42 results about "Centroid algorithm" patented technology

The invention provides a light stripe center extraction method and system and a medium. The method comprises the following steps: judging the pixel state of each pixel point column by column according to gray values of pixel points in an image containing light strips, and determining an inflection point of a current column and a first end point of a peak interval where the inflection point is located; based on the gray value of the first end point of the peak interval, calculating the absolute value of the difference between the gray value of each pixel point of the inflection point of the current column in the rear row and the gray value of the first end point, and judging a second end point of the peak interval; and constructing a curve fitting model based on the gray values and the row coordinates of the pixel points in the peak interval, obtaining stagnation points, which are greater than zero, of a fitting curve, and selecting the stagnation point closest to the inflection point as a light stripe center pixel point in the peak interval. The dynamic peak interval and cubic curve fitting are combined for use, the accuracy of the finally extracted light stripe center is higher than the calculation results of a gray centroid algorithm, Gaussian fitting and Rayleigh fitting, and the method has good accuracy and robustness.

The invention discloses an inspection method of an industrial inspection system based on multi-modal fusion, which belongs to the technical field of industrial inspection, and comprises the following steps: S1, obtaining positioning data of a node to be positioned by using an HDS-TWR distance measurement algorithm and a Taylor full centroid algorithm, and outputting a positioning coordinate; s2, using a sensor to measure the acceleration and angular velocity of the inspection system so as to align the motion and posture for evaluation; s3, improving a Kalman filtering algorithm to enable an output prediction result to be close to a true value; s4, infrared-vision dual-mode cooperative tracking is adopted, and meanwhile, a time synchronization displacement compensation mechanism is utilized to calculate a relative displacement difference between two tracking modes, and dynamic calibration is carried out; s5, a YOLOv8 model is adopted to process video frames or pictures collected in the inspection process, a target box and category probability are output, and finally an inspection result is obtained; according to the inspection method of the industrial inspection system based on multi-modal fusion, a high-precision and low-delay automatic inspection solution is provided for severe environments such as mines and nuclear power stations.

The invention discloses a treadmill user center track identification method based on Kalman filtering, and belongs to the technical field of ultrasonic measurement. An ultrasonic sensor array installed above a running belt is used for collecting user position data, an improved weighted centroid algorithm is adopted for calculating a position observation value, and the algorithm is combined with an exponential attenuation item and a power weight to improve the positioning precision. A second-order motion state model containing position and speed is established, adaptive estimation of observation noise covariance is realized based on statistical characteristics of an innovation sequence, and process noise covariance is dynamically adjusted according to residual sequence changes. A Kalman filtering framework is adopted for state prediction and updating, and real-time estimation of the position and the speed of the user is achieved. The method accurately recognizes the transverse position track of the user in the running process, adapts to different noise environments and motion modes, and has the advantages of being high in precision, high in adaptability and good in reliability.

The invention discloses a dynamic multi-dimensional fast track association method and device based on a millimeter wave traffic radar. The method comprises the following steps: acquiring point cloud data acquired by a millimeter wave traffic radar; clustering the point cloud data through a DBSCAN clustering algorithm to obtain a point cloud clustering result; adopting a centroid algorithm to complete the condensation of trace points in the point cloud clustering result, and obtaining the condensed trace points; and carrying out association calculation on the condensed trace points and the currently managed track or historical trace points based on an association algorithm of an optimal association coefficient to obtain matched track information.

The application discloses a multi-modal fusion-based industrial inspection system inspection method, and belongs to the technical field of industrial inspection, comprising the following steps: S1, obtaining positioning data of a to-be-positioned node by using an HDS-TWR ranging algorithm and a Taylor total centroid algorithm, and outputting positioning coordinates; S2, measuring the acceleration and angular velocity of the inspection system by using a sensor, so as to evaluate the alignment movement and posture; S3, improving a Kalman filtering algorithm, so that the output prediction result is close to the true value; S4, adopting infrared-vision dual-mode cooperative tracking, and simultaneously using a time synchronization displacement compensation mechanism to calculate the relative displacement difference of the two tracking modes and perform dynamic calibration; and S5, processing video frames or pictures collected in the inspection process by using a YOLOv8 model, outputting a target frame and a category probability, and finally obtaining an inspection result; the multi-modal fusion-based industrial inspection system inspection method provided by the application provides a high-precision and low-delay automatic inspection solution for mines, nuclear power plants and other harsh environments.

