591 results about "Aerial image" patented technology

The invention relates to the technical field of computer vision and deep learning, in particular to an aerial image target detection method based on frequency domain decoupling multi-scale feature fusion, which comprises the following steps of: acquiring an aerial image of an unmanned aerial vehicle, establishing a data set, and performing preprocessing and data division; an aerial image target detection network is constructed, and the aerial image target detection network receives an input image and outputs a target category and a bounding box position; loss functions are determined, wherein the loss functions comprise classification loss representing matching quality, coordinate loss representing prediction coordinate relevancy and bounding box regression loss representing bounding box positioning accuracy; training the aerial image target detection network based on the data set and the loss function; inputting a to-be-detected aerial image into the trained aerial image target detection network to obtain a to-be-detected target category and a bounding box position; the method can improve the feature fusion degree, retains high-frequency details, and enhances the small target recognition rate.

The invention discloses a target detection method based on a YOLO model, electronic equipment and a storage medium, and relates to the technical field of target detection. Comprising the following steps: inputting an aerial image of an unmanned aerial vehicle into a trained target YOLO model; the target YOLO model comprises a backbone network, a neck network and a head network, and a feature extraction module in the backbone network performs multi-scale feature extraction by adopting a double-branch architecture attention mechanism; performing multi-scale feature extraction on the aerial image by adopting a dual-branch architecture attention mechanism through a feature extraction module in the backbone network, and constructing to obtain a plurality of layers of first comprehensive image features of the aerial image; performing feature fusion on the first comprehensive image features of different levels through a neck network to obtain second comprehensive image features of multiple levels; and inputting the multiple levels of second comprehensive image features into a head network to obtain a target detection result of the aerial image. According to the invention, the accuracy of small target detection can be improved.

The invention relates to the technical field of flight path planning, in particular to an autonomous flight path planning method and system for a road slope inspection unmanned aerial vehicle, and the method comprises the following steps: controlling the unmanned aerial vehicle to fly in a road slope region according to a preset flight path, collecting the aerial image data of the unmanned aerial vehicle, analyzing the aerial inspection adaptive feature value of the unmanned aerial vehicle, and determining the flight path of the unmanned aerial vehicle. Judging whether the feature value is lower than a preset inspection adaptation feature threshold value or not; analyzing a field-of-view missing feature set of the unmanned aerial vehicle based on the aerial image data of the unmanned aerial vehicle if the field-of-view missing feature set is lower than the preset routing inspection adaptive feature value threshold; performing flight path planning processing on the unmanned aerial vehicle based on the field-of-view missing feature set; according to the method, the unmanned aerial vehicle is subjected to flight path planning processing on the basis of the field-of-view missing feature set, so that when an aerial picture is deviated, fuzzy or blind areas, judgment and automatic adjustment are performed in real time, self-adaptive optimization of the flight path is realized, and the road slope inspection efficiency is remarkably improved.

The invention relates to a robust unmanned aerial vehicle detection method based on dynamic feature fusion and context attention, and belongs to the technical field of image processing. Aiming at the problems of small target feature loss, semantic gap, background noise interference and the like caused by a fixed convolution kernel scale, one-way feature fusion and a static attention mechanism in an existing unmanned aerial vehicle aerial image target detection method, the method comprises the following steps: constructing a detection model comprising a backbone network, a neck network and a detection head network; a feature rearrangement and extraction module is designed in the backbone network to enhance feature learning, an enhanced double-flow feature fusion pyramid is designed in the neck network to optimize multi-scale feature fusion, and a dynamic multi-scale context attention mechanism is designed in the detection head network to suppress irrelevant background noise. The method effectively improves the accuracy and robustness of small target detection, and achieves a clearer and more stable detection effect in a complex environment.

The invention relates to a computer vision and target detection technology, in particular to an enhanced target detection method and device based on feature fusion and a medium. The method comprises the following steps: acquiring aerial image data; extracting a multi-level initial feature map through an initial feature extraction network; respectively extracting local features and global features of the initial feature map of each level through a local feature extraction network and a global feature extraction network which are deployed in parallel; self-adaptively fusing the local features and the global features through a gating fusion module to generate a fused feature map, and calculating a feature competition map and generating a spatial dimension gating weight map by adopting a local-global feature competition mechanism oriented to an aerial photography scene to realize feature weighted fusion of spatial positions one by one; and performing multi-scale fusion on the multi-level fusion feature map, and finally outputting a target detection result. According to the invention, the detection precision and robustness of the multi-scale target in the aerial image are effectively improved.

