155 results about "Spatial modulation" patented technology

The invention provides a spatial light modulator, a self-checking device thereof and a light computing device, and relates to the technical field of optical modulation.The spatial light modulator comprises an incident waveguide structure, a covering plate, a liquid crystal layer, a substrate and an emergent waveguide structure; the incident waveguide structure is optically coupled with the cover plate and is used for guiding externally incident laser into the cover plate at a preset angle; the covering plate is used for enabling laser to be propagated in the covering plate in a total reflection form; the substrate bears the liquid crystal layer, integrates the driving circuit and is used for applying regulation and control voltage to the liquid crystal layer; the liquid crystal layer is arranged on the lower side of the covering plate and used for conducting phase modulation on the laser transmitted to the liquid crystal layer in a total reflection mode in response to the regulation voltage. And the emergent waveguide structure is optically coupled with the covering plate and is used for guiding the laser modulated by the liquid crystal layer out of the covering plate to a free space. According to the spatial light modulator, efficient spatial modulation and accurate control of the light field are realized, so that the integration and function accuracy of the spatial light modulator in self-inspection and light calculation are improved.

The invention discloses a visual detection method and system for tea impurities, electronic equipment and a storage medium, and the method comprises the steps: obtaining a to-be-detected tea image and a training image data set, and carrying out the preprocessing of image data in the training image data set; inputting a to-be-detected tea image into a trained tea impurity detection model to obtain a plurality of tea impurity detection results containing confidence coefficients of a to-be-detected target and a detection frame, the trained tea impurity detection model comprising a backbone network, a neck network and a plurality of detection heads, the backbone network comprises a ScharrStem module, a first SGCB improved module and a second SGCB improved module, and the neck network comprises a first SGCB improved module, a second SGCB improved module and a channel sensing spatial modulation module; and determining a target tea leaf impurity detection result according to the confidence coefficient and the detection frame in the plurality of tea leaf impurity detection results. The detection accuracy of the tea impurities can be improved.

The invention discloses a multi-scale azimuth perception feature enhancement network SAR target identification method and system, and belongs to the technical field of image processing. The method comprises the following steps: obtaining a feature map of an SAR image; generating a channel parameter and a dynamic convolution kernel according to a target azimuth angle of the SAR image, and performing channel and spatial modulation on the feature map by adopting the channel parameter and the dynamic convolution kernel to obtain an enhanced feature map; fusing the enhanced feature maps of different scales to obtain a final feature map; in the training process, a difference azimuth angle feature consistency loss function, a dynamic angle interval loss function and a cross entropy loss function are adopted to calculate an overall loss function of the network, and iterative training is carried out to obtain a trained multi-scale azimuth angle perception feature enhancement network; identifying a target in the SAR image according to the trained multi-scale azimuth perception feature enhancement network; the method can adapt to feature changes caused by changes of azimuth angles of different targets in radar image recognition, and the robustness of SAR target recognition in a complex environment is improved.

The invention discloses a synchronous perception and backscatter communication method and device based on spatial modulation and backscatter, and relates to the technical field of communication, and the method comprises the steps: transmitting a pilot signal to a label through a transmitting array, modulating the load impedance of the label, and then backscattering the signal to a receiving array; the receiving voltage generated by the receiving array and the original channel matrix are adopted to determine the reflectivity of the clutter tag and the reflectivity of the tag antenna; positioning and identifying the tag based on the reflectivity of the clutter tag; extracting a target channel matrix for constructing a tag from the original channel matrix according to the tag antenna reflectivity; performing spatial modulation on the information bits through a transmitting array, generating constellation symbols and determining an activation label; and backscattering the constellation symbols to the receiving array through the activation tag by adopting the transmitting array, and demodulating and decoding by combining the target channel matrix to output information bits. The backscatter label is accurately positioned and identified in a disordered environment, and spatial modulation is integrated into backscatter communication, so that the reliability of backscatter communication is improved.