The invention discloses a self-adaptive free space light tracking and pointing system and method. The system comprises but is not limited to a CMOS camera, a grayscale image module, a signal light mode field, an optical fiber mode field and a controller. The method comprises the following steps: calibrating an optical fiber mode field, generating a fixed-size convolution kernel and storing calibration data; capturing a light spot image and preprocessing to output a stable light spot image; dynamically searching a light spot region to output an effective region coordinate; calculating a convolution matching degree to generate a power matrix; and feedback control, adjustment and driving of the optical execution mechanism. According to the method, the optimal power point is accurately calculated according to tracking and pointing of the turbulence self-adaptive spatial light, a power point calculation mode of a traditional centroid algorithm is replaced, the influence of complex environmental factors such as atmospheric turbulence and system distortion on the transmission efficiency is effectively reduced, and therefore efficient coupling transmission of the signal light is achieved.

The invention discloses a high-dynamic trailing astronomical image exposure center moment image plane positioning method, which belongs to the field of telescope optical detection image processing, and comprises the following steps of: firstly, filtering a high-dynamic star image through an anisotropic Gaussian kernel to obtain an enhanced image and a directional diagram, and accurately extracting an exposure center position of an astronomical image based on the enhanced image and the directional diagram; performing linear fitting on each extracted astronomical image by using a least square method to obtain a movement track of the astronomical image; and calculating an integral midpoint as the exposure center position of the astronomical image along the movement track of the astronomical image through the gray accumulated sum in the normal direction of each point on the track. According to the method, the accurate pixel position of the trailing star image under the high-dynamic background is determined, the problem that the center position of the star image is difficult to accurately solve under the complex trailing condition caused by high-dynamic star image trailing and platform rotary table shaking can be effectively solved, and compared with a traditional centroid algorithm, the image surface positioning accuracy at the star image exposure center moment is remarkably improved; and finally, the performance of the star tracker is improved.

The invention relates to a memory data line training method, a memory controller, a processor and electronic equipment, and belongs to the field of electronic circuits. The method comprises the steps that a two-dimensional eye pattern corresponding to a target data line of the memory is acquired, the two-dimensional eye pattern comprises a plurality of test points located in a two-dimensional space formed by a voltage dimension and a time delay dimension, and each test point corresponds to a combination of a voltage value and a transmission time delay value; determining the center of mass of the two-dimensional eye diagram; and based on the coordinates of the centroid in the voltage dimension and the time delay dimension, determining a reference voltage configuration value and a transmission time delay configuration value corresponding to the target data line. By introducing the centroid algorithm in mathematics, automatic selection of the optimal equilibrium point of the eye pattern is realized, the centroid reflects the equilibrium center of the effective area of the eye pattern in geometric and physical distribution, and the centroid is taken as the sampling center point, so that natural equilibrium between the time domain and the voltage domain can be realized; therefore, the influence of external interference on the eye pattern shape is reduced to the minimum.

The application relates to a method for processing muon detection data of a triangular prism scintillator. S1: obtaining uncorrected energy deposition information and uncorrected hit position of each muon event in each layer detection unit; S2: based on the actual structure parameters of the detector, performing geometric correction on the traditional centroid algorithm; S3: based on the statistical relationship between the energy deposition of the muon in the scintillator and the track length, establishing a signal channel gain coefficient model, and correcting the energy response of each detection channel; S4: based on the fitting results of the muon track length and the energy of each channel, determining the energy consistency standard, and performing rejection processing on the muon events determined as abnormal; S5: performing muon track reconstruction and imaging, evaluating the position resolution of the detector, and if the convergence condition is reached, ending, otherwise returning to S2 for iteration. The technical scheme can improve the muon track reconstruction accuracy and imaging quality through geometric structure correction, energy correction and abnormal energy event screening.