The invention relates to the technical field of topographic change analysis, in particular to an unmanned aerial vehicle-based topographic change detection system, which comprises a slope direction sensing track control module, a texture structure extraction module, a crack evolution track construction module, a direction trend comparison module and a patrol recheck positioning module. According to the method, a continuous elevation point column of an unmanned aerial vehicle scanning area is extracted, laser reflection point coordinates are fused, a space relation of transition point distribution is constructed, dynamic adjustment of a ground-imitated flight path is achieved, and a texture structure area with continuous directivity is recognized in combination with a high-angle image boundary communication relation; texture boundary evolution is compared at different time nodes to form a crack path, the stability of the path and the slope direction is judged through an included angle sequence, recognition and sorting of areas with the consistent direction are completed, a space comparison result is registered in a three-dimensional coordinate system, terrain change areas are accurately marked, and rapid positioning and continuous tracking of high-risk areas are achieved.

A method for detecting printing defects in a photolithography mask that will print on a wafer when using the photolithography mask in a specific photolithography system to print semiconductor structures on the wafer, the method comprising: acquiring a first aerial image of the photolithography mask using a mask inspection system; generating a second aerial image of the photolithography mask by applying a machine learning model (26) to the first aerial image, wherein the machine learning model is trained to map a first aerial image acquired by a mask inspection system to a second aerial image that emulates the application of the specific photolithography system to the photolithography mask; and detecting printing defects in the photolithography mask by comparing the second aerial image to a reference image.

The invention provides a ship target detection method based on nonlinear network enhancement, a medium and equipment, and belongs to the field of artificial intelligence. The method comprises the following steps of: performing type conversion, data division and enhancement on image data in a data set by adopting the ship data set acquired in a real river channel scene; a nonlinear network enhanced YOLO ship detection model is constructed; adopting a multi-task joint loss function to train the YOLO ship detection model; and inputting a port monitoring image and a sea surface aerial image into the trained YOLO ship detection model, outputting a category number, a confidence value and bounding box coordinates of each prediction box, and forming a visual image. According to the method, a nonlinear network enhanced YOLO ship detection model is constructed, and the modeling expression capability of the network on a ship target in a complex scene is effectively improved, so that the robustness and the accuracy of target detection are enhanced, and particularly, the performance is better under the conditions of small target detection and image degradation.

The invention discloses a three-dimensional reconstruction method and device based on an aerial image. The method comprises the following steps: collecting a plurality of aerial images of a to-be-reconstructed region and corresponding position data and attitude data; generating an initial three-dimensional point cloud according to the plurality of aerial images and the corresponding position data and attitude data; and converting the initial three-dimensional point cloud into a Gaussian original body, carrying out iteration on the Gaussian original body to obtain a Gaussian sputtering three-dimensional model, and generating a three-dimensional reconstruction result by using the Gaussian sputtering three-dimensional model. According to the scheme, through combination of traditional point cloud reconstruction and Gaussian sputtering modeling, efficient conversion from the aerial image to the high-precision three-dimensional model is realized. The method not only improves the reconstruction precision of the three-dimensional model, but also has high stability and calculation efficiency when processing complex terrains and large-scale scenes.

A method and system for constructing a three-dimensional (3D) aerial survey of a city street scene include obtaining a plurality of video frames from a calibrated multi-camera setup covering a 360-degree view mounted on a moving vehicle. The plurality of video frames is split into a plurality of 3D parts containing a subset of the plurality of video frames and preprocessing the subset of the plurality of video frames of each part of the plurality of parts to obtain a calculated information. Further, constructing, by the processing circuitry, a 3D representation of each part of the plurality of parts based on the calculated information to obtain a plurality of local 3D reconstructed scene intervals. The method includes stitching and filtering, by the processing circuitry, the plurality of local 3D reconstructed scene intervals to construct the 3D city street scene.

The invention relates to the technical field of image target detection, and discloses an unmanned aerial vehicle aerial image small target detection model construction method comprising the following steps: preparing an aerial image data set, preprocessing the aerial image data set, and generating an aerial image sample set; the method comprises the following steps: establishing a basic model on the basis of a YOLOv8 network model, removing a P5 detection layer in a head network Head of the basic model, introducing a P2 detection layer, replacing a specified position of a Conv module in a backbone network Backbone by adopting an ACMConv feature enhancement module, replacing a specified position of a C2f module in a neck network Neck by adopting a C2fMixStructure mixed structure module, and constructing an improved model by adopting a lightweight enhanced detection head structure; and dividing the aerial image sample set into a training set, a verification set and a test set in proportion, training and verifying the improved model, and generating an unmanned aerial vehicle aerial image small target detection model based on AMLP-YOLOv8. According to the invention, the method has higher perception capability when extracting fine target features, and improves the stability and robustness of small target detection of the aerial image of the unmanned aerial vehicle.