The invention discloses a springback compensation control system for a metal stamping part, and belongs to the technical field of metal plate forming. The system comprises a stress memory prediction module, a fractal compensation analysis module, a phase change compensation regulation and control module, a compensation decision fusion module and an execution feedback optimization module. The stress memory prediction module extracts stress evolution characteristics of the stamping process through a multi-scale memory network and a time decay attention mechanism; the fractal compensation analysis module adaptively adjusts the compensation grid density based on the fractal dimension to realize optimal configuration of computing resources; the phase change compensation regulation and control module predicts the phase change behavior of the material through the thermal-mechanical coupling constitutive model and calculates the compensation correction amount caused by phase change; the compensation decision fusion module intelligently fuses three compensation strategies by adopting Bayesian reasoning and a spatial modulation function; and the execution feedback optimization module realizes continuous optimization of the system through online learning. According to the method, the influence of three dimensions of stress history, geometric complexity and material phase change is comprehensively considered, accurate prediction and compensation of the springback behavior are achieved, and the forming precision and production efficiency of the metal stamping part are remarkably improved.

A spatial adaptive multi-user service modulation and demodulation method suitable for various channel environments is characterized in that an uplink channel is configured with multiple paths of general carriers and multiple paths of helicopter special carriers in a frequency division mode, and the general carriers support uplink adaptive information transmission of terminals of various types and sizes in an MF-TDMA (Multimedia Frequency-Time Division Multiple Access) system; the on-satellite demodulator obtains a modulation coding mode of a burst data segment by identifying mode word information in a physical layer frame header, and demodulates a corresponding mode; the carrier special for the helicopter uses a continuous carrier mode, and multi-rate channel error correction coding and decoding under rotor wing shielding are supported in a channel interleaving mode. Downlink transmission is carried out in a fixed time slot and free time slot time division mode, and downlink information transmission for universal multi-type terminals and helicopter terminals can be supported at the same time. Compared with a common spatial modulation and demodulation method, the modulation and demodulation scheme provided by the invention can support adaptive transmission of general services of multiple types of terminals and helicopter service information on both an uplink and a downlink by using one satellite-borne modem.

The present invention discloses a low-complex, reliable multidimensional orthogonal time-frequency space (ND-OTFS) modulation-based communication system for uplink and downlink wireless communication between resource-constrained Internet of Things (IoT) and machine-to-machine (M2M) devices comprising at least one transmitter configured to transmit bit stream involving N-Dimensional signal mapper and OTFS frame and at least one receiver configured to be integrated into said IoT or M2M devices, featuring a low-complexity detector for receiving the transmitted bit stream.

The invention provides a coding aperture snapshot spectral imaging system based on orthogonal double chromatic dispersion and a hyperspectral image reconstruction method, and the system obtains horizontal and vertical compression measurement images through the synchronous compression sampling of horizontal and vertical chromatic dispersion channels which are orthogonally arranged and a common spatial modulation mask. The reconstruction method comprises the following steps: in combination with an imaging mathematical model, establishing an optimization equation containing a priori item, constructing a double-color dispersion depth expansion framework (DDDU), and decomposing the optimization equation into a data sub-problem and a priori sub-problem. Data sub-problems are independently optimized through horizontal and vertical dispersion closed solutions, and a dual-channel complementary information structure is kept; according to the prior sub-problem, double-color scatter information fusion Transform (DDIFT) is adopted to carry out joint denoising and feature fusion. And high-fidelity hyperspectral image reconstruction is realized through iterative solution. According to the method, the sampling rate is improved by innovatively combining the orthogonal double-color dispersion channel design, and the imaging and reconstruction performance is improved by effectively utilizing complementary spectral information through an algorithm combining a physical model and deep learning.

A flying-over beam pattern scanning hologram microscope device includes: a scan beam generation unit which converts of a first beam and a second beam to a first spherical wave; a scanning unit, which includes a spatial modulation scanner for controlling the scan beam in the horizontal direction, and a translation stage for moving an object in the vertical direction at the rear end of the projection unit; the projection unit projecting the scan beam onto an object plane; and a light collection unit which detects a beam that has passed through the objective lens again after being reflected or fluoresced from the object.

The invention provides a sonar target detection method and device for a complex marine environment. According to the method provided by the invention, a C3k2SFD module maps a convolution kernel weight to a frequency domain through frequency weight decomposition, sets are divided according to a frequency mode length and then recombined into a convolution weight, a global attention branch and a first local branch are constructed in combination with spatial modulation to generate a dynamic convolution weight, and sub-band features are adjusted through frequency band modulation and a spatial attention graph; the MSDA module performs normalization and channel expansion on the input features, then splits the features, extracts spatial attention features through deep convolution of different directions and sizes, fuses various receptive field features, and multiplies the features with another feature; the C2DYT module carries out feature nonlinear compression control on extreme value echo features through the first dynamic activation layer, carries out dynamic range compression and channel-level affine adjustment on the second dynamic activation layer, combines residual connection, and effectively improves the accuracy and robustness of target detection in a complex marine environment through the combined action of the three.

