251 results about "Pixel matrix" patented technology

The invention provides a building main body construction quality tracing method and system, and relates to the technical field of building construction management, and the method comprises the steps: dividing region blocks based on a pixel matrix of a construction site image, carrying out the matching of each region block with a preset control node coordinate in a building information model, recognizing and positioning a target node, generating node image data; based on a target node associated with the node image data, collecting construction environment parameters and physical state parameters of the target node, and generating a multi-dimensional parameter set based on the timestamp index; performing parameter fitting calculation on the multi-dimensional parameter set to generate a parameterized label sequence for representing the construction progress and the quality state; encrypting the parameterized tag sequence and storing the parameterized tag sequence into a distributed account book to form encrypted evidence storage data; and based on the encrypted evidence storage data, quality responsibility tracing and construction record verification are carried out, and a quality tracing verification result is generated. And comprehensive and accurate real-time monitoring on the construction progress and the quality state can be realized.

The invention discloses a monitoring camera image anomaly detection method and system, and relates to the technical field of intelligent monitoring, and the method comprises the steps: obtaining an original video stream, extracting a pixel matrix of a current frame, carrying out the joint coding of the pixel matrix and a motion vector field of an adjacent frame into a quantum state bit vector, inputting the quantum state bit vector into a pre-constructed anomaly detection operator, and carrying out the detection of the anomaly. Executing quantum state evolution calculation in the Hilbert space, and outputting a two-dimensional distribution matrix of each pixel region; integrating the marked binary image and the physical verification conclusion, and when the quantum anomaly probability value reaches the alarm standard and passes the physical verification, outputting an anomaly alarm signal; and the abnormal alarm signal is converted into a confrontation training sample, a cross-scene virtual sample is generated through quantum noise injection, and an abnormal detection operator is dynamically updated. According to the method, the Hilbert space quantum state evolution step is combined, unitary transformation is applied to the quantum state bit vector, the orthogonal projection operation is executed, the accurately quantified abnormal probability distribution matrix is generated, and accurate assessment of the abnormal risk is achieved.

The invention provides an image encryption method based on chaotic mapping and dynamic positioning generalized knight cruise, and the method comprises the following steps: reading a plaintext image P, and converting the plaintext image P into a pixel matrix with the size of the plaintext image P; a chaotic pseudo-random sequence is generated through three-dimensional Logistic chaotic mapping, and three integer pseudo-random sequences are obtained through integer processing and serve as parameters of a pixel matrix for a knight cruise scrambling method; two of the obtained integer pseudo-random sequences are used for controlling cruising of a rider on rows and columns respectively, so that scrambling operation is completed; and generating a chaos matrix by using a third integer pseudo-random sequence, carrying out XOR operation on the chaos matrix and the scrambled image matrix, changing an image pixel value, and completing image encryption to generate a ciphertext image C. The dynamic positioning generalized knight cruise scrambling method designed by the invention is simple in process, and the encryption efficiency can be remarkably improved.

A method and device for remote sensing image classification using quantum pixel matrix entanglement is provided. The method comprises: preprocessing an acquired raw remote sensing image to obtain image data containing multiband fusion; and calculating a pixel matrix entanglement coefficient μ and a Euclidean distance d between a cluster center and other pixel matrices based on a preprocessed remote sensing image, performing an iterative self-organizing classification on remote sensing image data based on the pixel matrix entanglement coefficient and a Euclidean distance threshold, and obtaining remote sensing image classification results.

The invention discloses a color image encryption method based on deep learning and a four-dimensional Lorentz chaotic system, and belongs to the field of information security and image encryption. According to the method, a CNN deep learning network is constructed, a color image is divided into RGB channels, features of a pixel matrix of each channel are extracted as initial conditions of a chaotic differential equation, and a data set is obtained through solving by a variable-step-size fourth-order Runge-Kutta method and used for training the CNN network. And carrying out first round encryption on the pixel matrix by using a key sequence generated by the trained CNN network, carrying out Arnold scrambling and forward and reverse recursive diffusion on the pixel matrix by using a random chaos sequence generated by an improved four-dimensional Lorentz equation to complete second round encryption, merging the RGB three-channel pixel matrix, and outputting a final color encrypted image. The problems that an existing color image encryption method is low in encryption strength and cannot make full use of color image characteristics are solved.

