2352results about How to "Improve detection accuracy" patented technology

The invention discloses a small target detection method based on a Fourier mixed attention mechanism. An improved small target image detection network model based on RT-DETR is researched and designed, and a Basic Block module in a backbone network is replaced by a self-developed Fourier mixed attention enhancement module (FTABlock). The module is composed of a Fourier transform attention module (FTAModule) and a hybrid dynamic convolution feedforward network (DKMixFFN). The FFTModule calculates the attention weight in the frequency domain through Fourier transform, and strengthens small target feature expression in combination with position coding and multi-head attention; the DKMixFFN adopts a dynamic convolution kernel and a multi-scale dynamic convolution layer to realize adaptive modeling of multi-scale features. According to the method, under the synergistic effect of frequency domain attention and dynamic convolution, the small target feature extraction and perception capability of the RT-DETR model is effectively enhanced, and the small target detection precision is improved.

The invention discloses a tight-lock coupler intelligent tensile testing machine and a testing method thereof, and belongs to the technical field of tight-lock coupler test.The tight-lock coupler intelligent tensile testing machine comprises a mounting bottom plate and an electric control cabinet, and a support for supporting a tight-lock coupler to be detected is mounted on the surface of the mounting bottom plate; a tensile connecting frame for connecting the tight-lock coupler is mounted on the surface of the bottom plate on one side of the support, an integrated design of'tensile force detection and visual detection 'is innovatively adopted, mechanical parameter acquisition and visual deformation observation are combined, traditional mechanical data such as tensile force and displacement are obtained, and visual characteristics such as fine deformation and crack initiation are captured; precise synchronization of pulling force application, displacement collection and camera shooting is achieved through the synchronous controller, the real-time corresponding relation of pulling force-displacement-visual deformation is established, the problems that an existing device is single in detection dimension and poor in data relevance are solved, and the potential quality hazard checking capacity and the persuasion of a detection result are improved.

The invention discloses a calibration method of a polarization focal plane imaging system and an imaging system, relates to the field of photoelectric imaging and remote sensing detection imaging, and solves the problems of low correction efficiency and difficulty in correcting a system with a changed polarization state in the conventional polarization error correction technology. According to the calibration method, a plurality of coding patterns and corresponding polarization modulation matrixes are obtained based on the collected uniform light field and serve as calibration results. The polarization error correction method comprises the steps of obtaining an observation intensity vector based on a target scene; a corresponding polarization modulation matrix is obtained based on the coding pattern according to a coding polarization modulation matrix lookup table, and then a down-sampling Stokes vector after polarization error correction is obtained; and completing the correction of the polarization error through an imaging reconstruction method. A polarization focal plane imaging system comprises an imaging lens, a DMD digital micromirror array coding module, a focal plane DoFP polarization detector and an imaging reconstruction module. The method is suitable for the fields of remote sensing detection, target identification, environment monitoring and the like.

The invention relates to the technical field of aluminum alloy detection, and discloses clamping equipment for aluminum alloy detection, which comprises a base, two electric guide rails are symmetrically mounted on the base, an area for placing an aluminum alloy profile is formed between the two electric guide rails, brackets are mounted on the two electric guide rails, a bearing seat is mounted on each bracket, and elastic rolling pieces for pressing the aluminum alloy profile are mounted on the bearing seats; the electric guide rail can drive the support to drive the elastic rolling piece on the bearing base to roll on the surface of the aluminum alloy profile. And a hardness detection piece is mounted in the elastic rolling piece. Rapid identification of surface unevenness is achieved through the elastic rolling part and the displacement sensor, local hardness can be quantitatively detected through the built-in hardness detection part, internal defects caused by insufficient hardness are prevented, dynamic pressing and friction resistance detection of the profile in the full-length range are achieved in cooperation with the adjustable limiting detection part and the pressure sensor, and the detection precision is improved. The comprehensiveness of pre-screening before flaw detection is improved, and the condition of missing detection is effectively prevented.

The invention provides a ground insulation resistance detection circuit, a detection method and a product, which are applied to a detection scene of ground insulation resistance of a direct current side of a non-isolated high-voltage energy storage converter. According to the ground insulation impedance detection circuit provided by the embodiment of the invention, a complex isolation circuit in an isolation type scheme is not needed, an additional bridge arm resistor in a non-isolation type scheme is also avoided, the detection circuit is constructed through a small number of resistors, switches and voltage acquisition modules, the number of devices and the occupation of PCB space are reduced, the design and manufacturing cost can be effectively reduced, and the detection circuit is suitable for large-scale popularization and application. And the compact layout requirement of the non-isolated high-voltage energy storage converter is met. On the basis, the embodiment of the invention further provides a ground insulation impedance detection method, the voltage acquisition module stably collects the first voltage and the second voltage through on-off of the switch, whether the risk of insulation failure exists or not is judged, the situation of sampling signal distortion in a non-isolation scheme can be reduced, and the detection precision is effectively improved.

