined. This method provides a new, fast, accurate and non-contact way to measure light intensity for wafer waveguide performance evaluation, solving the problem of long detection time in conventional silicon photonic wafer testing methods. It is also worth mentioning that the present invention can position the laser (including intensity, type, etc.) based on parameters (beam spot characteristics) such as spectrum, beam structure, two-dimensional cross section (2D cross section), three-dimensional beam profile (3D beam profile), beam size, monochromatic (single wavelength), temporal (continuous wave to femtosecond pulses), coherence (consistent phase between all light elements), highly concentrated (focusable to extremely small spots), and beam irradiance profile (unique spatial power or energy distribution), but the present invention is not limited thereto. The methods S0 and S2 of the present invention for enhancing the light coupling quality of the light spot at the exit end of a silicon photonic wafer by using artificial intelligence multiplexing image recognition technology can be implemented through a computer program product. The order of the various implementation steps described in the above embodiments can be adjusted, combined or omitted according to actual needs. The above embodiments can be implemented using a computer program product, which may include a machine-readable medium (non-transitory computer-readable storage medium) storing a plurality of instructions, and these instructions can program a computer to perform the steps in the above embodiments. The machine-readable medium can be, but is not limited to, a floppy disk, an optical disk, an optical read-only disk, a magneto-optical disk, a read-only memory, a random access memory, an erasable programmable read-only memory (EPROM), an electronically erasable programmable read-only memory (EEPROM), an optical card or a magnetic card, a flash memory, or any machine-readable medium suitable for storing electronic instructions. Furthermore, embodiments of the present invention may also be downloaded as a computer program product, which may be transferred from a remote computer to a requesting computer via a data signal using a communication link (eg, a network connection). It can be seen from the above embodiments that the present invention has the following advantages: First, an innovative system and method combining optical imaging and artificial intelligence algorithms are used to analyze the light spot at the exit end of the silicon photonic wafer. By analyzing the shape, size, symmetry and other characteristics of the light spot, the transmission mode, coupling efficiency and other performance of the waveguide can be evaluated, and its intensity can be judged. Second, the system and method of the present invention provide a new way to quickly, accurately and non-contactly measure light intensity for the performance evaluation of wafer waveguides, so as to solve the problem of long detection time in conventional silicon photonic wafer testing methods. Third, the image data of the light spot at the exit end of the silicon photonic wafer is first collected (such as spectrum and infrared images, etc.), and after artificial intelligence and deep learning, it can be used to judge the intensity and quality of the coupled light. Although the present invention has been disclosed above in terms of embodiments, this is not intended to limit the present invention. Anyone skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the invention shall be determined by the scope of the appended patent applications. 100: System for enhancing the light coupling quality of the light spot at the exit end of a silicon photonic wafer using artificial intelligence multiplexing image recognition technology 110: Silicon photonic wafer 112: Output end 200: Image capture device 210: First light spot image 220: Second light spot image 300: Processing device 310: Image preprocessing operation 320: Light spot segmentation and feature extraction operation 330: Intensity judgment model establishment operation 340: Model training and testing operation 350: Practical application operation IE: Entry end OE: Exit end S0, S2: Method for enhancing the light coupling quality of the light spot at the exit end of a silicon photonic wafer using artificial intelligence multiplexing image recognition technology S01, S02, S03, S04, S05, S06, S21, S22, S23, S24, S25, S26: Steps Figure 1 is a schematic diagram of a system for enhancing the light coupling quality of the light spot at the exit end of a silicon photonic wafer by using artificial intelligence multiplexing image recognition technology according to the first embodiment of the present invention; Figure 2 is a block diagram of the system for enhancing the light coupling quality of the light spot at the exit end of a silicon photonic wafer by using artificial intelligence multiplexing image recognition technology according to Figure 1; Figure 3 is a flow chart of a method for enhancing the light coupling quality of the light spot at the exit end of a silicon photonic wafer by using artificial intelligence multiplexing image recognition technology according to the second embodiment of the present invention; and Figure 4 is a flow chart of a method for enhancing the light coupling quality of the light spot at the exit end of a silicon photonic wafer by using artificial intelligence multiplexing image recognition technology according to the third embodiment of the present invention. S0: Method for enhancing the light coupling quality of the light spot at the exit of silicon photonic wafer using artificial intelligence multi-tasking image recognition technology S01, S02, S03, S04, S05, S06: Steps