The invention discloses a smart city planning three-dimensional scene reconstruction optimization system combined with semantic segmentation, and belongs to the technical field of three-dimensional scene reconstruction optimization. The system comprises a data sensing module which collects aerial images and laser point clouds, and complements a sheltered area to obtain city modeling data; the preprocessing module performs denoising, data registration and data fusion on the city modeling data to generate texture point cloud data; the semantic understanding module realizes multi-modal semantic segmentation of the texture point cloud data through a fine-tuned SAM network and an improved RandLA-Net, and semantic tag data is obtained through restoration and optimization; the reconstruction optimization module constructs an initial three-dimensional grid model based on the texture point cloud data, and optimizes a ground feature boundary and a missing region of the model in combination with a semantic tag; and the output application module converts the optimized model into a standard format and outputs the model. Through deep coupling of semantic segmentation and reconstruction optimization, the precision of the three-dimensional scene model is improved, and reliable support is provided for smart city planning.

The invention discloses an unmanned aerial vehicle small target detection method based on feature enhancement and selective attention and a related device, and relates to the technical field of computer image target detection, and the method comprises the steps: obtaining an aerial image of an unmanned aerial vehicle, adjusting the image to a preset resolution, and obtaining an adjusted image; inputting the image into a backbone network of a network architecture, extracting a multi-level feature representation through a plurality of CALBlock feature extraction modules, and enhancing features of the multi-level feature representation through an EMIT edge enhancement architecture to obtain an enhanced multi-level feature representation; inputting the enhanced multi-level feature representation into a neck network of the network architecture, and performing cross-scale feature fusion and enhancement through an FSAFPN structure to obtain a multi-scale enhanced feature map; and inputting the multi-scale enhanced feature maps into a detection network of a network architecture, and performing target detection by using a decoder to obtain target category probability distribution and bounding box coordinates.

A georegistration (a.k.a. georectification) of an image captured by a camera in an aerial vehicle, such as a satellite, is based on identifying multiple features using descriptor sets, and sending to a ground station only the descriptors of the identified features and the associated locations in the captured image, without sending of the captured image itself, thus requiring a low communication bandwidth. Using a database of geosynchronized reference images, the ground station uses the received descriptors sets and the associated image locations to localize the features on a selected geosynchronized reference image from the database, and forms a mapping function that map any locations in the captured image to geographical coordinates on Earth. The mapping may be used to geosynchronize an additional feature identified in the aerial vehicle, or to geo synchronize a region that may be cropped from the captured image and sent to the ground station.

The invention discloses an unmanned aerial vehicle aerial image imaging optimization method and device fusing a deep learning perception mechanism and physical modeling. The method comprises the following steps: acquiring an original image frame obtained in a flight process of an unmanned aerial vehicle; inputting the image into a MobileViT illumination estimation network, extracting local convolution perception and multi-scale global semantic features, and outputting a scene illumination intensity estimation value; constructing a differentiable imaging parameter reasoning module based on an illumination physical modeling relationship, reversely deducing an optimal exposure parameter combination of a current frame, and constructing a parameter optimization module based on a perceptual error; combining the difference between the reconstructed image and the target image in the semantic perception space to construct a multi-loss function joint training model, and optimizing an exposure combination; deploying an edge computing platform for the trained network model to complete parameter prediction, control feedback and image acquisition link closed loop; according to the method, exposure optimization is realized before imaging, image gamma decoding and target enhancement are realized after imaging, and the image quality in low-light and backlight scenes is improved.

The invention discloses an improved RT-DETR-based small target detection method and system under a visual angle of an unmanned aerial vehicle, and aims to solve the problems of low detection precision and the like caused by complex background and large scale change in small target detection in an aerial image of the unmanned aerial vehicle. A backbone network of the RT-DETR network is improved to enhance the multi-scale feature representation capability and the space detail retention capability of the RT-DETR network; in order to obtain a deep feature map with richer semantic information, a new adaptive image feature integration module is introduced, and finally, a new feature fusion network is designed, so that feature maps with different resolutions can be better subjected to bidirectional fusion, a final feature map is generated, and the target detection precision is improved.