The invention relates to the technical field of bill text recognition, and discloses a bill text recognition system and method based on deep learning. The method comprises the following steps: acquiring target bill original image data containing a multi-channel pixel matrix and spatial resolution information; based on a matching result of the bill edge features and a preset template, geometric distortion correction is carried out on the original image, and a corrected bill image is generated; inputting the corrected image into a pre-training text region detection network to obtain positioning information containing text line boundary coordinates and region confidence; text line image blocks are extracted according to the positioning information, character segmentation preprocessing is executed, and a character-level image sequence is generated; calling a deep character recognition model to classify the sequence character by character, and generating an initial text recognition result; semantic verification and error correction are performed on the initial result based on the bill type knowledge base, final structured text data are generated, and the processing requirements of bills of different types and qualities can be met.

A method includes receiving, by a computing device, an image indicative of a cross-section of a thermally-sprayed layer. The thermally-sprayed layer defines a porosity comprising a void volume of the thermally-sprayed layer. The image comprises a matrix of pixels, each pixel in the matrix of pixels defining a respective luminance value of a plurality of luminance values. The method includes identifying, based on the luminance values, at least one pixel that is indicative of a void volume in the thermally-sprayed layer. The method includes calculating, based on the at least one pixel that is indicative of the void volume in the thermally-sprayed layer, a total porosity of the thermally-sprayed layer. The method includes determining, by the computing device and based on the at least one pixel that corresponds to a void volume in the thermally-sprayed layer, a quantification of a spatial homogeneity of the porosity of the thermally-sprayed layer.

The present application relates to a kind of for the determination method of liquid alloy density in electrostatic suspension state in space station, after obtaining the photo of liquid alloy in electrostatic suspension state in space station, first photo is converted into pixel matrix and carries out gray scale processing, then deduct photo background by matrix operation.Further, the best binary grayscale threshold of pixel matrix is calculated, and pixel matrix is carried out binary processing.Then all boundary coordinates in pixel matrix are obtained by matrix operation, and each connected boundary is distinguished.Excluding bright spot boundary, the boundary coordinates and center coordinates of liquid alloy are obtained.Then the boundary equation of liquid alloy is obtained by using multi-order Legendre equation fitting liquid alloy boundary coordinates.Finally, the volume of liquid alloy is obtained by rotating integral to the boundary equation of liquid alloy, and then the density value of liquid alloy is obtained.The density value of the present application can be obtained by processing liquid alloy photo, and can be used for the determination of containerless liquid alloy density in electrostatic suspension in space station.

The invention relates to the technical field of video processing, in particular to a target area video compression quality evaluation method based on video dynamic semantic segmentation. The method comprises the following steps: acquiring an original video sequence; performing pixel-level semantic segmentation on each frame of the original video sequence by using a pre-trained video semantic segmentation model, and outputting a binary target area mask; performing compressed encoding on the original video sequence, and outputting a compressed video sequence; according to the binary target area mask, extracting a pixel matrix of a target area in the original video frame and the compressed video frame; and calculating the PSNR of the original video frame and the compressed video frame in the target area pixel by pixel based on the pixel matrix, and solving the average value of the PSNR values of all pixels to obtain the average PSNR value of the single-frame target area. According to the method, quality degradation during scene switching or strenuous movement is accurately captured, and the evaluation sensitivity of the target area is improved compared with that of a traditional method.

The invention relates to the technical field of intelligent peripherals, in particular to a user-oriented interaction information expansion method, which comprises the following steps that: each keycap acquires displacement and pressure, generates an event packet, uploads the event packet through an optical wireless link, and obtains a clock and supplies power; constructing a keycap graph by a master controller, obtaining an intention vector through graph convolution and pooling, inputting the intention vector into a cyclic network to generate a pixel matrix and a phase matrix, and downlink after Bernoulli sparse coding; performing key cap side orthogonal matching pursuit decoding and four-round gradient optimization, and outputting a differential pixel set, a differential phase set and a synchronization mark; the master control refreshes the digital micromirror array and the ultrasonic phased array in a differential manner, locks a frame-level time sequence through a synchronization pulse, compensates link delay and hardware drift in real time based on Kalman filtering, and keeps an optical-acoustic alignment error to be less than 100 microseconds; degradation monitoring enables a link to be automatically switched to a pre-carved character layer and to be quickly recovered when the link is abnormal, and a keyboard interaction information expansion scheme with low delay and high consistency is provided.