The embodiment of the present invention discloses a water mist stabilization method and device suitable for a capacitive touch screen of a smart wearable device, the method comprising: extracting three consecutive frames of data after monitoring a line-drawing action on the capacitive touch screen; performing dynamic weighted calculation based on the three consecutive frames of data to obtain the coordinate value of each touch point; connecting the coordinate values ​​of multiple touch points to form a new line-drawing trajectory. Compared with the prior art, when there are a small amount of water droplets or a certain amount of water mist on the capacitive touch screen of a smart wearable device (such as a watch), the water mist stabilization solution for the capacitive touch screen provided by the embodiment of the present invention has the following advantages: 1. The front, middle and back three frames of data are used to perform dynamic weighted coordinate calculation on the capacitive screen data of a smart watch with water mist. Compared with a direct coordinate calculation algorithm, the situation in which the center of gravity algorithm is unstable due to water mist can be effectively suppressed. 2. The user experience can be improved. 3. The durability of the watch can be enhanced. 4. The application field can be expanded.

The invention discloses an indoor positioning system based on an ultra-wideband communication technology. The indoor positioning system comprises a base station, a label and an upper computer, the upper computer carries out data interaction with the data processing and storage module, the monitoring and data display module and the communication module, the upper computer receives data from a base station, the data processing and storage module extracts positioning information and comprises a preliminary positioning stage and an accurate positioning and error correction stage, and the positioning stage utilizes a trilateration method to realize accurate positioning and error correction. Calculating the initial position of the label according to the distance information between the label and the at least three base stations, and then further optimizing the initial position by utilizing a weighted centroid algorithm and combining the position information and the relative distance of the plurality of base stations to obtain an estimated position; in the accurate positioning and error correction stage, the position obtained in the initial positioning stage is used as the initial position estimation of a Taylor series expansion method, and the Taylor series expansion method is utilized to gradually approach the real position in an iterative approximation mode.

The application relates to the field of space laser communication, in particular to a face array APD multi-spot overlapping state simultaneous identification method and system based on code division multiple access, which comprises the following steps: three slave optical terminals respectively emit three beacon light beams to a master optical terminal; the three beacon light beams pass through a converging lens and converge to a 5*5 face array APD detector photosensitive surface to generate 25 APD pixel photoelectric current signals; the 25 APD pixel photoelectric current signals are input to a signal conditioning circuit, and a digital signal corresponding to each APD pixel photoelectric current signal is output; the digital signals are all input to a preprocessing unit to obtain a summary signal; the preprocessing unit inputs the summary signal to a code tracking loop to obtain an output signal; integrated processing is conducted on the output signal output by the code tracking loop and the digital signal output by the signal conditioning circuit to obtain an integrated processing result; and a centroid algorithm is used to solve the spot center position of the integrated processing result. The number of code tracking loops, the complexity and the scale of the circuit system are reduced.

The invention relates to the technical field of underwater wireless optical communication, in particular to a full-duplex underwater wireless optical communication device and method based on a beacon-free capturing and tracking technology. The method comprises the following steps: transmitting optical signal data with a specific divergence angle through a transmitting device to cover an opposite-end communication terminal area; a CCD camera is used for collecting light signal data, the collected light signal data is preprocessed, whether a light spot is captured or not is judged through the CCD camera, and if the light spot is not captured by the CCD camera, the light spot is captured by a photon counting camera; s2, adjusting the emission light power based on the light spot captured in S2 and the number of received photons obtained by a photon detector; based on the optical signal data after the optical power is adjusted, processing the light beam through the FPGA to obtain an actual angle deviation; and adjusting the deflection angle of the FSM based on the actual angle deviation to complete link tracking. And accurate positioning is realized through a weighted centroid algorithm, the problem of rapid light beam offset caused by shaking of the underwater platform is solved, and the tracking stability is improved.