The invention discloses a road facility disease evolution prediction and maintenance decision-making method and system based on a CNN-Transform hybrid architecture. The method comprises the following steps: acquiring a multi-period aerial image, constructing a time sequence data set, and carrying out spatial alignment and simulation degradation processing; a hybrid network model of an encoder-decoder structure is constructed, an encoder adopts a dual-channel structure, local features are extracted through a convolution block, global semantic features are extracted based on an attention mechanism and integrated with an edge enhancement component, and a decoder adopts a bidirectional propagation unit and fuses the features through a cross-scale attention mechanism; carrying out cooperative training and compression on the model to obtain a simplified prediction model; predicting a facility state by utilizing the model, constructing a multi-criterion optimization target, and solving by adopting a genetic algorithm to obtain a non-dominated solution set; and finally, generating a maintenance decision report according to the decision preference and visually displaying the maintenance decision report. According to the invention, accurate prediction and scientific maintenance decision making of road facility disease evolution are realized.

The invention belongs to the field of computer vision, and provides an aerial photography small target lightweight detection method fusing RepVGG and GhostConv. The method comprises the steps that a re-parameterized convolution module RepVGG is introduced into a detection model, feature representation is enhanced through multi-branch convolution in a training stage, and multi-branch structures are combined into single convolution in an inference stage, so that efficient inference is achieved; a lightweight convolution module GhostConv is adopted to reduce model parameters and calculation amount, and multi-scale feature extraction and fusion are carried out in combination with an attention mechanism; feature extraction, network structure optimization and training are carried out on aerial images, and finally model training and small target detection are completed on an optimized aerial data set. According to the method, the detection speed can be remarkably improved while the detection precision is ensured, light-weight, efficient and high-precision small target detection of the model is realized, and the method is particularly suitable for small target rapid identification in an aerial photography scene.

The invention provides a multi-category small target dangerous area intrusion detection method based on unmanned aerial vehicle aerial images, which adopts a large-scale unmanned aerial vehicle visual target detection data set to carry out training, and comprises the following steps: firstly, providing a dynamic channel attention convolution (DyCACConv) for the problem of low small target detection precision caused by weak key feature capture, and solving the problem of low small target detection precision caused by weak key feature capture; key features can be adaptively focused, so that feature expression is enhanced, and small target detection precision is improved; secondly, aiming at the problems that the network multi-scale target detection capability is insufficient and multi-class tasks are difficult to process, an improved feature fusion module (BiFPNConcat2) is provided, the feature importance is adjusted through a learnable weight, and the multi-class target detection capability is improved; and finally, aiming at the problems of insufficient scale feature capture and weak long-distance dependence perception of the existing method, a multi-scale dilated convolution attention mechanism (DilateBlock) is provided, and dilated convolution and self-attention are combined, so that the scale features can be efficiently captured and the long-distance dependence perception can be enhanced.

Methods, systems and apparatus, including computer programs encoded on computer storage media for unmanned aerial vehicle flight operations near physical structures or objects. In particular, a point cloud of the physical structure is generated using aerial images of the structure. The point cloud is then referenced to determine a flight path for the UAV to follow around the physical structure, determine whether a planned flight path to desired locations around the structure is possible, determine the fastest route to return home and land from a given position around the physical structure, determine possibility of inflight collision to surface represented in point cloud, or determine an orientation of a fixed or gimballed camera given a position of the UAV relative the point cloud.

The invention relates to the technical field of computer vision and target detection, and discloses a small target detection network for adaptive fine-grained feature mining. The invention provides a network, and the network overcomes the contradiction between the calculation efficiency and the detection precision of a traditional method through collaborative design of adaptive fine-grained feature mining and RoI feature interaction. The network can detect a high-resolution image by using difficult areas in a conventional-level feature map and a high-resolution shallow-layer feature map at the same time, the difficult areas with dense information are automatically positioned through a foreground probability discriminator, background redundancy calculation is avoided by using an iterative mining strategy, and the small target detection speed is increased; meanwhile, the Inter-RoI feature interaction module realizes bidirectional complementary enhancement of deep semantics and shallow details in a key difficult area, and in combination with a high-resolution detection head and a result fusion mechanism, the feature characterization capability of a small target can be enhanced under a complex background, and finally, the feature characterization capability of the small target can be enhanced while the high reasoning efficiency is kept. And the detection precision and robustness of small targets which are non-uniformly distributed and have weak features in the aerial image are improved.