The invention discloses an intelligent glass fiber reinforced plastic curing temperature control method and system in combination with visual detection, and relates to the field related to intelligent manufacturing, and the method comprises the steps: carrying out the collection of a morphological spectral image and an infrared thermogram of a glass fiber reinforced plastic curing surface through a multispectral visual detection sensor; establishing a multi-dimensional pixel matrix including surface texture, reflection spectrum and infrared temperature distribution; carrying out polymerization analysis on optical coherence and temperature distribution continuity, constructing a photo-thermal coupling consistency tensor, and adaptively generating a curing region partition map; calculating a temperature gradient vector field of each adjacent region, and constructing a control mapping function by taking a morphological response characteristic index as a weight parameter; and a differential heating instruction is configured to control the multi-point array heating unit to execute self-adaptive heating control. The technical problem that existing glass fiber reinforced plastic curing temperature control cannot comprehensively and accurately sense and dynamically adapt to complex changes of the curing process is solved, and the technical effect of intelligent, accurate and differential control over the glass fiber reinforced plastic curing temperature is achieved.

The invention discloses a pixel extraction and conversion method of an LEN format image, and belongs to the technical field of digital image processing. In order to solve the technical problems that the LEN format has no public specification and general software cannot read data, binary analysis is carried out through an LEN sample, and the core binary layout of a file is found for the first time: the image width and height are stored in specified offset positions, and a small-end byte order is adopted; pixel data comprises a compressed data segment and a row information segment, and an RLE coding rule is adopted. The method comprises the following steps: acquiring an LEN format file; analyzing the file header to obtain an image size; performing segmented analysis on the image data; synthesizing the analyzed data into image data; and packaging the pixel matrix into a standard TIFF format file. The LEN core binary layout is disclosed for the first time, zero distortion and cross-platform conversion are achieved, reading and storage of screening parameters are not involved at all, existing patents are effectively avoided, and the method can be used for preview, filing and third-party system integration in the flexographic printing process.

A method includes receiving, by a computing device, a raw image indicative of a cross-section of a thermally-sprayed layer. The image includes a matrix of pixels, and each respective pixel in the matrix of pixels defines a luminance value. The method may further include determining, based on the luminance values, at least one pixel that corresponds to an oxide component in the layer and removing the at least one pixel that corresponds to the oxide component in the layer to generate a modified matrix of pixels. The method may further include generating an oxide-filtered image based on the modified matrix of pixels. The method may further include converting, by the computing device and based on the luminance values, the oxide-filtered image into a binary image and determining, by the computing device and based at least partially on the binary image, a porosity of the coating layer.

The application discloses a high-uniform seamless splicing system and a brightness adjusting method for light field display, and belongs to the technical field of light field display. The system uses a minimum display unit modular splicing architecture to realize super-high resolution display, adopts a composite microlens array to realize axial displacement of an imaging surface, covers a gap between adjacent display units through optical expansion of a pixel matrix, and eliminates parallax faults caused by physical splicing seams. The system integrates a distributed dynamic uniform light module, collects brightness data of a display area in real time, performs brightness compensation on a low-illumination area through array light source driving control, changes a light transmittance, and regulates and controls pixel-level brightness uniformity of the system. The scheme effectively solves the problems of pixel continuity loss and adjacent dimming area light coupling in high-resolution light field reconstruction. The system is widely applicable to light field three-dimensional display, and is a scheme for improving three-dimensional display effect.

A CMOS image sensor includes a unit pixel array including a photodiode array, a color filter array, a micro-lens array, and a grid isolation structure laterally separating adjacent color filters. The grid isolation structure includes a first low-n grid, a second low-n grid underlying the first low-n grid, and a metal grid within the second low-n grid, the first low-n grid being narrower than the second low-n grid. The color filter array includes color filter matrixes, all color filter matrixes have the same arrangement pattern. Sizes of color filters in each color filter matrix vary depending on locations of the color filters in the color filter matrix. In an edge portion, a distance between a center of a color filter matrix and a center of a corresponding unit pixel matrix in plan view varies depending on a location of the unit pixel matrix in the CMOS image sensor.

