98 results about "Visual contrast" patented technology

The invention discloses a phase residual error-based shallow defect detection method and device and a storage medium, which are used for improving the detection precision of shallow defects. Obtaining a reflection fringe pattern of the display screen to be detected; performing phase recovery on the reflection fringe pattern to generate first phase recovery data; calculating and generating a first phase gradient amplitude diagram according to the first phase recovery data; constructing a local average background gradient at each pixel according to the first phase gradient magnitude image; generating a first gradient residual image according to the first local average background gradient image and the first phase gradient magnitude image; performing visual contrast enhancement processing on the first gradient residual image; self-adaptive threshold judgment is carried out on the first anomaly enhancement graph, and a first structure defect mask graph is generated; performing feature fusion on the first phase gradient amplitude image, the first anomaly enhancement image and the first structure defect mask image; and inputting the first feature fusion image into a target shallow layer defect identification model for shallow layer defect detection, and generating a shallow layer defect detection result.

The invention discloses a few-sample classification method research based on feature cache optimization, and the method comprises the steps: (1), employing an improved ResNet-based SwAV model as a backbone, and enabling the SwAV model to be used for generating auxiliary features; (2) taking the image data and the text description as original input, and generating text input with rich downstream language semantics by utilizing GPT-3 to serve as text prompt of a CLIP model; (3) through a feature selection method based on feature similarity and difference, the problems of feature redundancy and inaccurate selection in the feature selection process are solved, feature dimensions with high selection value are identified, and normalization and enhancement zooming processing are performed on the features; and (4) utilizing visual contrast knowledge of SwAV, introducing a learnable cache model, and adaptively mixing prediction results from CLIP and SwAV. According to the method, under the CLIP framework without extra training, the accuracy of few-sample image classification is effectively improved through feature cache optimization and a self-adaptive hybrid strategy.

The invention discloses a traffic video question and answer method based on inter-frame difference, and the method comprises the steps: innovatively constructing an inter-frame difference extraction module on the basis of an image pre-training model based on the local abrupt change characteristic of a traffic video: effectively capturing the abrupt change characteristic of a moving target through the pixel-level difference calculation of adjacent frames; and designing a gating feature fusion mechanism, dynamically aligning spatial distribution of differential features and visual features on the basis of fusing problem features, and realizing collaborative characterization of static scenes and dynamic changes. In order to enhance cross-modal feature alignment, a hierarchical comparison learning strategy is provided: differential-visual feature comparison is implemented at a bottom layer to enhance local dynamic perception, and text-visual comparison is performed at a high layer to establish global semantic association. Experimental results show that the classification performance is improved on the basis of the SUTD-TrafficQA reference data set.

The invention relates to the technical field of intelligent control, and discloses an ultra-low power consumption miniature LED visible light intensity self-adaptive regulation and control method and system, and the method comprises the steps: obtaining an analog light intensity signal, and carrying out the smoothing processing of the analog light intensity signal, and obtaining an environment light intensity value; calculating a light intensity change rate, and if the light intensity change rate exceeds a dynamic threshold value, marking an environment state and outputting a light intensity difference value; generating a preliminary adjustment instruction according to the state matching control parameter and the difference value quantization mapping, and calculating target light intensity in combination with the basic brightness; after executing the instruction, calculating actual and target deviation, and iteratively correcting to obtain a light intensity control signal; if the power consumption exceeds the limit, optimizing the signal amplitude; the input driving circuit analyzes the synthesized pulse current, excites pixels to emit light and gathers to obtain adjustment data; and if the visual contrast is lower than a threshold value, the change rate is adjusted again to realize closed-loop refining, and the light intensity adjustment accuracy is improved. The method can improve the real-time sensing capability of ambient light.

The invention discloses an aspheric keratoprosthesis lens column, a keratoprosthesis and an application method of the aspheric keratoprosthesis lens column, the whole lens column is cylindrical, the front surface of the lens column is a convex surface, the rear surface of the lens column is a concave surface, the rear surface of the lens column is a surface close to the retina, the front surface of the lens column is a surface far away from the retina, and the front surface and the rear surface of the lens column are both curved surfaces. And the spherical surfaces are rotationally symmetrical aspheric surfaces. The artificial cornea comprises a lens column and a support, and the lens column is installed on the support. The fixed position of the lens column is that the distance between the vertex of the rear surface of the lens column and the retina is 10-120mm, the focal plane is perpendicular to the optical axis, and the inclination angle of the focal plane is 0 degree. According to the lens column, spherical aberration and coma can be effectively corrected, and the imaging quality is improved; light scattering is reduced, the visual contrast at night and in a low-light environment is improved, and the visual experience of a patient is improved; the deviation tolerance of the implantation position is higher, and the operation difficulty and the risk of postoperative complications are reduced.