The invention relates to a digital simulation platform, method and device of a wavefront sensor and a storage medium. The system comprises a Zernike polynomial calculation module, a micro-lens array simulation module, a centroid extraction module and a wavefront reconstruction module. The Zernike polynomial calculation module is used for analyzing and generating incident wavefront by adopting a Zernike polynomial, the micro-lens array simulation module is used for simulating a micro-lens array by adopting a mixed diffraction propagation model to segment the incident wavefront to generate a light spot image, and the centroid extraction module is used for extracting the light spot image by adopting a weighted centroid algorithm. The light spot image acquisition module is used for acquiring a light spot image and calculating a mass center position of the light spot image in combination with global and local dynamic threshold mechanisms, and the wavefront reconstruction module is used for establishing a relationship between the mass center position and a wavefront slope, solving a Zernike coefficient by using a numerical optimization algorithm, generating wavefront aberration distribution and outputting a reconstructed wavefront phase. Through the modular design, the independence of each function is ensured, the simulation efficiency is improved, and the data generation capability and the system flexibility are enhanced.

The invention discloses an accurate positioning system and method for mine personnel and mobile terminal equipment, the system can automatically correct a reference error caused by roadway deformation or equipment displacement by introducing a dynamic baseline real-time measurement technology, a multi-source data fusion mechanism is adopted, the arrival time difference and a received signal strength value are synchronously utilized, and the positioning accuracy of the mine personnel and the mobile terminal equipment is improved. And resolving is carried out through a weighted centroid algorithm, so that the robustness and the positioning stability of the system are remarkably improved. An inertial measurement unit is innovatively integrated, a signal intensity value is intelligently compensated by sensing a walking direction of a person and combining a human body shielding model, a measurement error caused by human body shielding is effectively reduced, a multipath interference index is extracted by analyzing a channel pulse response, an algorithm weight is dynamically adjusted accordingly, and the accuracy of human body shielding is improved. According to the method, environment self-adaptive positioning optimization is realized, the technical bottlenecks of insufficient precision and poor environment adaptability of the traditional positioning technology are effectively solved, and reliable technical guarantee is provided for safe production of mines.

The invention discloses a remote fault on-line positioning processing method, system and platform based on a floating tap sliding window and a storage medium, and aims to realize instrument-free and on-line automatic positioning of a remote fault of a high-speed link by configuring a floating tap in a decision feedback equalizer into a slidable detection window. The detection limitation that a standard fixed tap is only about 5cm is broken through, the effective diagnosis range is extended to 30cm through a parameter self-adaptive window sliding strategy, and a back plate and a key area at the tail end of a cable are completely covered. The spatial resolution is improved to a submillimeter level by combining energy detection with a centroid algorithm; through an original distance energy double-criterion mechanism, false peak values generated by multiple reflections can be effectively identified and inhibited, and only real physical faults are reported. Existing hardware resources of the chip are utilized in the whole process, external testing equipment or service interruption is not needed, and an efficient and low-cost automatic solution is provided for mass production testing, on-site operation and maintenance and monitoring of the high-speed link.

The application discloses a kind of element pin detection methods based on machine vision, comprising the following steps: step one: using industrial camera obtains the 3D point cloud of the detected element including pin;Step two: based on the 3D point cloud obtained in step one, obtain height map;Step three: in height map, select a height reference value, and form a selection interval on the height reference value according to the set height difference value;Step four: the point cloud of the sampling point in the selection interval determined in step three is used to form a gray scale image;Step five: in the gray scale image, the center of gravity of the pin is obtained using Blob analysis algorithm and center of gravity algorithm;Step six: use the center of gravity of the pin obtained in step five to judge defects, so use height map to set selection interval to select effective point cloud, so noise point cloud can be filtered, avoid noise point cloud adversely affect gray scale image, to form the real gray scale image of pin, and then judge the defects of pin.