The invention discloses a method for calculating large-area displayable graphic and text information under a point light source. The method comprises the following steps: S1, digitizing the graphic and text information; s2, point light source parameter modeling; s3, carrying out basic modeling on the microscopic fringes; s4, stripe width correction is carried out in combination with an image-text information model; and S5, carrying out block calculation and splicing on the large-area region. By digitally processing the graphic and text information, the complex graphic and text can be converted into an accurate pixel matrix, a clear data basis is provided for subsequent calculation, and the accuracy of the graphic and text information at a processing source is ensured; point light source parameter modeling is carried out, the position, illumination intensity and other key parameters of a point light source can be accurately captured, a reliable optical basis is provided for subsequent microcosmic fringe design, and calculation deviation caused by fuzzy light source parameters is avoided. And a brand new microscopic stripe basic model is constructed, so that stripe distribution can more flexibly adapt to point light source characteristics and graphic and text information requirements.

The invention relates to the field of image processing, in particular to a laser marking positioning method and system based on vision, and the method comprises the steps: obtaining a template RGB image and a to-be-positioned RGB image of a bottom plate of an electric iron; constructing a pixel matrix set of the target hole, and calculating the distinction degree between the target hole and any other hole; calculating the positioning confidence of the target hole under the target scale, traversing to obtain the positioning confidence of the target hole under each scale, and determining a positioning hole according to the positioning confidence; calculating the positioning weight of any pixel matrix in the pixel matrix set of any positioning hole, and taking any steam hole in the RGB image to be positioned as a hole to be matched so as to calculate the positioning matching degree of the hole to be matched and any positioning hole at the target scale, and calculating the comprehensive matching degree of the to-be-matched hole and any positioning hole based on the positioning matching degree and the positioning confidence accumulated value. According to the technical scheme, the marking and positioning precision can be improved.

The application relates to the fields of information security and image processing, and discloses a multi-image parallel encryption processing method based on an improved chaotic mapping, which comprises the following steps: separating and stacking each color channel of multiple images to be encrypted to construct a three-dimensional pixel matrix; extracting a matrix hash digest to be converted into a sub-key; inputting the sub-key into an improved chaotic mapping model of a historical state variable to iteratively generate a scrambling and diffusion sequence; performing sorting scrambling on the three-dimensional matrix in multiple orthogonal dimension sections by using the sequence; mapping the sequence into a shift step length to perform a cyclic shift and a bitwise XOR linkage diffusion on the pixels; and finally, outputting ciphertext images by segmentation. The application breaks the physical isolation of a single image to realize spatial depth mixing by constructing a three-dimensional matrix and performing cross-dimension scrambling; the chaotic dynamics degradation under limited precision is overcome by means of a historical state delay; and the efficiency of multi-image encryption and the system attack resistance are improved by using a shift and XOR linkage to realize bit-level depth confusion.

The present invention relates to a light emitting display device (200) comprising a matrix of pixels wherein each pixel is composed of three basic light emitting regions (201a, 201b, 201c), each basic light emitting region comprising an OLED stack (205a, 205b, 205c) having a different color, and each region being protected by a photoresist (280); two adjacent control electrodes (202a, 202b, 202b, 202c, 202c, 202a) belonging to two adjacent light-emitting regions (201a, 201b, 201b, 201c, 201c, 201a) of different basic light-emitting region groups are separated by a trench (211), and have insulating surfaces (223) covering the side walls of the adjacent electrodes (202a, 202b, 202b, 202c, 202c). Two adjacent control electrodes (202a, 202a, 202b, 202b, 202c, 202c) of two adjacent light-emitting regions (201a, 201a, 201b, 201b, 201c, 201c) belonging to the same basic light-emitting region group are separated by a trench (211) which is filled with a filling element (203), and the filling element (203) is arranged between the two adjacent control electrodes (202a, 202a, 202b, 202b, 202c, 202c) of the two adjacent light-emitting regions (201a, 201a, 201b, 201b, 201c, 201c) belonging to the same basic light-emitting region group. Wherein, in each group of basic light-emitting regions, the combination of the OLED stack and the upper electrode thereof is separated from the adjacent combination by an encapsulation layer (290).

The present application relates to organic light emitting display data processing technical field, especially high resolution organic display driving waveform generation method, through distributed parameter analysis, pixel matrix is divided into spatial block, using GPU node parallel extraction parasitic parameter and construct cross block coupling model, generate block parasitic parameter mapping table and characteristic vector matrix;Based on the residual charge data and dynamic gray scale segment control generation contains curvature parameter segmented waveform parameter set;Call brightness and voltage response data to execute nonlinear brightness correction, combined with light intensity feedback loop to generate compensation pulse parameters;According to the display content feature classification switch multi mode driving strategy, through phase locked loop phase synchronization to eliminate time sequence fault, the present application effectively suppresses parasitic effect interference and brightness mutation overshoot, improves the gray scale compensation accuracy and refresh rate synergy, enhances the display uniformity and dynamic picture smooth.