The invention relates to the technical field of image updating, in particular to a multi-modal image real-time updating method and system for temporal bone surgery. The method comprises the following steps: acquiring a real-time multi-modal image, performing pixel-by-pixel frequency domain reconstruction optimization, and constructing a frequency spectrum enhanced fusion image; performing reverse geometric transformation compensation on the spectrum enhancement fusion image to obtain a space alignment optimization image; performing real-time visual contrast enhancement on the space alignment optimization image, and constructing a visual enhancement image; carrying out inter-frame difference calculation on the vision enhanced image, carrying out real-time increment updating optimization, and constructing an increment updating image sequence; and performing multi-level cache rendering management and parallel execution based on the incremental updating image sequence. The temporal bone surgery safety and efficiency are improved through real-time and efficient image updating.

The invention discloses a character interaction detection method based on multi-granularity context aggregation and occlusion reconstruction, and the method comprises the steps: extracting multi-scale features through a Swin Transform, and carrying out the modeling of long-range dependence through a deformable encoder; the features pass through a texture-local collaborative enhancement module, and in combination with texture covariance enhancement and a gating local attention mechanism, discrimination is improved, and background interference is suppressed. The decoder generates an instance query and bounding box prediction. For the occlusion problem, a space adapter is introduced to extract multi-granularity features, and the semantics of an occluded area is restored through random erasure and residual gating in combination with an occlusion reconstruction module. In the interactive reasoning stage, fusing reconstructed features and query to construct interactive global embedding, and analyzing a character-action-object relationship; and meanwhile, CLIP text semantic embedding is introduced as classification initialization, and semantic alignment is optimized by combining visual contrast loss. According to the method, the robustness of feature representation and shielding is enhanced, and the detection precision and semantic consistency in a complex scene are improved.

The embodiment of the invention relates to a page component processing method and device, equipment and a storage medium, and the method comprises the steps: after a playing operation for a first video is received, if the brightness change intensity indicated by the brightness information of the first video is determined to be greater than preset change intensity, according to the brightness value of each video frame in the first video, determining that the brightness change intensity is greater than the preset change intensity; and determining the background transparency value of the target page component on the corresponding video frame, and controlling the visual contrast between the target page component and the video frame. And displaying a target page component on a video picture of the video frame according to the background transparency value. According to the method and the device, the transparency value of the background color displayed on the video frame by the target page component can be determined according to the brightness value of the video frame of the video with relatively high brightness change intensity, so that the visual contrast of the target page component and the video picture is controlled; the information displayed on the target page component is efficiently transmitted to a user in a scene that the user plays a video.

The invention provides an automatic wafer defect detection method and system based on visual driving, and relates to the technical field of image data processing, and the method comprises the following steps: processing a butt joint piece for defect detection by importing design shape data of a wafer; the butt joint piece is assembled on a splicing assembly mechanism to be spliced with the edge of the wafer, and a spliced piece is obtained; driving a macro vision lens to collect a spliced image set of the splicing piece; and calling a defect identification model to carry out defect feature identification on the spliced image set, returning a defect detection result, and carrying out wafer sorting treatment according to the defect detection result after the splicing assembly mechanism separates the spliced pieces. The technical problem that in the prior art, due to the fact that the wafer edge defects are small in size and low in visual contrast ratio, slight edge defects such as micro cracks and micro notches are difficult to effectively recognize is solved, and the technical effect of achieving high-sensitivity and high-precision wafer defect automatic detection and recognition is achieved.

A synthetic image device comprises an array of line focusing elements and an image layer (20) having linearly extending cells The image layer is positioned at a distance relative to the array of line focusing elements such that a limited part of the image layer becomes visible when the image layer is viewed through the array of line focusing elements. Each cell comprises at least two sub-cells, provided side-by-side in a transverse direction (T). The sub-cells of the image layer have individual visual-contrast creating items in patterns (21) forming sub-images. Composed images, each associated with a set composed by one sub-image from each cell, are visible from different viewing angles in the transverse direction. Within overlap areas (42E-G), an average pattern density of the visual-contrast creating items in at least two sub-cells giving rise to the overlapping composed images is less than a complete coverage.