The invention discloses an optical fiber single-pixel imaging method and device based on sparse optimization and neural network fusion, and the method comprises the steps: constructing a collimation illumination system, carrying out the spatial modulation of incident light, introducing a collimation illumination mechanism, and obtaining a spatial speckle illumination light field of a multimode optical fiber; performing sparsity extraction analysis on the spatial speckle illumination light field by using a sparse optimization algorithm, removing a high redundancy mode, and selecting a speckle illumination mode with the lowest cross correlation; a speckle illumination mode after sparse optimization is used as input, a neural network model constructed based on a soft threshold iterative algorithm is trained, and an optical fiber single-pixel image reconstruction model is established; performing sparse optimization speckle illumination on a target object, acquiring corresponding total light intensity response data, inputting the data into the trained optical fiber single-pixel image reconstruction model, realizing image reconstruction and outputting a final target image; the method has anti-defocusing capability and anti-disturbance robustness, and can still keep good optical fiber single-pixel image reconstruction performance under the condition of low sampling rate.

The invention relates to the technical field of photoelectric detection, in particular to a photocurrent imaging system based on a probe station and an electrometer. The system generates a focused detection light beam through a light path system, the focused detection light beam irradiates a sample carried by a probe station after being spatially modulated by a mask plate, weak light current signals generated by the sample are collected by an electrometer, and a high-resolution light current distribution image is generated by combining scanning motion of a high-precision electric control displacement platform and intelligent processing of computer software. A physically-driven signal enhancement model, a multi-dimensional noise suppression algorithm and a dynamic imaging optimization formula are innovatively designed, the problems that in the prior art, weak light current detection precision is low, the scale connection error is large, and imaging efficiency is insufficient are solved, accurate quantitative analysis of the photoelectric characteristics of materials is achieved, and the method is suitable for popularization and application. The method is suitable for semiconductor device research and development, solar cell performance evaluation, biosensor detection and other scenes, and has both detection precision and application flexibility.

A system and method for contact-based spatial frequency domain imaging. A medium to be imaged is contacted by a sheet comprising an array of lighting elements generating illumination light patterns illuminating the medium, and comprising an array of sensor elements measuring spatially resolved images of resulting light patterns reflected back after interaction of the illumination light patterns with the medium. The lighting elements generate illumination light patterns that are spatially modulated in light intensity along a surface of the illuminated medium according to a respective different spatial frequency. A set of spatially modulated images is determined based on the measured resulting light patterns caused by the different illumination light patterns. An image of the medium can be generated based on a combination of the spatially modulated images.

The application discloses a long-wave infrared dynamic space modulation device and method based on double coding apertures, and belongs to the technical field of optics. The application avoids the diffraction effect in the long-wave infrared space modulation process by selecting a large-size coding pixel coding aperture, adopts a double coding aperture space light modulation structure, separately performs space modulation on target light beams, and realizes coding effect superposition through beam combination, so that the space modulation resolution of the large-size coding pixel coding aperture is improved. On the premise of guaranteeing the space modulation resolution requirement and without increasing the coding aperture area, the diffraction effect in the long-wave infrared space modulation process is avoided by selecting the large-size coding pixel coding aperture. The application introduces the use of a displacement table, and the position of the moving coding template is transformed for multiple times within the single information acquisition time of the system, so that the effective coding area in the optical path is transformed, more coding superposition effects are generated, and the coding resolution and the coding gray scale are further improved.

The invention discloses a boundary perception gating cross-scale feature fusion method for a remote sensing image, belongs to the technical field of computer vision and remote sensing image processing, and particularly relates to the field of remote sensing target detection. Aiming at the problems of insufficient cross-layer alignment, boundary confusion, insufficient small target recall and the like in the prior art, the invention provides a boundary perception gating cross-scale feature fusion method for a remote sensing image. The method comprises the following steps: constructing an Epag module containing Scharr edge prior and logit space fusion, and realizing dual-branch feature adaptive fusion; designing an SAFM module with channel separation and multi-scale pooling, and completing spatial adaptive enhancement; the two modules are integrated at the neck of a detection network, and multi-scale features are output through top-down gating fusion and spatial modulation. According to the method, the target boundary positioning precision and the multi-scale adaptation capability are remarkably improved, multiple technical pain points are solved, mAPs on DOTA1.0 and DIOR-R data sets are improved respectively, multiple detectors are adapted, and accurate technical support is provided for remote sensing application such as urban planning and traffic monitoring.