The invention provides a rice monitoring method, device and system based on multi-modal data fusion, and the method comprises the steps: obtaining the spectrum and image data of a rice canopy, respectively extracting the features of the data, inputting the features to a trained rice monitoring model, and obtaining a monitoring result; the rice monitoring model is obtained based on the following steps: preprocessing rice canopy spectrum and image data to obtain a feature vector and a pixel matrix; and training a multi-task learning model composed of a one-dimensional convolutional neural network and a two-dimensional coupon neural network based on the sample feature vector, the pixel matrix and the corresponding label matrix. Aiming at the problems of tedious flow, low efficiency, poor real-time performance and the like in the existing rice nitrogen and chlorophyll detection and pest and disease damage identification technology, rapid in-situ detection of the nitrogen and chlorophyll contents of rice leaves and efficient pest and disease damage identification are realized through a map fusion technology; the technical problems that an existing rice monitoring method is prone to misjudgment and missed judgment and low in efficiency can be solved.

The application relates to the technical field of image processing, and discloses a light field image dataset processing method and system based on differential privacy. The application groups light field image datasets to obtain a light field image data group and a group number, numbers sub-images to obtain image numbers and first pixel matrices through pixel extraction; encodes the image numbers, the first pixel matrices and the group number to obtain one-dimensional encoding data of the sub-images and stores the one-dimensional encoding data; processes the one-dimensional encoding data based on a differential privacy processing algorithm to obtain one-dimensional encoding data of the sub-images; and rearranges the sub-images based on the processed one-dimensional encoding data of the sub-images to obtain a light field image dataset protected by differential privacy. The one-dimensional encoding and rearrangement ensure the integrity of light field image array information, avoid the loss of position information caused by separately processing independent images through differential privacy, and ensure the availability of the light field image dataset while protecting privacy.

The invention relates to a naked-eye 3D display system based on a multi-layer phase modulation composite grating, belongs to the technical field of stereoscopic display, and aims to overcome the technical bottlenecks of low brightness of an existing parallax barrier grating, large crosstalk caused by large thickness of a cylindrical lens and serious dispersion of a diffraction element. The system comprises a two-dimensional display panel and a composite diffraction grating, pixel units of the display panel are composed of red, green and blue sub-pixel matrixes and are coded according to viewpoints. The composite diffraction grating is attached to the light emitting side of the panel, is made of an optical anisotropic material and is of a multi-layer structure, and a first micro-nano structure layer and a second micro-nano structure layer are sequentially arranged along a light path; the first layer is an aspherical focusing array, and the phase delay of the aspherical focusing array meets a quadric surface function so as to collimate light waves; the second layer is a dispersion compensation type deflection array, and the grating period of the second layer is accurately matched with the central wavelength of the sub-pixels, so that light of different colors is projected at the same deflection angle. According to the invention, full-color dispersion-free display is realized, the thickness of the device is less than 50 microns, and crosstalk is obviously reduced.

This paper proposes a chaotic image encryption method based on the combined scrambling of the improved Josephus transform and the Knight's Tour. First, a chaotic pseudo-random integer sequence is generated and integerized using a memristive chaotic system. Then, the pixel matrix is ​​divided into several small cell pixel matrices. Within each cell pixel matrix, the chaotic pseudo-random integer sequence is used to control the counting starting point, counting period, and counting direction, respectively. Block scrambling is performed using the improved Josephus scrambling algorithm. Subsequently, each cell pixel matrix after the improved Josephus block scrambling is treated as a checkerboard point, and the direction of the Knight's Tour is controlled using the chaotic pseudo-random integer sequence. The image is scrambled using the improved Knight's Tour algorithm and the improved Josephus scrambling algorithm, effectively reducing the correlation between adjacent pixels in the scrambled image and improving the global encryption effect of the image. Finally, the chaotic pseudo-random integer sequence generated and integerized by the memristive chaotic system is used to diffuse the pixels, enhancing the security of the encrypted image.