The invention discloses a storage battery based on a two-dimensional code traceability function. The storage battery comprises a storage battery body, an exclusive two-dimensional code identifier, a cloud database, a special management APP and an intelligent polishing tool, the exclusive two-dimensional code identifier is printed through acid and alkali resistant UV ink, covers the polycarbonate transparent wear-resistant coating and is permanently fixed on the battery cover of the storage battery body, and the color of the two-dimensional code and the battery cover form obvious visual contrast; the exclusive two-dimensional code identifier is associated with a unique identity identifier, and the unique identity identifier and storage battery full life cycle data are bound and stored in a cloud database. According to the method, the data reading efficiency is improved, the full-life-cycle data association rate is improved, manual one-by-one retrieval is not needed, complete data of the battery from production to operation and maintenance can be directly obtained through code scanning, errors of data recording and querying are reduced, meanwhile, physical coating and encryption dual protection is adopted for the two-dimensional code, the tampering rate is decreased to 0, and the anti-fake capacity is remarkably improved.

The invention relates to the technical field of machining, in particular to a machining method of a support bearing, which comprises the following steps: after finish turning of the support bearing, sequentially carrying out rough grinding of an inner hole, accurate grinding of an outer circle and an end surface, accurate grinding of a large end surface by adopting a surface grinding machine, final accurate grinding of the inner hole and milling of an oil wedge surface. According to the method, the traditional process route of'reference in advance 'is changed, clamping scratches before final accurate grinding of the inner hole are effectively avoided, and the stability of the roughness of the large end face is guaranteed through surface contact machining of a surface grinding machine; and meanwhile, temporary coloring treatment is conducted on the large end face obtained after accurate grinding, a visual contrast layer easy to identify is formed, the problems of inner hole scratches, end face roughness fluctuation and machining reflection interference are effectively solved, and the machining quality, consistency and production efficiency of the supporting bearing are improved.

A synthetic image device (1) comprises an array (10) of line focusing elements (11), an image layer (20) having linear extending cells (22), and a coloured background layer assembly (25). A limited part of the image layer is visible through the array of line focusing elements. Each cell comprises at least three sub-cells (24A-C). The at least three sub-cells having individual visual- contrast creating patterns (21), forming sub-images. Composed images, associated with a set composed by one sub-image from each cell, are visible from different viewing angles. The composed images together present a movement effect. The coloured background layer assembly covers a surface facing away from the image layer. The background layer assembly has a first colouring (28A), visible through the image layer. The first colouring presents a colour varying over the extension of the background layer assembly in registry with movement effects. A producing method is also disclosed.

The dating system for molds comprises at least one date stamp (12) connected to a knob (4) through a connection cable (1), wherein the date stamp (12) comprises an engraving formed only by numbers and / or letters and a position indicator. Furthermore, said knob (4) is mounted on a control block (2) comprising a reading engraving (3) located around said knob (4), in correspondence with the engraving of said date stamp (12). This makes it possible to manufacture date stamps with the information, without relief, since the simple contrast of the areas of the date stamp polished with high gloss and the nuanced numbers or letters, offer a visual contrast that allows clearly identifying the information that the date stamp must transmit.

The present disclosure provides a face plate for securing onto an electrical receptacle box, an electrical device recognition safety system, and a method of making a wall mounted electrical device safer and more easily identified. The face plate can include a plate body having an opening and a plurality of zones. The plurality of zone can surround the opening. The plurality of zones can include a first zone and the second zone. The first zone can have a visual contrast with the second zone.

A method of providing a dark viewing mode with enhanced visual contrast includes evaluating the luminance and chroma of background colors, including highlight colors, of digital content prior to converting the digital content for the dark viewing mode to identify highlight colors having a luminance value that exceeds a luminance value threshold and a chroma value that exceeds a chroma value threshold. When a highlight color has luminance and chroma values that exceed the respective threshold values, the luminance value of the highlight color is clamped to a predetermined lower luminance value before converting the value for the dark viewing mode. The luminance value is converted for dark viewing mode by taking the opposite luminance value. As a result, the luminance value of the converted highlight color has a higher luminance value and greater luminance contrast with respect to the dark background in dark viewing mode.