The invention discloses a space-time modulation hydride vapor phase epitaxy reaction chamber, equipment and a method. Wherein the reaction chamber comprises a reaction chamber which is provided with a water cooling side wall and is limited by a flange to form a cavity, and the interior of the cavity is divided into an upper part, a middle part and a lower part; the metal source reaction boat assembly is positioned at the upper half part of the reaction chamber cavity; the space-time modulation reaction growth device is positioned in the middle of the reaction chamber cavity; the heating device and the tail gas collecting device are positioned at the lower half part of the reaction chamber cavity; and the central gas circuit device is positioned at the axis position of the cavity of the reaction chamber and is connected with the metal source reaction boat assembly at the upper half part of the cavity and the space-time modulation reaction growth device in the middle of the cavity. According to the invention, the growth time is controlled through the combination of time modulation and space modulation, so that the thickness of the material thin layer is accurately controlled, the flexibility of growth program setting is improved, the rapid epitaxial growth of the ultra-thin layer and ultra-thick layer material stacking structure is realized, and the low-cost epitaxial growth requirement of a semiconductor device with a complex structure is met.

A semiconductor device comprising: a silicon substrate; at least one P+ / N-well structure formed on the substrate, the at least one P+ / N-well structure comprising a first set of N-well layers coupled to an extra positive extrinsic electric field, and a second set of N-well layers coupled to an extra negative extrinsic electric field; and whereby the extra positive extrinsic electric field and the extra negative extrinsic electric field accelerate movement of carriers in the layers.

A system and method of generating a perceptual reflex encryption key (PRE-Key) may include the following steps: 1) delivering a spatially modulated auditory stimulus from a mobile device to a human subject via a secure audio output interface; 2) capturing, using a MEMS sensor subsystem, an involuntary physical response of the human subject to the auditory stimulus; 3) determining a response latency Δt between stimulus delivery and the captured response; 4) extracting a perceptual feature vector based on neocortical response approximations; and 5) computing the PRE-Key by hashing a combination of the auditory stimulus parameters, the perceptual feature vector, the physical response, and the response latency.

The invention discloses an instantaneous polymorphic target high-confidence visual detection and identification method, and belongs to the field of instantaneous target intelligent identification. The method comprises the following steps: firstly, continuously collecting infrared images of an instantaneous target area; then carrying out change detection on the current frame image and the initial frame image, and judging a change area; screening to obtain an instantaneous target screening sequence; inputting the screening sequence into an instantaneous target recognition network for training; and finally, inputting a regional infrared image in an actual scene into the trained instantaneous target recognition network to obtain a target detection and recognition result. For an instantaneous target, a continuous multi-frame change region screening method is adopted, an instantaneous target recognition problem is converted into a continuous frame image behavior recognition problem, and an instantaneous target recognition network comprising a basic backbone network, a spatial modulation module, a time sequence modulation module, a spatial-temporal feature fusion module and a prediction module is constructed. Spatial features, time features and feature fusion can be processed in a targeted mode, and generalization is high.

The embodiment of the invention provides an optical system and optical processing equipment. The optical system comprises a light source module, a polarization modulation module, a micromirror array and an objective lens assembly. The light source module is used for outputting femtosecond laser pulses. The polarization modulation module is used for receiving the femtosecond laser pulse and modulating the polarization state of the femtosecond laser pulse into a circular polarization state. And the micro-mirror array is used for receiving the femtosecond laser pulse in the circular polarization state and performing spatial modulation on the femtosecond laser pulse to generate different patterns. The objective lens assembly is used for receiving and focusing the pattern to the surface of the sample. Thus, the femtosecond laser pulse passing through the polarization modulation module has the same polarization state in all directions, so that the absorptivity and the working efficiency of different positions of the sample M are consistent in the laser processing process, and the processing quality and effect are improved. Moreover, the micro-mirror array is sensitive to the polarization state, so that energy loss and non-uniformity caused by polarization can be avoided, and the processing uniformity and the processing precision can be further improved.