The application discloses an image encryption method and system based on a chaotic system, and relates to the technical field of image processing.The method comprises the following steps: a chaotic system is constructed by combining a cellular automaton with a tent map; the coordinate values of pixel points are subjected to coordinate scrambling; all the pixel points are formed into a pixel matrix, the pixel matrix is traversed, and an initial encryption matrix is obtained; a pseudo-random sequence is obtained through the chaotic system, arranged into a row matrix and a column matrix, and subjected to exclusive or operation to convert a plaintext image into a two-dimensional matrix; the initial encryption matrix is converted into a two-dimensional encryption matrix; based on a CA model of the cellular automaton, the initial encryption matrix and the two-dimensional encryption matrix are subjected to pixel point value calculation and assignment, and a target encryption image is obtained.The application can solve the technical problems of the prior art, such as that the image encryption method based on the chaotic system is not sensitive to the plaintext image, cannot effectively defend against differential attacks, has poor robustness, and has the problems of periodicity and a generally small key space.

The invention relates to a method and system for preventing burning of an infrared detector, and the method comprises the steps: reading the data of the infrared detector, and obtaining pixel points in a continuous multi-frame image and monitoring points in a pixel dot matrix; judging the burning state of the infrared detector according to the monitoring points in the pixel dot matrix; and according to the burning state of the infrared detector, protecting the infrared detector according to preset anti-burning measures. According to the anti-burn method disclosed by the invention, the burn state of the detector can be detected in real time, the burn state and degree of the detector can be obtained, burn is actively blocked, the protection effect is good, and the detector can be protected for a long time until burn is finished.

This invention discloses a video tracking and recognition method for the spatial trajectory of hoisted objects based on multi-source data fusion, belonging to the field of machine vision monitoring technology. The method includes acquiring pre-loaded physical boundary constants of the tower crane and simultaneously collecting multi-source data during tower crane operation. Based on the physical boundary constants, kinematic calculations are performed on the multi-source data to obtain the relative vertical displacement, the three-dimensional absolute spatial coordinates of the luffing trolley center point, and the absolute coordinates and three-dimensional rotation matrix of the camera's optical center in the global three-dimensional coordinate system. Edge detection is performed on the temporally aligned color video frame matrix to extract candidate wire rope contours and generate a binary edge pixel matrix. Collinear edges are mapped to polar coordinate space using Hough transform. This invention reduces the uncertainty of monocular vision depth scale and improves stability and accuracy under conditions of large-amplitude lifting and complex swinging.

A double-light-source pixel matrix lamp capable of being spliced is characterized in that the double-light-source pixel matrix lamp comprises a plurality of first light sources which are arranged on a first lamp panel and are evenly distributed in an array mode, and a plurality of second light sources which are arranged on a second lamp panel and are evenly distributed in an array mode; the first lamp panel and the second lamp panel are distributed in parallel at intervals in the vertical direction, and the first lamp panel is located under the second lamp panel. The number of the second light sources is four times that of the first light sources, and the first light sources are located on the central axis of a square defined by the four first light sources closest to the first light sources.

A kind of power equipment image processing method based on optical fiber bundle imaging, storage medium and system, method is: signal output by infrared fiber bundle is collected by infrared camera pixel matrix, unit in matrix is divided into effective unit and invalid unit, and generate unit division binary matrix data file;Again processing obtains temperature correction matrix data file, temperature segmented point corresponding response value matrix data file, and establish the correction value of response value matrix under different radiation scene;Through numerical fitting and interpolation compensation method, the video stream of optical fiber bundle transmission image is carried out real-time restoration and optimization, completes the inversion from thermal map to temperature field distribution;The present application also includes the storage medium and system of storing the power equipment image processing method based on optical fiber bundle imaging of running this, the present application carries out real-time restoration and optimization to the video stream of optical fiber bundle transmission image by numerical fitting and interpolation compensation method, completes the inversion from thermal map to temperature field distribution.

This invention belongs to the field of laser ultrasound imaging technology, specifically disclosing a laser ultrasound imaging method based on neural networks, comprising the following steps performed sequentially: S1, acquiring the A-scan signal; S2, decomposing and reconstructing the A-scan signal through empirical mode decomposition; S3, extracting the temporal features of the ultrasound signal using pooling; S4, establishing a neural network and training it multiple times to obtain multiple neural networks with different representation capabilities; S5, inputting the temporal features of the ultrasound signal of the sample to be tested into each trained neural network, and filling the recognition results into the corresponding positions according to the detection point order to form a pixel matrix; S6, superimposing all pixel matrices, setting a threshold to evaluate the state of each detection point, and generating the corresponding pixel image. The method of this invention is insensitive to noise and can be used to process low signal-to-noise ratio signals, improving the accuracy of material state identification. This invention is applicable to detecting the state of materials.