The invention relates to the technical field of intelligent manufacturing, in particular to an intelligent manufacturing unit-oriented integrated industrial robot control system, which comprises an integrated control center, an initial interaction module, a task scheduling module, a dynamic collaboration module, a collaboration presentation module and an early warning response module, according to the method, identification, matching and testing of robot communication protocols are automatically completed, the equipment access time is remarkably shortened, meanwhile, it is ensured that all robots can stably complete data interaction, system access dead angles caused by protocol incompatibility are avoided, candidate equipment is screened based on production task requirements and the real-time states of the robots, and the production efficiency is improved. Meanwhile, the task execution error probability is reduced through multi-dimensional comprehensive evaluation and in combination with a load capacity deep screening robot, in addition, in the material grabbing link, center deviation is judged through visual comparison, re-grabbing or abnormal early warning is automatically triggered, production stagnation caused by material grabbing errors is reduced, along with reasonable path planning, the production efficiency is improved, and the production cost is reduced. And the robot driving path safety is ensured.

The utility model discloses an electric vehicle hub motor appearance detection device based on a computer vision system, which comprises a conveying mechanism, a carrier, a camera module, a light source module and a computer system, the carrier is connected to the conveying mechanism and can move along the surface of the conveying mechanism, the carrier is used for placing and fixing a hub motor to be detected, and the camera module is connected to the camera module. A rack is fixedly connected above the conveying mechanism, the camera module, the light source module and the computer system are all connected to the rack, the camera module is used for shooting and identifying a hub motor, the light source module is used for illuminating the hub motor, and the computer system is connected with and controls the camera module and the light source module. The camera module can automatically detect and judge the problems of omission, deformation and the like of laser etching characters of products by using an OCR technology, and judges whether the problems of omission of oil seal installation, reverse installation of oil seal, missed installation of screws, valve hole dislocation after assembly of a hub and a side cover and the like exist or not by using a visual comparison technology, so that omission and misjudgment caused by manual inspection are completely eradicated.

The invention relates to the technical field of packaging bag production and detection, and discloses a packaging bag defect visual inspection system and a method thereof. The visual shooting unit is used for shooting and collecting the target graph appearance of a specific area, the area formed by the #-shaped structure is used for being compared with the area of a packaging bag, the edge comparison module is used for visually comparing the boundary dimension deviation of the packaging bag and adjusting the detection area, and the conveying module is used for conveying the packaging bag to the detection area. The conveying belt of the conveying module is used for providing supporting conditions during packaging bag detection. According to the packaging bag thickness detection device, the broken and leaked area of the defective packaging bag is used, the contact roller is in direct rolling contact with the packaging bag, the supporting laser sensor is used for measuring and calculating generated height difference data, if the height difference data exceed the thickness size of the packaging bag, the packaging bag is judged as a defective product, and therefore uniform detection of the thickness of the packaging bag is achieved; and bubbles or uneven surfaces are avoided after heat sealing.

This invention relates to the field of welding inspection technology, and specifically to a method and system for detecting welding defects. The method includes: acquiring a depth map of a workpiece weld; performing adaptive pseudo-color encoding on the depth map to obtain a pseudo-color image, wherein a lookup table is constructed based on the cumulative distribution function of the depth map, and depth values ​​are mapped to color values ​​through the lookup table; calculating pixel-level normal vectors based on the depth map, and performing illumination rendering on the pseudo-color image according to the pixel-level normal vectors to obtain a rendered image; inputting the rendered image into a defect detection network, and outputting defect detection results. This invention, by constructing a lookup table based on the cumulative distribution function of the depth map for adaptive pseudo-color encoding, can automatically adjust the color mapping relationship according to the actual depth distribution of each depth map, making different depth values ​​visually present a more uniform and obvious color difference, effectively improving the visual contrast and feature representation of the depth map.

The present application relates to the field of intelligent control and photoelectric detection of aviation obstruction light, in particular to an aviation obstruction light adaptive dimming system based on wireless networking; the system comprises an environment sensing module, a feature coupling module, a control solving module and an adaptive driving module; the ambient light intensity value, the background environment color temperature value and the space scattering evaluation value are collected, the multi-dimensional light feature coupling compensation coefficient is solved, the background environment light comprehensive interference value is corrected and generated, and the target light source effective light intensity and its ratio are set as the dynamic ambient light signal-to-noise ratio; then the preset visual compensation model is input, the visual saliency maintenance index is solved and the dimming control instruction is generated, the driving current and the pulse duty cycle are adjusted, so that the visual contrast of the target light source is maintained within the preset safety interval, the present application can ensure the long-distance warning and recognition ability, suppress the near-distance glare and halo aggravation, and take into account the warning effect, thermal stability and device life.