The invention discloses a crop leaf disease detection method and system based on an improved YOLOv11 model, and the method comprises the steps: obtaining image data of crop leaves, carrying out the preprocessing of the image data, and inputting the preprocessed image data into an improved and trained YOLOv11 model; performing feature extraction on the image data to obtain an input feature map, and performing multi-scale processing through an adaptive down-sampling module to obtain a multi-scale primary feature map; performing space and channel attention weighting on the multi-scale primary feature map through a multi-scale local attention module; and / or in at least another fusion layer, performing spatial modulation and channel attention fusion on the multi-scale primary feature map through a space-channel fusion attention module to obtain a multi-scale fusion feature map; and outputting a detection result according to a detection head in combination with the multi-scale fusion feature map. According to the invention, the precision and robustness of disease detection can be improved, and the calculation overhead is reduced.

The invention relates to the technical field of photoelectronic imaging, in particular to a snapshot spectrum camera shooting device and method.The device comprises an imaging modulation device, an acquisition device and a computing device.The imaging modulation device comprises an imaging lens, an optical filter array and a spatial modulation device; the optical modulator is configured to generate an optical signal with modulation at an image plane position of an imaging lens; the acquisition device comprises a two-dimensional image detector configured to convert an optical signal into image data and / or video data to generate a measurement image and / or a measurement video; and the computing equipment is connected with the acquisition equipment, and is configured to read and store the measurement image and / or the measurement video from the acquisition equipment, and invert the spectral image and / or the spectral video of the target scene from the measurement image and / or the measurement video. Therefore, the problems of difficulty in improving the spatial resolution, high cost and the like caused by exponential increase of on-chip encoder device cost and process complexity when pursuing higher spatial resolution in related technologies are solved.

The disclosure provides an electro-optic phase modulator for a displacement measurement system and the displacement measurement system. The electro-optic phase modulator comprises a housing with an accommodation space, an entrance and an exit are formed on the housing, and polarized measurement light enters the accommodation space through the entrance; a modulation crystal is arranged in the accommodation space and is configured to modulate the polarized measurement light from the entrance, and a polarization direction of the polarized measurement light is 45° to a modulation electric field direction of the modulation crystal; and a quarter-wave plate is located between the modulation crystal and the exit, an optical axis of the quarter-wave plate is 45° to the modulation electric field direction of the modulation crystal, the polarized measurement light passes through the quarter-wave plate and is emitted to a measured object through the exit, the measured object reflects the polarized measurement light back to the modulation crystal through the quarter-wave plate, and a polarization component direction of the polarized measurement light reflected back to the modulation crystal is changed by 90° compared with that of the polarized measurement light emitted to the modulation crystal through the entrance.

The invention discloses a high-dynamic three-dimensional reconstruction method based on cooperative regulation and control of a projector and a camera. At a camera end and a spatial light modulator, according to a pre-calibrated spatial mapping relation between the camera and the spatial light modulator, the corresponding position of an overexposure pixel on the spatial light modulator is accurately positioned, local gray scale attenuation is carried out on the overexposure pixel, and spatial domain intensity suppression of reflected light is realized; and when the adjustment of the spatial light modulator is insufficient to completely eliminate the overexposure area, the system starts the brightness modulation of the projector end, analyzes the image brightness distribution in real time through the camera, and dynamically adjusts the projection brightness of the local area by adopting a self-adaptive projection strategy to avoid overexposure of the high-reflectivity area. The process forms a closed-loop linkage control strategy of projection self-adaption and spatial modulation compensation, and the fringe imaging quality and the three-dimensional reconstruction precision of the system in a high dynamic range scene are remarkably enhanced.

This invention relates to the fields of physiological signal processing and emotion recognition technology, and discloses a method for recognizing physiological and emotional states of diverse populations. The method includes: acquiring facial video streams, infrared thermal imaging, and electrodermal signals; performing spectral adaptive enhancement based on skin color to obtain a spectrally enhanced video stream; inputting the spectrally enhanced video stream into a spatially modulated attention network to obtain a remote photoplethysmography waveform; extracting temperature distribution features from infrared thermal imaging; extracting electrodermal response features from electrodermal signals; aligning the three signals spatiotemporally and constructing alternative features for the failed signal based on the signal-to-noise ratio; inputting the three signals and alternative features into a multimodal signal pattern recognition network for weighted fusion, outputting a fused physiological feature; inputting the fused physiological feature into a population-level verification network for population adaptive matching and individual baseline correction, outputting the final emotional state. This invention achieves robust, fair, and accurate emotion recognition for diverse populations.