The invention relates to the technical field of computer vision, and discloses a hybrid supervision collaborative learning-based power transmission line identification method under a small sample condition, and the method comprises the steps: obtaining an aerial image of an unmanned plane, and carrying out the preprocessing and marking of the image, and obtaining a data set; a self-supervised learning framework of a self-adaptive mask enhancement strategy is proposed, linear geometric priori knowledge is obtained through a lightweight U-Net segmentation model, a linear element self-adaptive mask mechanism driven by weak supervised learning is constructed, and a masked image is generated; a scale-width-angle collaborative enhancement multi-dimensional line feature attention module is provided, the multi-dimensional line feature attention module is fused into a sparse convolution encoder to extract visible region features, and the perception and feature extraction capability of the model on the slender structure of the power transmission line is enhanced; a masked area is reconstructed through a decoder, a composite loss function optimization strategy with weight collaboration is provided, a segmentation task under the condition of foreground and background extreme imbalance is met, the overall convergence speed of the model is increased, and power transmission line recognition precision rise under the condition of small samples is achieved.

The invention belongs to the technical field of unmanned aerial vehicle autonomous navigation, and discloses an unmanned aerial vehicle visual positioning navigation method based on a deep twin network and multi-modal fusion, and the method comprises the steps: obtaining a real-time aerial image of an unmanned aerial vehicle and a pre-stored satellite reference image, carrying out the self-adaptive scale zooming of a satellite image according to the output height of a barometer, and obtaining the real-time aerial image of the unmanned aerial vehicle; eliminating scale difference; the images after scale unification are input into a deep twin network for feature extraction and matching, and two-dimensional plane transformation parameters between the two images are obtained; and calculating the parameters as initial absolute geographic coordinates, adopting a tight coupling factor graph optimization frame, fusing a re-projection factor formed by the visual coordinates, an IMU pre-integration factor, a barometer height factor and a magnetometer course factor, and estimating and outputting six-degree-of-freedom pose information of the unmanned aerial vehicle through nonlinear optimization. According to the invention, the problems of poor matching robustness of different-source images and incomplete pose output in a GNSS denial environment are solved, and high-precision autonomous positioning navigation is realized.

The invention relates to the technical field of image recognition, and discloses a grassland degradation recognition method and system based on aerial images, and the method comprises the following steps: S101, obtaining a three-band reflectivity image; s102, extracting a red edge main direction and red edge anisotropy intensity; s103, performing screening to obtain a main response band-pass texture map; step S104, calculating to obtain a red edge texture coherence ratio index; step S105, marking the types of the pixels; and step S106, obtaining the area and proportion of each type of region. Structural features such as red edge main direction and anisotropic strength of a red edge form comparison graph are extracted, adaptive texture features are screened in combination with multi-scale filtering and main response scale, the optimal displacement vector is calculated through phase correlation to achieve space alignment, pseudo-coherent interference such as splicing seams and micro-displacement is effectively inhibited in cooperation with weighted covariance, and the method is suitable for large-scale and large-scale imaging. Therefore, grassland degradation identification precision is improved.

The invention discloses a lightweight real-time target detection method for aerial images of an unmanned aerial vehicle. The lightweight real-time target detection method comprises the steps of constructing a mixed attention module (MAM), designing a shape interchange-to-sum ratio loss function (SI-IoU), constructing an LART-DETR model overall architecture, processing an input layer, optimizing backbone network features, processing encoder features, detecting a decoder and a detection head target and optimizing model training. Comprising the following steps. According to the lightweight real-time target detection method for the aerial image of the unmanned aerial vehicle, in terms of lightweight and real-time performance, through MAM and architecture optimization, the model parameter quantity is only 13.51 M (reduced by 32% compared with RT-DETR and reduced by 17.9%-47.7% compared with YOLO series), the calculated quantity is 44.1 GFLOPs (reduced by 22.6% compared with RT-DETR), the single-batch reasoning speed reaches 37 FPS, limited hardware resources of the unmanned aerial vehicle can be adapted, and the real-time detection requirement can be met; in regressive performance and robustness, the problem of imbalance of the width-height ratio of a boundary frame is solved by combining an innovative shape intersection-to-sum ratio loss function (SI-IoU) with L1 loss to form SI Loss, and the regression error of a distortion target is reduced by 20% compared with that of RT-DETR (Reverse Transcription-DETR).