The invention relates to a livestock and poultry positioning and tracking method and device based on the Internet of Things, equipment and a medium. The method comprises the following steps: acquiring historical position and relative distance data through an Internet of Things terminal deployed on livestock and poultry, and constructing a group topological relation model; when a terminal failure is detected, M normal terminals with the highest correlation degree are dynamically screened based on the model to form a collaborative node set; fusing the real-time position data of the cooperative nodes and the association strength parameters in the model, and calculating the position coordinates of the livestock and poultry where the invalid terminal is located in real time through a weighted centroid algorithm; and finally, performing state prediction and observation updating on the continuous position estimation points based on a Kalman filtering algorithm, and generating a smooth and continuous motion track during a terminal failure period. According to the scheme, a dynamic cooperation mechanism of group intelligence is utilized, space correlation characteristics among individuals are converted into a positioning compensation basis, complete trajectory tracking can still be maintained when equipment fails or signals are interrupted, and the robustness and data continuity of a livestock and poultry positioning system in a complex environment are improved.

This invention discloses a machine vision-based method and system for monitoring the seal of an internal floating roof, belonging to the field of machine vision technology. It includes: acquiring laser stripe images collected by a monitoring unit; performing binarization and connected component analysis on the laser stripe images to delineate the region of interest, and extracting the center line of the laser stripes with sub-pixel precision using a grayscale centroid algorithm; establishing a triangulation model based on a pre-calibrated camera intrinsic parameter matrix and laser plane equation, converting the center line into a three-dimensional contour point cloud of the seal, and performing time-series filtering; retrieving the reference contour point cloud corresponding to the circumferential encoder angle, calculating the normal deviation value and tangential gap value; determining the fault type based on multi-level comparison results with preset thresholds, and generating a fault spatial index by fusing circumferential azimuth angle and liquid level data. This invention obtains high-contrast stripes under extremely dark lighting and metal reflection environments in storage tanks, enabling quantitative measurement of the three-dimensional deformation of the seal and preventative maintenance.

The invention relates to a microsphere high-precision positioning method based on deep learning deblurring, and aims to solve the problem of defocusing blurring caused by movement of microspheres in the optical axis direction in the microscopic imaging process so as to effectively improve the three-dimensional positioning precision of the microspheres. According to the method, a deep learning image restoration network is introduced, real-time sharpening processing is carried out on a blurred image, and gray features used for microsphere center positioning are effectively reconstructed. In the network model training process, a structural similarity index is introduced as a performance evaluation standard, and the restoration quality of the image is improved by optimizing a loss function. The transverse positioning achieves nanoscale precision by adopting a gray centroid algorithm, and the axial positioning is realized by constructing a radial gray vector of an image and matching the radial gray vector with a calibration model. The method provided by the invention has the characteristics of high measurement precision, strong robustness and the like, and is suitable for particle tracking, micro-nano operation, robot micro-assembly and other precise positioning tasks with a relatively large space motion range.

The present invention relates to the technical field of computer vision and image processing, and particularly relates to a method and system for evaluating the matching of court tactical segments based on graphic similarity. The method includes: obtaining a court video and dividing it into a plurality of court tactical segments through frame segmentation, constructing a coordinate system based on the court, and obtaining the position data of each moving target in each frame; converting the position data into a plurality of closed polygons, and determining the corner function of the closed polygons; calculating the corner function of any court tactical segment, corresponding and matching it with the tactical formations in the tactical template library, and determining the tactical graphic similarity between the court tactical segment and the corresponding and matched tactical formation to obtain the score of any court tactical segment; quantifying the graphic similarity index through the centroid algorithm and the corner function, and performing matching scoring, which solves the problem of low efficiency of traditional methods and eliminates subjective biases.