The invention discloses a communication method for realizing orthogonal time chirp space modulation based on discrete Fresnel transform, which is suitable for a mobile scene, and is characterized in that a data symbol is modulated on a two-dimensional complex chirp carrier on a delay-Fresnel Doppler plane, and the system performance is improved by using the interference resistance of the chirp wave. The system for realizing orthogonal time chirp spatial modulation based on discrete Fresnel transform provided by the invention has better bit error rate performance under a time-varying channel, especially under a high signal-to-noise ratio condition, is especially suitable for high-speed mobile application, and only increases a small amount of phase rotation complexity compared with OTFS.

A method of decoding information encoded in respective activation modes of an activated subset of antennas of a MIMO transmitter transmitting orthogonally modulated signal components, the method comprising: decoupling a real part and an imaginary part of a signal component of a received signal vector, respective channel matrices of real and virtual channels are received, determined, or estimated, and real and virtual signal components are represented as respective products of transmitted symbol components and activation unit vectors. The respective activation unit vectors of the decoupled real and imaginary parts are then represented with corresponding random variable vectors that are dependent on a probability quality function of the activation pattern available at the transmitter. Then, a Gaussian belief propagation-based messaging process is applied to the random variables to jointly estimate activation unit vectors of the respective real and imaginary parts of the received signal vector, and finally the activation unit vectors are decoded to obtain the information.

The invention provides a bee colony unmanned aerial vehicle communication system based on an enhanced OTFS frame and a hybrid topology, and relates to the technical field of unmanned aerial vehicle cluster wireless communication, and the system comprises autonomous wireless communication modules which are deployed in all unmanned aerial vehicles and are used for generating and processing orthogonal time-frequency space OTFS modulation signals; a star-mesh-linear IP hybrid topology network dynamically constructed by unmanned aerial vehicle nodes supports autonomous optimization of a communication path according to channel quality; embedding an enhanced OTFS frame structure of reduced-order cyclic prefix (RCP) and reduced-order zero filling (RZP) pilot frequency; a ground control station GSC integrating an OTFS module and a remote controller RC establish a low-delay link through UDP multicast; the invention relates to a RadCom joint module with integrated radar sensing and communication functions. According to the invention, a single-point fault risk is eliminated through decentration topology, inter-symbol and inter-carrier interference caused by multipath interference is effectively suppressed by an enhanced OTFS frame, the anti-interference capability of a complex environment is improved by a dual-band frequency hopping spread spectrum technology, and target detection and data transmission millisecond-level synchronization are realized by a RadCom module.

According to the polycrystalline silicon solar cell crack detection method based on the improved LDC edge detection model, the target data set is obtained by performing enhancement processing on the initial data set, a high-quality data basis is provided for model training, overfitting can be effectively avoided, and the detection accuracy is improved. The model can still show good performance when facing unseen data, double-dynamic snakelike convolution is designed and integrated in a backbone network in an LDC edge detection model, the perception ability of tiny edges and defects on the surface of the polycrystalline silicon solar cell is improved, a multi-scale attention mixer module is constructed in a branch network, and the multi-scale attention mixer module is used for detecting the surface of the polycrystalline silicon solar cell. Structural information under different receptive fields is obtained through multi-scale feature decoupling and dynamic space modulation, efficient interaction of cross-scale semantics is achieved, a dual loss function mechanism is added into a target network structure, and finally the accuracy of a crack detection result of the polycrystalline silicon solar cell by the obtained target improved edge detection model is guaranteed.

This application discloses a visual detection method, system, electronic device, and storage medium for tea impurities. The method acquires an image of the tea to be detected and a training image dataset, preprocesses the image data in the training image dataset, and inputs the image of the tea to be detected into a trained tea impurity detection model. This yields multiple tea impurity detection results containing confidence scores and detection boxes for the target tea. The trained tea impurity detection model includes a backbone network, a neck network, and multiple detection heads. The backbone network includes a ScharrStem module, a first improved SGCB module, and a second improved SGCB module. The neck network includes a first improved SGCB module, a second improved SGCB module, and a channel-aware spatial modulation module. Based on the confidence scores and detection boxes in the multiple tea impurity detection results, the detection result for the target tea impurity is determined. This application can improve the detection accuracy of tea impurities.