292 results about "Image field" patented technology

The invention relates to the technical field of data processing, and provides a regional soil composition detection method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a hyperspectral image of a to-be-detected region; acquiring LIBS spectral data of spatial point locations in the to-be-detected area based on a laser-induced breakdown spectroscopy (LIBS) technology; based on the spatial position information, matching the hyperspectral image with the LIBS spectral data, and performing spatial point location soil component labeling on the hyperspectral image based on a matching result to obtain a labeled hyperspectral image; and detecting the soil components of the to-be-detected area based on the labeled hyperspectral image to obtain a soil component detection result. According to the regional soil component detection method and device, the electronic equipment and the storage medium, low-cost and high-efficiency automatic labeling of the hyperspectral image is realized, and the method and the device have important significance in promoting development of the field of remote sensing image field large-area environment detection.

The invention belongs to the field of image anomaly detection, and particularly relates to a zero sample anomaly detection method based on prompt learning tuning, which comprises the following steps: S1, acquiring to-be-detected image data, and preprocessing the to-be-detected image data; s2, constructing a zero sample anomaly detection model which comprises a text encoder, an image encoder, a prompt learning module, a multi-scale feature extraction module and the like; s3, performing fine adjustment on the zero sample anomaly detection model by using the typical data set of the to-be-detected image field; and S4, detecting an image to be detected by using the fine-tuned zero sample anomaly detection model, and describing anomaly information by using a basic large model to obtain anomaly classification, anomaly positioning and anomaly description results. Compared with a traditional zero sample anomaly detection model, the method not only constructs more comprehensive and generalized adaptive prompts, but also realizes pixel-level accurate positioning of anomaly regions of different sizes and shapes, and finally fills the blank of an existing method in the aspect of anomaly description, thereby providing richer judgment basis for zero sample image anomaly detection.

The invention discloses a sound authenticity identification method based on multi-modal feature deep interactive fusion. According to the method, a double-flow architecture is adopted, and a pre-trained BEATs model and a CNN14 network are respectively utilized to extract Transform sequence features and convolution time-frequency embedding features of an audio; an adaptive MobileFormer fusion device is innovatively proposed, an original MobileFormer structure used in the image field is transformed into a bidirectional cross-modal interaction module suitable for one-dimensional time sequence audio features, dynamic complementary modeling of local details and global semantic features is achieved through a cross attention mechanism, and dynamic nonlinearity is introduced to activate and enhance the expression ability; and the fused enhanced features are input into a hierarchical graph attention network ASSIST, and high-order semantic modeling and authenticity classification are completed by combining spectrogram and time sequence double-flow reasoning. Experimental results show that the performance of the method on an ASVspoof2021LA data set is superior to that of an existing baseline model. The method can effectively detect AI generation voice, replay attack and other forged voice, and is suitable for a voice authentication system.

The invention discloses a multi-feature fusion method for sustainable learning, and the method comprises the steps: determining a plurality of target training pictures in a to-be-learned target image field in a plurality of image fields, and determining a plurality of pieces of target frequency domain feature sub-data of each target training picture; establishing a cross loss function of the image processing model according to a feature fusion process of the plurality of frequency domain feature sub-data in the image processing model; and training the image processing model by taking minimization of the cross loss function as a training target so as to obtain an image processing model which learns a target image field. According to the method, through the synergistic effect of frequency domain feature extraction, double loss function construction and continuous optimization training, the dynamic adaptive capacity and recognition precision of the image processing model in the multi-modal fusion task are improved, and the stability and expansibility of the model are enhanced.

An imaging EUV optical unit serves for imaging an object field into an image field. The optical unit has a plurality of mirrors for guiding EUV imaging light at a wavelength shorter than 30 nm along an imaging beam path from the object field to the image field. The plurality of the mirrors includes at least two normal incidence mirrors and at least two grazing incidence mirrors. An overall transmission of the plurality of the mirrors is greater than 10%. This yields an imaging EUV optical unit whose usability for an EUV projection exposure apparatus can be improved.

An imaging optic (10) for projection lithography serves to image an object field (5), in which a reticulum to be imaged can be arranged, onto an image field (11), in which a substrate to be exposed can be arranged. The imaging optic (10) has a plurality of mirrors (M1 to M8) for guiding imaging light (16) along an imaging beam path from the object field (5) to the image field (11). The image-side numerical aperture of the imaging optic (10) is greater than 0.75. The optical conductivity of the imaging optic (10) is greater than 45 mm. 2 The result is an imaging optic whose usability for a projection exposure system is improved.