The application provides a strabismus nine-orientation image classification method and system, which comprises the following steps: based on a preset recognition algorithm and a pre-constructed coordinate system, boundary recognition is performed on a pre-acquired eye image to obtain initial left eye boundary box coordinates and initial right eye boundary box coordinates, a deviation angle is calculated and the eye image is corrected according to the deviation angle to obtain target left eye boundary box coordinates and target right eye boundary box coordinates; based on a preset centroid algorithm and the corrected eye image, target left eye iris centroid coordinates and target right eye iris centroid coordinates are calculated; a preset prediction model is used to obtain a nine-orientation distribution probability of the eye image; based on preset nine-orientation mutual exclusion requirements and the nine-orientation distribution probability of the eye image, the eye image is classified to obtain a strabismus nine-orientation image classification result. The application realizes accurate positioning of the iris position of the eye image, reduces irrelevant information interference, and classifies the iris by combining the mutual exclusion characteristics among the nine orientations, thereby improving the reliability of the eye image classification.

The invention relates to the technical field of optical measurement and signal processing, and discloses a high-precision center control target scoring system which comprises a laser phase shift interference array, a high-speed phase demodulation unit and a center control processing unit. According to the system, a laser phase shift interference array is used for forming a laser sensing surface behind a target surface, and the laser sensing surface is used for detecting air refractive index change caused in a transient wake of a projectile after the projectile penetrates through the target in a non-contact manner and converting the change into an instantaneous phase shift signal. And the central control processing unit accurately calculates the coordinates of the impact point through a weighted centroid algorithm based on the peak value of the signal. In addition, the system also constructs a phase disturbance characteristic vector by extracting a plurality of characteristic parameters of the phase signal, and performs mode matching with an expert knowledge base to realize synchronous diagnosis of the flight state of the projectile. The method does not depend on a special target plate, is high in anti-interference capability, not only realizes high-precision target scoring, but also provides an additional ballistic analysis function, and has the characteristics of high stability and low operation cost.

The invention relates to the technical field of sensor detection, in particular to a non-contact human body posture detection device based on an ultrasonic array. Comprising a shell, an ultrasonic transmitting unit, a two-dimensional ultrasonic receiving array, a signal conditioning module, a distance detection and multi-mode adaptive control module, a dual-mode automatic calibration module, an STM32 single-chip microcomputer control module, an output interface module and a power supply module. The ultrasonic transmitting unit is used for transmitting a 40kHz-60kHz pulse signal by adopting a dynamic beam forming technology; the two-dimensional receiving array obtains spatial amplitude distribution information; the multi-mode self-adaptive control module dynamically adjusts the gain and the threshold by combining the distance and the signal-to-noise ratio; the dual-mode automatic calibration module corrects channel errors through starting / periodic calibration; the single-chip microcomputer module combines an energy centroid algorithm and an attitude feature library to realize attitude recognition with response less than or equal to 50ms. The device has high detection precision and strong privacy, has power consumption of less than or equal to 2W, and is suitable for the scenes of old-age nursing, game interaction, smart home and the like.

The present invention discloses a wavelength division multiplexing centroid detection method and system, comprising generating two-dimensional modulation matrices A, B, and C, loading them into a wavelength division multiplexing optical modulator, modulating illumination light of three different wavelengths according to three two-dimensional modulation information patterns, and then illuminating a target object through a light combining prism and a projection lens; separating the reflected or transmitted light of the target object into three reflected or transmitted light of different wavelengths, which are received by a single-pixel detector assembly; and calculating the centroid position based on the received optical signal information and a centroid algorithm. The present invention can directly detect the centroid of a target object without imaging. Combined with a wavelength division multiplexing method, the frame rate of centroid detection using a single-pixel detector is no longer limited by the modulation frequency of the optical modulator, further improving the centroid detection speed.