The invention discloses an RDU-YOLO-based lightweight marine small target detection method, device and equipment and a medium, and relates to the technical field of image target detection. The method comprises the following steps: constructing an RDU-YOLO network model, wherein the RDU-YOLO network model comprises a backbone network, a neck network and a detection head; training the RDU-YOLOY network model by using the disclosed aerial photography data set of the marine unmanned aerial vehicle; obtaining a to-be-detected image; and inputting the to-be-detected image into the pre-trained RDU-YOLO network model to obtain the position and category of each target corresponding to the to-be-detected image. By implementing the technical scheme provided by the invention, the problems that the details of the target are difficult to accurately capture and the detection precision is obviously reduced due to the fact that the target is small in size and shows a low-resolution characteristic in an aerial image when a traditional target detection algorithm is used for processing a small target in an open water area are solved.

The invention discloses an image processing method and device for an aerial remote sensing image and a storage medium, and relates to the field of image analysis, and the method comprises the steps: obtaining an aerial image shot by an unmanned plane in real time, and carrying out the image preprocessing of the aerial image, and obtaining a preprocessed aerial image; performing rigid framework extraction on the preprocessed aerial image to obtain an original incomplete framework graph after gravel interference is removed; carrying out fracture framework repair and completion on the original incomplete framework diagram to obtain an unmanned aerial vehicle visual angle topological structure diagram; performing elastic topological mapping matching on the unmanned aerial vehicle visual angle topological structure diagram and a pre-stored satellite reference diagram to obtain a matching result; and carrying out pose calculation according to a matching result, and outputting accurate coordinate information of the unmanned aerial vehicle in the world coordinate system. According to the method and the device, high-precision positioning can be realized in an extreme environment in which the GPS is rejected and the surface texture is subjected to unstructured deformation.

The invention discloses an urban agglomeration anti-seismic damage prediction method and system based on an unmanned aerial vehicle technology and a deep learning model. The method comprises the following steps: constructing a typical building multi-angle aerial image database and a building earthquake damage database under different load coupling effects; training an urban building group identification model based on the aerial photo database to identify the type, geometry and texture information of the building; constructing and training horizontal and vertical seismic oscillation amplification coefficient prediction models for correcting seismic oscillation intensities under different topographic conditions; a typical building earthquake damage prediction model is constructed and trained, and the building damage grade is predicted in combination with the corrected earthquake vibration intensity; and finally, realizing three-dimensional visualization of the damage condition of the urban building group through a damage grade imaging module. According to the method, the problems of inaccurate urban building group modeling information, large earthquake damage analysis workload and low precision in the prior art are solved, and a scientific basis is provided for construction and evaluation of tough cities.

The invention discloses an aerial target detection method based on dynamic attention and double-frequency feature enhancement. The method comprises the following steps: 1) extracting multi-scale features by using a backbone network in combination with a low-light feature fusion module, and enhancing detail expression in a low-light and low-contrast scene; 2) sparse modeling is carried out on deep features through a dynamic sparse attention module, key attention connection is reserved, and the global semantic ability is improved while the calculation amount is reduced; 3) inputting the deep enhanced features and the shallow features into a double-frequency feature enhancement module, and respectively modeling low-frequency background and high-frequency details to realize foreground highlighting and background suppression; according to the unmanned aerial vehicle aerial image target detection method, the problems of low illumination, complex background and small target detection are effectively solved, and the precision and robustness of unmanned aerial vehicle aerial image target detection are remarkably improved.

The invention belongs to the field of computer vision and deep learning, and discloses an aerial image small target detection method based on bidirectional feature fusion and adaptive downsampling, which comprises the following steps: obtaining standardized input data for small target detection training, and adopting a lightweight trunk feature extraction network as a trunk feature extraction network to extract a small target; the method comprises the following steps: carrying out feature extraction on an aerial image in a standardized label format to obtain a multi-scale basic feature map, constructing a bidirectional feature fusion network, carrying out bidirectional fusion on the multi-scale basic feature map with different spatial scales, introducing an ADown module to carry out self-adaptive downsampling to obtain a self-adaptive downsampling feature map, and carrying out self-adaptive downsampling on the self-adaptive downsampling feature map. A multi-scale coding feature map is obtained through the hybrid efficient encoder, the decoder decodes the multi-scale coding feature map, and accurate detection and positioning of a small target are achieved. According to the method, the precision of small target detection is remarkably improved on the premise of keeping high efficiency, and the method is particularly suitable for complex scenes with high real-time requirements such as unmanned aerial vehicle monitoring and intelligent transportation.