Claims

1. A system for enhancing the light coupling quality of a light spot at an exit end of a silicon photonics wafer using artificial intelligence multiplexing image recognition technology comprises: an image capture device for capturing a light spot at an exit end of a silicon photonics wafer to generate a first light spot image and a second light spot image; and a processing device electrically connected to the image capture device and acquiring the first light spot image and the second light spot image, the processing device being configured to perform the following operations: performing a preprocessing operation on the first light spot image to generate a preprocessed light spot image; extracting a light spot region from the preprocessed light spot image using an image segmentation algorithm, and extracting a key feature from the light spot region to quantify an intensity distribution of a light spot; A mapping relationship is established between the key feature and a light intensity of the light spot to establish an intensity interpretation model corresponding to the first light spot image; and a model training and testing operation is performed, wherein the model training and testing operation includes training the intensity interpretation model based on a training data set and testing the intensity interpretation model based on a test data set to obtain a trained test intensity interpretation model with artificial intelligence, and the trained test intensity interpretation model is used to interpret the second light spot image.

2. A system for enhancing the light coupling quality of a light spot at an exit end of a silicon photonics wafer using artificial intelligence multiplexed image recognition technology as described in claim 1, wherein the image capture device comprises one of a charge-coupled device (CCD) image sensor and a complementary metal oxide semiconductor (CMOS) image sensor, the image capture device having a dynamic range, a linear response, a magnification, and a field of view to capture changes in the light intensity and covering an output end face of the silicon photonics wafer, the output end face facing the exit end.

3. A system for enhancing the light coupling quality of the light spot at the exit end of a silicon photonic wafer using artificial intelligence multiplexed image recognition technology as described in claim 1, wherein the preprocessing operation includes at least one of a background correction operation and a denoising operation, the image segmentation algorithm includes one of a threshold segmentation method and an edge detection method, and the key feature includes one of a brightness mean, a peak value, and a standard deviation.

4. A system for enhancing the light coupling quality of a light spot at the exit of a silicon photonic wafer using artificial intelligence multiplexed image recognition technology as described in claim 1, wherein: The intensity judgment model includes one of an analytical model based on physical principles and a data-driven model based on machine learning, the data-driven model based on machine learning includes one of a regression model and a neural network model, the regression model includes one of a simplification swarm optimization (SSO), a linear regression and a support vector machine, and the neural network model includes a convolutional neural network; and the model training and testing operations are repeatedly performed to obtain the optimized trained and tested intensity judgment model.

5. A system for enhancing the light coupling quality of a light spot at the exit end of a silicon photonic wafer using artificial intelligence multiplexed image recognition technology as described in claim 1, wherein the image capture device captures the light spot at a first time to generate the first light spot image, and captures the light spot at a second time to generate the second light spot image, the first time being different from the second time, and the light coupling quality of the light spot corresponds to the light intensity.

6. A method for enhancing the light coupling quality of a light spot at an exit end of a silicon photonics wafer using artificial intelligence multiplexing image recognition technology comprises: capturing a light spot at an exit end of a silicon photonics wafer using an image capture device to generate a first light spot image and a second light spot image; performing a pre-processing operation on the first light spot image using a processing device to generate a pre-processed light spot image; extracting a light spot region from the pre-processed light spot image using an image segmentation algorithm by the processing device, and extracting a key feature from the light spot region to quantify an intensity distribution of the light spot; The processing device establishes a mapping relationship between the key feature and a light intensity of the light spot to establish an intensity interpretation model corresponding to the first light spot image; and the processing device performs a model training and testing operation, wherein the model training and testing operation includes training the intensity interpretation model based on a training data set and testing the intensity interpretation model based on a test data set to obtain a trained test intensity interpretation model with artificial intelligence, and the trained test intensity interpretation model is used to interpret the second light spot image.

7. A method for enhancing the light coupling quality of a light spot at an exit end of a silicon photonics wafer using artificial intelligence multiplexed image recognition technology as described in claim 6, wherein the image capture device comprises a charge-coupled device (CCD) image sensor and a complementary metal oxide semiconductor (CMOS) image sensor, the image capture device having a dynamic range, a linear response, a magnification, and a field of view to capture changes in the light intensity and covering an output end face of the silicon photonics wafer, the output end face facing the exit end.