The utility model discloses a shooting device for remote sensing imaging, and relates to the field of remote sensing imaging. The shooting device for remote sensing imaging comprises an unmanned aerial vehicle, a support unit and a shooting unit, the support unit and the shooting unit are arranged below the unmanned aerial vehicle, the support unit comprises a circular sliding rail, a sliding ring and supporting legs, the sliding ring is arranged below the circular sliding rail in a sliding mode, and the supporting legs are fixedly installed below the sliding ring; the shooting unit comprises a driving motor, a connecting frame and a shooting camera, the driving motor is fixedly installed at the bottom of the unmanned aerial vehicle, the connecting frame is connected with the output end of the driving motor and connected with the supporting legs, and the shooting camera is connected with the connecting frame. According to the shooting device for remote sensing imaging, the driving motor fixedly installed at the bottom of the unmanned aerial vehicle drives the connecting frame, the shooting camera and the supporting legs to rotate, so that only the unmanned aerial vehicle needs to hover in the air, when the shooting camera rotates by a shooting angle, the supporting legs do not shield the unmanned aerial vehicle, and the shooting stability and the imaging quality are improved.

An imaging optics for projection lithography has a plurality of mirrors (M8, M9, M10) for imaging an object field in an object plane into an image field (8) in an image plane (9) with imaging light (3) guided along a beam path between the object field and the image field. A penultimate mirror (M9) in the beam path has no passage opening for the imaging light (3). An aperture (19) is arranged in the beam path between the penultimate mirror (M9) and a last mirror (M10) in the beam path. The aperture (19) defines, at least in sections, an edge contour of a pupil of the imaging optics (7) in the beam path between the penultimate mirror (M9) and the last mirror (M10). The aperture (19) is fixed to a component (M10) of the imaging optics via a fastening element (20) which runs at least in sections along the beam path between the penultimate mirror (M9) and the last mirror (M10).In another embodiment of the imaging optics, the aperture is part of a shell component of the imaging optics surrounding the beam path, or is fixed to a shell component of the imaging optics surrounding the beam path, or to a frame component of the imaging optics, or to the second-to-last mirror in the beam path, or to the last mirror in the beam path. In each case, the resulting imaging optics has a well-defined pupil and is optimized for projection lithography.

The invention relates to the technical field of agricultural remote sensing classification, and provides an intelligent field extraction and fusion method for high-resolution satellite images. The method comprises the following steps: step 1, collecting high-resolution satellite images of different areas and preprocessing the high-resolution satellite images to obtain a high-resolution satellite image data set; 2, constructing a high-resolution image field intelligent extraction and fusion network, and training the high-resolution image field intelligent extraction and fusion network by using the high-resolution satellite image data set to obtain a high-resolution image field intelligent extraction and fusion model; wherein the high-resolution image field intelligent extraction fusion network comprises a cultivated land information intelligent extraction module, a cultivated land boundary multi-task deep learning module and an improved mean shift multi-scale segmentation module. According to the method, layered extraction of cultivated land information extraction, cultivated land block boundary determination and crop field block segmentation is carried out in the cultivated land fragmentation area by fusing multiple remote sensing methods, and the refined extraction requirement of precision agriculture can be met.

The invention relates to a multi-modal data construction method based on VLM-LLM cooperative verification and instruction driving, and the method achieves the full-process automation through a VLM and LLM cooperative verification design mechanism, reduces the manual labeling link, and enables the construction efficiency of complex scene data to be improved. Furthermore, through a multi-crossing rechecking mechanism of LLM, the self-error correction capability is established, the error labeling rate is remarkably reduced, and the later correction cost is greatly reduced; by automatically scheduling the expert identity in the LLM, the excellent domain migration capability is realized, the problem of artificial expert dependence is effectively solved, the method can automatically adapt to the new image field, manual intervention is not needed, and the flexibility and adaptability of multi-modal data construction are improved. Compared with a traditional manual labeling scheme and a semi-automatic labeling scheme, the technical disadvantages of insufficient semantic understanding, modal splitting, poor dynamic adaptability and the like are overcome, the manpower disadvantages of high cost, large quality fluctuation, low iteration efficiency and the like are overcome, and a solution is provided for efficiently producing a high-quality multi-modal data set.