The invention discloses a galaxy form classification method and system based on asymmetric cross-modal momentum contrast learning, and belongs to the technical field of image classification, and the system comprises a visual encoder which extracts image features and stores the image features in a momentum feature queue; the text encoder is used for encoding the text information to generate initial text embedding; the dynamic text embedding projection layer is used for embedding the initial text into the projection to obtain a text projection; the classification head is used for predicting a galaxy form classification result; the processing module is used for calculating total loss including visual contrast loss and cross-modal alignment loss, and updating model parameters of the visual encoder, the dynamic text embedding projection layer and the classification head by using the total loss; for the total loss, the weighted weight of the visual contrast loss is greater than the cross-modal alignment loss. According to the method, asymmetric interaction of images and texts can be realized, text information is effectively fused in a visual dominant galaxy classification task, the classification precision is improved, and resource consumption is optimized.

The invention belongs to the technical field of gas meter detection devices, and particularly relates to a device special for gas meter sealing performance detection, which comprises a box body, a gas meter to be detected is arranged in the box body, an L-shaped plate is fixed on one side of the box body, a cylinder is fixed at the top of the L-shaped plate, a U-shaped plate is fixed at the output end of the cylinder, and a groove is formed in the top of the U-shaped plate. By arranging the swing mechanism, the gas meter periodically swings in a reciprocating mode in water, so that the posture of the gas meter is changed, bubbles of all parts have the opportunity to be exposed, the situation that the bubbles are blocked is reduced, the sealing condition of the gas meter can be comprehensively and accurately detected, and the sealing performance of the gas meter is improved. Meanwhile, the motion trail of the bubbles in the water is more obvious through swinging, stronger visual contrast is formed between the bubbles and the surrounding water body, observation and recognition are facilitated, and compared with bubbles in static water, the bubbles in the motion process are easier to observe.

The application discloses a kind of cross-scale, multi-mode, visual biological detection device and detection method, belong to biological detection technical field, including: X-ray fluorescence imaging module, mechanical spectroscopy adhesion monitoring module, signal integration processing module.X-ray fluorescence imaging module is used for the acquisition, optimization, coding and identification of biological sample blue, green, red and near-infrared color pattern, mechanical spectroscopy adhesion monitoring module is used for real-time, dynamic monitoring and recording to biological sample active adhesion, through the collaborative processing analysis of X-ray optical signal and mechanical spectroscopy signal of the biological sample to be measured, realize the fluorescence visual of biological sample morphology, structure component and the accurate acquisition of biological activity data, for the research and application of cross-scale, multi-mode, visual contrast agent of mechanical spectroscopy and X-ray excited fluorescence imaging provides equipment support, promote its development optimization and conversion application in biomedical field such as inflammation infection, malignant disease, tumor diagnosis.

The utility model relates to a fogging out-of-focus composite lens and a pair of frame glasses with the fogging out-of-focus composite lens, and belongs to the field of out-of-focus lenses. A fogging out-of-focus composite lens comprises: a lens substrate; a plurality of out-of-focus micro lens groups, wherein the plurality of out-of-focus micro lens groups are sequentially nested on the surface of one side of the lens substrate from inside to outside; the plurality of groups of annular fogging defocus areas are arranged on the surface of the other side of the lens base body, and the orthographic projection of each group of annular fogging defocus areas is located between the two connected groups of defocus micro lens groups. The structure of the lens is optimized, and thousands of light scattering micro points exist from the edge to the center of the lens. And the visual contrast of each distance is reduced. The fogging out-of-focus technology effectively reduces'extension 'signals sent by the eyes to the brain, so that the growth of the ocular axis is delayed. Due to the design of scattered light, when the user uses the eyes for a long time, the visual pressure is relieved, the eyes feel more comfortable, and common visual fatigue and dry feeling are reduced.

The invention provides an aim point display method and device, electronic equipment and a storage medium. The method comprises the steps that an aim point graph is displayed on a display picture of an application program in an overlapped mode; acquiring a background image of an area where the aim point graph is located in real time, and extracting color features of the background image; determining a target color according to the color features and a preset color processing strategy, wherein the target color is used for forming visual contrast between the aim point graph and the background image; and updating and displaying the aim point graph by using the target color. According to the scheme, the color of the aim point can be adaptively adjusted along with the change of the background in real time, so that the aim point can be kept clear and distinguishable by forming strong visual contrast with the background in any complex game scene, and the aiming accuracy and the game experience are remarkably improved.