8. A method for enhancing the light coupling quality of the light spot at the exit end of a silicon photonic wafer using artificial intelligence multiplexed image recognition technology as described in claim 6, wherein the preprocessing operation includes at least one of a background correction operation and a denoising operation, the image segmentation algorithm includes one of a threshold segmentation method and an edge detection method, and the key feature includes one of a brightness mean, a peak value, and a standard deviation.

9. The method of enhancing the light coupling quality of the light spot at the exit end of a silicon photonic wafer by using artificial intelligence multiplexing image recognition technology as described in claim 6, wherein: The intensity judgment model includes one of an analytical model based on physical principles and a data-driven model based on machine learning, the data-driven model based on machine learning includes one of a regression model and a neural network model, the regression model includes one of a simplification swarm optimization (SSO), a linear regression and a support vector machine, and the neural network model includes a convolutional neural network; and the model training and testing operations are repeatedly performed to obtain the optimized trained and tested intensity judgment model.

10. A method for enhancing the light coupling quality of a light spot at the exit end of a silicon photonic wafer using artificial intelligence multiplexed image recognition technology as described in claim 6, wherein the image capture device captures the light spot at a first time to generate the first light spot image, and captures the light spot at a second time to generate the second light spot image, the first time being different from the second time, and the light coupling quality of the light spot corresponds to the light intensity.

11. A method for enhancing the light coupling quality of a light spot at a silicon photonics wafer outlet using artificial intelligence multiplexed image recognition technology comprises: capturing a plurality of light spots at a plurality of outlets of a silicon photonics wafer using an image capture device to generate a first light spot image and a second light spot image; performing a preprocessing operation on the first light spot image using a processing device to generate a preprocessed light spot image; extracting a plurality of light spot regions from the preprocessed light spot image using an image segmentation algorithm by the processing device, and extracting a plurality of key features from the light spot regions to quantify the plurality of intensity distributions of the light spots; The processing device establishes a complex mapping relationship between the key features and the complex light intensities of the light spots to establish an intensity interpretation model corresponding to the first light spot image; and the processing device performs a model training and testing operation, wherein the model training and testing operation includes training the intensity interpretation model based on a training data set and testing the intensity interpretation model based on a test data set to obtain a trained test intensity interpretation model with artificial intelligence, and the trained test intensity interpretation model is used to interpret the second light spot image.

12. A method for enhancing the light coupling quality of a light spot at an exit end of a silicon photonics wafer using artificial intelligence multiplexed image recognition technology as described in claim 11, wherein the image capture device comprises a charge-coupled device (CCD) image sensor and a complementary metal oxide semiconductor (CMOS) image sensor, the image capture device having a dynamic range, a linear response, a magnification, and a field of view to capture changes in the light intensities and covering an output end face of the silicon photonics wafer, the output end face facing the exit end.

13. A method for enhancing the light coupling quality of the light spot at the exit end of a silicon photonic wafer using artificial intelligence multiplexed image recognition technology as described in claim 11, wherein the preprocessing operation includes at least one of a background correction operation and a denoising operation, the image segmentation algorithm includes one of a threshold segmentation method and an edge detection method, and each of the key features includes one of a brightness mean, a peak value, and a standard deviation.

14. The method of enhancing the light coupling quality of the light spot at the exit end of a silicon photonic wafer by using artificial intelligence multiplexing image recognition technology as described in claim 11, wherein: The intensity judgment model includes one of an analytical model based on physical principles and a data-driven model based on machine learning, the data-driven model based on machine learning includes one of a regression model and a neural network model, the regression model includes one of a simplification swarm optimization (SSO), a linear regression and a support vector machine, and the neural network model includes a convolutional neural network; and the model training and testing operations are repeatedly performed to obtain the optimized trained and tested intensity judgment model.

15. A method for enhancing the light coupling quality of light spots at the exit end of a silicon photonic wafer using artificial intelligence multiplexing image recognition technology as described in claim 11, wherein the image capture device captures the light spots at a first time to generate the first light spot image, and captures the light spots at a second time to generate the second light spot image, the first time being different from the second time, and the light coupling qualities of the light spots respectively correspond to the light intensities.

Images