The invention relates to a visual scale adaptive unmanned aerial vehicle detection method based on HRRP physical perception, and belongs to the technical field of mobile communication. The method comprises the following steps: constructing a cooperative monitoring architecture of an ISAC base station and a visual sensor, and establishing a space-time mapping relation between a radio frequency sensing coordinate system and a visual sensor coordinate system; performing macroscopic parameter estimation based on echo signals received by the ISAC base station, resolving spatial position coordinates of a target, generating a servo control instruction to drive a visual sensor holder to rotate, and locking the target at the center of a visual image field of view; extracting a frequency domain channel response of the echo signal, generating a time domain HRRP through IDFT, and estimating a radial physical span of a target in combination with a constant false alarm detection algorithm; constructing a self-adaptive image slice model based on physical structure constraint, and performing resampling on slices by using a scale normalization scaling factor; and inputting the normalized image block into the feature fusion detection network to complete fine detection of the unmanned aerial vehicle target.

A method for operating a projection exposure system (1) comprises the following steps: providing a projection exposure system (1) with a radiation source (1) for generating illumination radiation (3), an illumination optic (11) for transferring the illumination radiation (3) from the radiation source (2) to an object field (8) extending in a scan direction (y) and a cross-scan direction (x), wherein the illumination optic (11) has at least one faceted mirror (6, 7) with a plurality of individual mirrors, and wherein at least a subset of the individual mirrors is configured such that they only partially illuminate the object field (8) in the scan direction (y), and a projection optic (10) for imaging a reticle (12) arranged in the object field (8) of the illumination optic (11) onto a wafer (19) arranged in an image field (17) of the projection optic (10), and specifying a set of at least one element of field-dependent aberrations. (Ai(x, y);i=(1..M)), specification of a target vector (Z(x) = (Z1(x), ..., Z; M (x) T ) for the target values ​​Z i (x) of the aberrations Ai(x, y) after averaging over the scan direction (y), dividing the extent of the object field (8) in the scan direction (y) into j ≥ 2 intervals ([y0; y1], ..., [y j-1 ; y j ]) and determining scan weights (gi(x, y)) which characterize a dependence of an illuminance at a location x in the cross-scan direction in the object field () on a position in the scan direction (y) such that a deviation of a weighted aberration vector A weighted ( x ) → : = ∑ gi A l → is reduced from the target vector (Z(x)).

In a projection exposure method for exposing a substrate arranged in the region of an image plane of a projection lens with at least one image of a pattern arranged in the region of an object plane of the projection lens, using radiation from a wavelength range around a design wavelength < 260 nm, a substrate (SUB) is coated with a relatively thick radiation-sensitive photoresist layer (RS) and exposed using a projection lens (PO). This lens produces a focus at a design focus position (FOCO) at the design wavelength and, for other wavelengths outside the wavelength range, offset focus positions from an axially extended focus area (ΔFOC) around the design focus position (FOCO). The projection lens is designed as a single-waisted or double-waisted system.During an exposure time interval, different wavelengths from the wavelength range around the design wavelength are used such that the axial extent of the focus area is at least as large as the layer thickness. The projection lens is optically corrected for chromatic aberration (CHV) such that, in the region of the design wavelength, the image scale is essentially independent of the wavelength, so that a change in wavelength of 1 picometer results in a maximum variation of 2 nm between the position of an image point corresponding to an object point within an image field.

An apparatus for producing incisions in the interior of an eye. In this case, the apparatus contains an image recording device embodied to record at least part of the image field and an image evaluation device embodied to evaluate recordings of the image recording device and produce signals for the control device and / or the operator. Furthermore, a method for producing incisions in the interior of an eye, wherein an image recording device is used to record at least part of the image field and an image evaluation device evaluates the recordings of the image recording device and produces signals for the control device and / or the operator.

The invention discloses an image display method, a construction system and a computing control unit of a head-mounted display, which are applied to the field of images. The first to-be-displayed image is corrected through the pre-correction network model, the visual quality of the image displayed by the first display module is improved, and the defects of geometric distortion, spherical aberration, field curvature and the like of the image seen by a user are reduced, so that the visual fatigue of the user is reduced, and the immersion of VR experience is improved. Moreover, in the process that the user uses the head-mounted display, the training data is collected online in real time, and the pre-correction network model is trained based on the collected training data, so that the training process of the pre-correction network model is simplified, and the training efficiency of the pre-correction network model is improved. And the training data in various actual use scenes can be continuously adopted in an online mode, so that the types of the training data for training the pre-correction network model are enriched, and the accuracy of the pre-correction network model is ensured.