The application discloses a weak pressure measurement method based on a stable buoyancy system and an image centroid algorithm, and belongs to the technical field of precise mechanics measurement. The method first constructs a self-stable buoyancy rotation platform supported by high-viscosity silicone oil and an optical lever amplification light path, and eliminates mechanical friction by means of the difference between the gravity moment and the buoyancy moment; a weak pressure to be measured is applied to make the platform produce a slight rotation angle, the rotation angle is amplified into linear displacement of a reflected light spot through the optical lever principle, and a video is collected; red channel extraction and multi-frame superposition and averaging are carried out on the video frames to suppress liquid surface fluctuation noise; a threshold segmentation and a morphological operation are used to lock a light spot connected domain, a weighted pixel centroid algorithm is used to calculate the light spot center coordinates; and the light spot displacement is converted into the pressure value to be measured according to a calibrated linear response model. The application realizes high-precision measurement of weak pressure in the order of micro-newton, and the measurement resolution is better than 1 muN.

The invention provides a geometric level automatic measurement method and system based on CCD encoding and decoding, and the method comprises the steps: scanning a bar code ruler through a CCD sensor, and obtaining a bar code digital signal; obtaining a current measurement distance to generate a corresponding interpolation step length, and performing sub-pixel linear interpolation on the bar code digital signal to obtain an interpolated bar code digital signal; extracting a wave crest and a wave trough of the interpolated barcode digital signal, and screening wave crest data meeting a matching condition based on a wave crest interval to obtain a wave crest section; based on the wave crest segment, the wave trough and the standard bar code, performing bar code matching through a matching algorithm to obtain an initial bar code sequence; correcting the initial barcode sequence based on a centroid algorithm to obtain a target barcode sequence; and decoding the target barcode sequence based on the coding rule to obtain a geometric level measurement result so as to complete automatic geometric level measurement. According to the method, sub-pixel-level positioning precision is provided by adopting bar code matching, and the accuracy of geometric leveling results is improved.

This invention discloses a dynamic multidimensional rapid track association method and device based on millimeter-wave traffic radar. The method includes: acquiring point cloud data collected by millimeter-wave traffic radar; clustering the point cloud data using the DBSCAN clustering algorithm to obtain point cloud clustering results; using the centroid algorithm to agglomerate the points in the point cloud clustering results to obtain agglomerated points; and using an association algorithm based on the optimal association coefficient to perform association calculations between the agglomerated points and currently managed tracks or historical points to obtain matched track information.

This invention discloses a method, apparatus, device, and storage medium for wireless optical communication beam alignment. The method includes: acquiring a real-time beacon image; calling a target detection model to perform target detection; and outputting the beacon beacon bounding box and corresponding detection score; determining an initial region of interest (ROI) based on the beacon beacon bounding box, determining an expansion factor based on the detection score, adaptively expanding the ROI, and performing boundary enhancement processing on the expanded ROI to obtain an enhanced ROI image; using a centroid algorithm to perform sub-pixel-level localization of the beacon center position in the enhanced ROI image to obtain the sub-pixel centroid coordinates of a single beacon beacon; comparing the sub-pixel centroid coordinates with the center coordinates of the camera imaging plane to obtain a deviation; calculating a control quantity based on the deviation to drive an actuator to adjust the beam direction, thereby achieving closed-loop beam alignment. This invention enables fast, stable, and sub-pixel-accurate beacon light recognition and beam alignment in complex environments.

The invention discloses an imaging type array ICMOS focal plane splicing method, and solves the technical problems that no splicing reference is available and the splicing precision is not ideal in the existing splicing method. The method comprises the following steps: converting a coordinate reference of a splicing instrument to a central point of a cross wire on a target plate; the splicing base is leveled; obtaining two-dimensional installation coordinates of all the ICMOS; the current optimal focal plane positions of all the ICMOSs are sequentially obtained, the lowest one of the current optimal focal plane positions of all the ICMOSs is used as a reference focal plane position, and the current optimal focal plane positions of other ICMOSs are adjusted, so that the coplanarity of the optimal focal plane positions of all the ICMOSs meets the design requirement; and the detection lens is combined with a centroid algorithm to complete adjustment of all ICMOS rotation deviations and straightness in sequence, so that array ICMOS focal plane splicing is realized. According to the method, high-precision splicing of the array ICMOS focal plane can be realized.