The invention discloses a nickel-cobalt ion extraction visual contrast detection method and an extract liquid detection device, the detection method comprises the following steps: a sampling pump pumps an extract liquid in an extraction tank into a sampling pipeline, a concentration sensor of the sampling pipeline obtains the total concentration of the extracted extract liquid, and the concentration information is transmitted to a terminal; the visual detection camera adjusts the exposure time according to the total concentration of the extracted extraction liquid, photographs and samples the extracted extraction liquid in the sampling pipeline, and then transmits a photographed and sampled picture to the terminal; the color distinguishing module performs color value analysis and decomposition on the RGB color of the extracted extraction liquid picture; and comparing the RGB value of the picture of the extracted extraction liquid with the RGB value of the standard sample of the nickel-cobalt mixed solution under different concentrations to obtain the concentration value of the nickel-cobalt ions in the extracted extraction liquid. According to the detection method provided by the invention, the extraction degree can be accurately and timely fed back, the continuity of the production process is ensured, and the extraction efficiency is improved.

A totally transparent glass jar, intended for containing foodstuffs, with a substantially cylindrical shape; the midsection is characterized by a smooth, continuous and cylindrical surface; the surface delineates a visually neutral area, which enhances, by contrast, the three-dimensional elements located both in the upper and in the lower parts; the underside is characterized by a decorative peripheral band, which consists of four horizontal annular ribs, that extend all around the entire circumference, circularly; a further peripheral decorative band is located symmetrically, at the top, just below the jar neck; this band consists of four horizontal annular ribs, which are identical in number, proportions and arrangement to those of the lower band; all the ribs – both in the upper and in the lower parts - have a purely aesthetic function; the jar neck is narrower than the body and has oblique helical ribs; the open part of the jar is circular and has a regular rim; the individual character of the design arises from: the symmetrical repetition of the decorative bands; the visual contrast between the bands and the smooth cylindrical midsection; and the proportional balance between the body, the decorative bands and the neck; all the reflections and shadows caused by the lights, shown in all reproductions of design No. 1, are disclaimed from the application for registration of design; the No. “22” and the wording “720 ml” shown on the bottom view, are also disclaimed from the application for registration of design; fig. No. 1.1 is a front view of the product; fig. No. 1.2 is a perspective view of the product; fig. No. 1.3 is a further perspective view of the product; fig. No. 1.4 is a bottom view of the product; fig. No. 1.5 is an enlarged view that shows the decorative peripheral band in the underside, which consists of four horizontal annular ribs that extend all around the entire circumference, circularly; fig. No. 1.6 is a top view of the product, which shows its open upper part; fig. No. 1.7 is an enlarged view that shows the decorative band just below the jar neck, which consists of four horizontal annular ribs, identical in number, proportions and arrangement to those of the lower band.

This invention provides a sensitive text characterization method based on contrastive learning, belonging to the field of text characterization technology. The method includes constructing positive and hard negative sample pairs; establishing a multi-task contrastive learning model; training the multi-task contrastive learning model; and obtaining a sensitive text characterization result. This invention solves the problem of inefficient contrastive training caused by existing contrastive learning models that use discrete data augmentation to obtain distorted semantic samples and directly use visual contrast frameworks for contrastive learning in the natural language domain.

The present invention discloses a method for tracking a target for autonomous aerial refueling of a UAV based on a falcon-like visual twin network: 1. Multi-scale feature extraction based on the contrast sensitivity mechanism and adaptive focusing mechanism of the falcon-like vision; 2. Similarity calculation and feature fusion based on the bifoveal visual interaction mechanism of the falcon-like vision; 3. Classification and regression loss function design based on the perceptual competition and decision-making mechanism of the superior colliculus pathway of the falcon-like vision; 4. Training the falcon-like visual twin tracking network and outputting classification loss and regression loss; 5. Testing and evaluating the falcon-like visual twin tracking network. The present invention can adapt to various complex environments and changes in the size of the UAV's appearance; can efficiently fuse multi-scale spatial features and focus on the most significant areas; can maximize tracking stability and improve the success rate and safety of autonomous aerial refueling of the UAV; the training strategy is adapted to the complex autonomous aerial refueling environment of the UAV; the feature extraction and fusion are efficient, robust and generalizable, and require less training data.