The application discloses a kind of based on multi-modal medical image's lumbar disc herniation postoperative recurrence prediction system, belong to medical image field.Use lumbar image set and each image corresponding feature true value and each patient corresponding recurrence prediction label to carry out training to based on multi-modal image's lumbar disc herniation postoperative recurrence prediction model, in training process, using random gradient descent algorithm to update model parameter, obtain the trained based on multi-modal image's lumbar disc herniation postoperative recurrence prediction model.Through the application, benefit from the rich image features provided by multi-modal image, the prediction network trained based on multi-modal data shows higher accuracy in lumbar disc herniation postoperative recurrence prediction.

According to the medical rotary cutting tool bit assembly, the rotary cutting catheter and the visual blood vessel volume reduction system, the rotary cutting tool bit assembly is provided with a front rotary cutting tool bit and a rear rotary cutting tool bit to conduct rotary cutting operation on blocked blood vessel plaques or thrombus and the like, and the effects of opening blocked blood vessels and enlarging the diameter of a lumen are achieved; a larger blood vessel lumen diameter can be obtained under the condition that surgical instruments do not need to be replaced or repeated rotary cutting is not needed, and the surgical operation steps and time are reduced; meanwhile, by means of an imaging assembly, especially an ultrasonic transmitting and receiving device, instrument positioning in the blood vessel is carried out, a real-time and accurate ultrasonic image view is provided, and the lesion condition can be accurately evaluated; under the scene that a blood vessel guide wire or other surgical instruments cannot effectively penetrate through the occluded lesion, a lesion area can be smoothly reached, and the occluded blood vessel can be opened, so that the operation is safer and more effective.

A magnifying imaging optic (20c) is part of a metrology system for examining objects. The magnifying imaging optic has a maximum of four mirrors (M1 to M4) that image an object field (4) in an object plane (17) into an image field (21) in an image plane (22). Reflecting surfaces of small-area mirrors (M2, M3), which are used to guide imaging light (3) along an imaging beam path, deviate from a spherical shape by a maximum of 10 µm. The result is a magnifying imaging optic that, for a given manufacturing effort, produces an imaging result that meets the strict requirements of a metrology system.

An endoscope and an endoscope camera system, the endoscope includes a mirror tube, an illumination light path and an imaging light path, a light homogenizer is installed in a branch tube, an optical fiber is installed in the mirror tube, the incident end of the optical fiber extends to the branch tube and is connected to the output end of the light homogenizer, and the output end of the optical fiber extends to the front end of the mirror tube. Since the endoscope is equipped with a light homogenizer at the front end of the optical fiber, the light homogenizer can efficiently couple the light beam of the light guide into the optical fiber, and the light homogenizer can also eliminate the light spot in the light beam and improve the uniformity of the light beam; the optical fiber is an ultra-wide-angle optical fiber, and the divergence angle of the optical fiber is greater than the imaging field angle. The ultra-wide-angle optical fiber can effectively reduce the light energy loss of the illumination light path, improve the light transmission efficiency, and provide more sufficient illumination for ultra-high-definition images. This ultra-high-definition endoscope achieves the goal of increasing the light flux while making the illumination distribution uniform in the observation field of view through the cooperation of the light homogenizer and the ultra-wide-angle optical fiber, thereby improving the contrast and signal-to-noise ratio of the endoscopic image, and then improving the imaging quality.

The present invention relates to the technical field of particle detection, and discloses an ultra-fast static image method particle detection system and method, comprising a main control module, a light source, a platform, an optical path system, an image acquisition module and a data processing module. A filter membrane with a sample is provided on the platform. When detecting particles of the sample, an image acquisition area is divided, the image depth of field of a single acquisition area is fused, image detection and jigsaw puzzle are performed in parallel, and ultra-fast data acquisition and analysis are performed, so that the time of the entire test process is greatly shortened, and the detection result is clear and accurate.

The invention discloses an event camera depth estimation method based on unsupervised domain adaptation, and is suitable for the technical field of event camera depth estimation in the field of computer vision. According to the method, the relation between the two modes is enhanced by comprehensively analyzing the difference between the image domain and the event domain and changing the image style. According to the method, the features in the implicit space can be more easily aligned, and the gap in the explicit space is reduced. Training is carried out in the marked image field, and then knowledge is transferred to unmarked event data; according to the method, knowledge is transferred from an image to an event by utilizing cross-domain feature alignment and depth alignment, style transfer is carried out by utilizing appearance, and the difference between different fields is minimized by utilizing edge invariant consistency. Through unsupervised domain adaptation, event monocular depth estimation is carried out only by using a depth label of an image.

An imaging optics for projection lithography has a plurality of mirrors (M9, M10) for imaging an object field in an object plane into an image field in an image plane with imaging light (3) guided along a beam path between these fields. A penultimate mirror (M9) in the beam path has no aperture for the imaging light (3). In the beam path (3 M9M10 ) between the second to last mirror (M9) and the last mirror (M10) in the beam path, an aperture (19) is arranged. This defines the outer edge contour of a pupil of the imaging optics in the beam path (3 M9M10 ) between the penultimate mirror (M9) and the last mirror (M10) at least in a pupil edge contour section, which is simultaneously in the beam path (3 M10W) between the last mirror in the beam path (M10) and the image field. The result is an imaging optic with a well-defined pupil and optimized for projection lithography.

An imaging EUV optic (10) has a plurality of mirrors (M1 to M6) for guiding imaging light (16) along an imaging beam path from an object field (5) to an image field (11). The imaging optics (10) have a total apodization tilt APO Z3 that is less than 10%. A beam path intersection region (K1, K2, K3) lies between a first partial beam path (M1M2, M2M3) and a second partial beam path (M3M4, M4M5) in the imaging beam path. The first partial beam path (M1M2, M2M3) lies between a mirror (M1, M2) and a mirror (M2, M3) directly following this mirror in the imaging beam path. The second partial beam path (M3M4, M4M5) lies between another mirror (M3M4) and a mirror (M4M5) directly following this further mirror in the imaging beam path. An object-image offset (d) OIS) is greater than 250 mm. This results in an imaging EUV optic whose usability for an EUV projection exposure system is improved.

An optical system for a projection exposure system comprises an illumination optic for guiding illumination and imaging light (3) from a light source to an object field in which an object to be imaged can be arranged, and an imaging optic for imaging the object field into an image field in which a substrate to be exposed can be arranged. The imaging optic has an exit pupil (36) with at least one obscuration exit pupil area (38) located within an outer pupil margin (37). The illumination optic is designed such that an illumination angle distribution of the object field with the illumination and imaging light (3) results, which can be described by an illumination pupil (12a) whose illumination pupil area (I) exposed to illumination light is adapted to the obscuration exit pupil area (38) of the imaging optic.The obscuration exit pupil area (38) of the imaging optics can have a non-central geometric center of gravity (GC). In one variant of the optical system, the illumination optics are designed such that an illumination angle distribution results in the object field which can be described via an illumination pupil (12a) with an obscuration illumination pupil area (31) whose envelope deviates from a circular shape in pupil coordinates (x, y) normalized to the outer pupil margin (37) and / or in absolute pupil coordinates (x, y). To specify the illumination pupil (12a), a diffraction behavior of at least one object structure of the object to be imaged is analyzed, a position of the obscuration exit pupil area (38) in the exit pupil (36) is determined, and the illumination pupil (12a) is adjusted so that at least one of the variants of the optical system results.This results in an optical system with improved resolution.

An OCT system for the retina (4) comprises: an illumination unit (36) providing SS-OCT source radiation, an illumination beam path leading to the object (4) and a measurement beam path coming from the object (4) to a detector (38) having pixels (74), wherein the measurement beam path is coupled out via a splitter element (42), a filter element (48) following the splitter element (42) which makes the numerical aperture of the image larger in a first direction (x) than in a second direction (y) lying transversely thereto, and a coupling device (52, 66) for coupling in the reference radiation (R) in the pupil and outside the optical axis (6), so that the reference radiation (R) is superimposed on the detector (38) with the measurement radiation (M) at an angle, wherein the control device (58) reads out pixels (74) of the detector (38) which correspond to a line-shaped area in the image field (22) correspond,which has a line length along the first direction (x) and a line height along the second direction (y), and the control device (58) realizes an image field (22) that is higher than the line height by moving the line-shaped region over the object (2) along the line height, wherein a depth resolution of the generated 3D image is determined by the spectral range and spectral bandwidth of the tuning, and a lateral resolution of the 3D image is determined by the numerical aperture in the line direction and by the line height.

An imaging optic (23) serves to image an object field arranged in an object plane, in which a lithography mask can be arranged as the object to be imaged, with imaging light via an imaging beam path into an image field arranged in an image plane, in which a substrate onto which the image is to be formed can be arranged. The imaging optic (23) has at least one correction device (32). This device has a correction element (30) arranged in the imaging beam path, the imaging effect of which varies across the image field. Furthermore, the correction device (32) has a correction element holder (33) and a correction element drive unit (34). The correction element holder (33) is driven by the correction element drive unit (34) during an imaging phase of the imaging optic (23) and can be displaced along a correction element displacement direction (35) extending transversely to the imaging beam path.The result is an imaging optic whose imaging effect is improved to achieve a predetermined imaging result.

The invention relates to a microscope with an imaging beam path comprising an objective (4) and imaging a sample (2) from an image field onto a two-dimensional detector (8), and an illumination beam path (B) illuminating the sample (2) through the objective (4), which illumination beam path comprises a light source and is reflected into the imaging beam path (M) via a beam splitter (6) in front of the objective (4), and a measuring device (10) followed by a light modulator (16) comprising at least one intensity-modulating, first line grating structure (26-29), wherein the light modulator (16) is arranged obliquely to a plane conjugate to the focal plane of the objective (4) and is projected onto the sample (2) through the illumination beam path (B) and the objective (4), wherein the at least one intensity-modulating, first line grating structure (26-29) is projected into a measuring area of ​​the sample (2) that only partially fills the image field,and the measuring device (10) has a processor (32) which reads out image data supplied by the detector (8) and divides these into measurement data and display data for displaying a sample image based on a position in the image field, wherein the processor (32) selects as measurement data those image data which correspond to the measurement area of ​​the sample (2) and evaluates the measurement data with regard to reflection images of the at least one line grating structure (26-29) projected onto the sample (2).

The present application discloses a method, system, device and storage medium for providing image collection suggestions, belonging to the field of artificial intelligence technology. The method includes obtaining an image field combination required to make the target business meet the information completeness requirements; according to the image field combination, the image type set available for collection, and the image fields contained in the image corresponding to the image type in the image type set, the image type of each image that respectively has at least one image field in the image field combination is determined from the image type set; based on the determined image type, the type of image recommended for collection is determined. Starting from meeting the information completeness requirements of the target business, an image field combination is obtained, and from the image type set available for collection, the image type of each image containing the image field combination is determined, so as to provide image collection suggestions, which can objectively and quickly provide image type collection suggestions that meet the current business processing needs.

An arrangement for detecting a possible deviation of an agricultural vehicle (10) from a target path (74) when working on crops grown in rows (58), a corresponding vehicle (10) and method comprise an imaging sensor system (78) whose field of view includes the wheels (18, 20) of the vehicle and simultaneously the rows of crops (58) adjacent to the wheels (18, 20).

The invention discloses an image contrast clustering method based on sample layering. Aiming at the problem of false positive samples in a contrast clustering process used in the image field at the present stage, the image contrast clustering method is improved from two aspects: on one hand, positive sample pairs are divided into different levels, namely prototype-level positive sample pairs, neighborhood-level positive sample pairs and enhancement-level positive sample pairs, and on the other hand, the positive sample pairs are divided into different levels, namely, the prototype-level positive sample pairs, the neighborhood-level positive sample pairs and the enhancement-level positive sample pairs; the positive sample pairs of different hierarchies are subjected to contrast clustering in different forms, the clustering mode accurately utilizes information of the positive sample pairs of different types, and the problem of false positive samples is relieved; on the other hand, classifier information is introduced into the positive sample pairs of the prototype samples and the positive sample pairs of the neighborhood samples to construct pseudo labels, more attention is paid to the positive sample pairs with the same label information, and the pseudo positive sample problem caused by similar different types of samples is relieved. The key process of the method is as follows: data preprocessing and enhancement, construction of an image comparison representation learning model, pre-training, construction of an image comparison clustering model, training of the image comparison clustering model, and testing of the image comparison clustering model. The invention provides an image clustering method by improving a contrast clustering model, and the method can accurately cluster different types of image data under the condition of no label.