The invention relates to the technical field of vehicle trajectory analysis, in particular to a vehicle trajectory anomaly detection method and system based on deep learning, and the method comprises the following steps: obtaining trajectory coordinates and boundary distance to generate symbol offset, constructing a road offset continuous field through interpolation and spatial embedding, calculating the ratio of instantaneous speed to reference speed, and obtaining a vehicle trajectory anomaly detection result. And generating an unbalance degree parameter, executing vector dimension stretching, extracting a trajectory embedding vector and an abnormal score by using a long short-term memory network, executing square summation and difference operation, and generating a trajectory energy gradient value. According to the method, migration characteristics are mapped to a continuous field to achieve track and road geometric constraint association, unbalance degree parameters are used for remarking directional quantity scales to enhance speed dynamic perception, energy gradient modulation scoring is cooperated, a confidence probability curved surface is constructed by using trilinear interpolation, and a curved surface peak value is positioned to output a detection result. Positioning noise is suppressed; and abnormal precision under sparse sampling is improved.

The invention belongs to the technical field of computer vision, and particularly discloses a monocular 3D object detection method for realizing depth enhancement based on a visual basic model, electronic equipment and a readable storage medium, and the method comprises the steps: S1, building a data set: employing a monocular camera to collect a pavement scene, and obtaining an RGB image in the pavement scene; s2, image preprocessing: preprocessing the RGB image for subsequent feature extraction and depth estimation; s3, performing feature extraction by adopting a dual-backbone network: performing visual semantic feature extraction on the preprocessed RGB image by using DINOv2; performing depth feature extraction on the preprocessed RGB image by using a DPT head; s4, generation of depth perception query points: inputting the visual semantic features and the depth features into a DETR network to generate the depth perception query points; and S5, target detection output: using an MLP-based detection head to obtain information of the category, the size, the center point position, the depth, the 3D size and the direction of the object.

The invention discloses a steel structural member local corrosion intelligent identification and evaluation method based on deep learning, and the method comprises the steps: obtaining a steel structure surface image through an imaging device, carrying out the feature extraction and corrosion region identification of an input image through a YOLOv12 semantic segmentation model which introduces a CBAM self-attention mechanism, and generating a pixel-level corrosion mask; the actual physical area of the corrosion area is automatically calculated based on an area calculation algorithm of conversion of mask integration and ground sampling distance; the residual bearing performance of the corroded component is evaluated in combination with the corrosion area ratio and the morphological characteristic prediction model; and finally realizing visual display and report output of detection results, area data and performance indexes through a graphical interface. According to the invention, full-process automatic processing of local corrosion of the steel structural member from detection, segmentation, quantitative measurement and calculation to member performance evaluation is realized, and closed-loop intelligent evaluation from visual detection to mechanical evaluation is realized.

This invention discloses a copper-based ternary chalcogenide semiconductor material, its preparation method, and its applications. The material is prepared by mixing CuSn(OH)6 nanomaterials, thioacetamide, ethylenediaminetetraacetic acid, and water and carrying out a hydrothermal reaction to obtain Cu3SnS4 nanomaterials. This semiconductor material exhibits excellent photoelectrochemical properties. Photoelectrochemical sensors constructed based on this material demonstrate high detection accuracy, fast response speed, low detection limit, and strong stability, enabling rapid and accurate real-time detection of microplastic content. The preparation method is simple, low-cost, and suitable for industrial production.

The invention provides a surface defect detection method for aero-engine blades, and the method comprises the steps: sequentially shooting each blade, and splicing the blades into a panoramic image; due to the fact that the images are partially overlapped, the images can be sequentially spliced into the panoramic image according to the shooting sequence on the basis of the overlapped parts. The panoramic image not only contains enough defect information, but also can fuse a plurality of blades into the same image, so that the problem that the blades are difficult to match with the corresponding blades after the defects are identified is greatly reduced. Inputting the panoramic image into a recognition model added with a space-frequency dynamic convolution module, a Mangban module and a joint attention mechanism; and finally, outputting defect parts of all blades of the engine to finish identification. On the basis of the scheme, all defects on all the blades can be identified only by carrying out an identification process on the identification model once; the process has enough detection accuracy and does not need a large amount of calculation; and relatively high identification and blade quality detection efficiency can be kept.

The invention discloses a police target tracking system based on deep learning, and relates to the field of police video monitoring. The system comprises a video sensing and front-end processing unit, an edge intelligent tracking server and a cloud model optimization and command platform, the video sensing and front-end processing unit collects visible light and thermal infrared video streams, the edge intelligent tracking server processes the visible light and thermal infrared video streams, and a stable track with an identity label is generated; and the cloud model optimization and command platform aggregates edge data based on a federated learning framework, iteratively optimizes a global model and realizes command scheduling. According to the system, the tracking robustness, the real-time performance and the multi-target identity keeping accuracy in a complex environment are effectively improved, and the actual combat requirements of police affairs are met.

The invention belongs to the technical field of oil and gas pipeline safety monitoring and unmanned aerial vehicle image analysis, and discloses a two-stage oil and gas pipeline occupation pressure monitoring method and system based on unmanned aerial vehicle space perception and deep learning fusion, and the method comprises the steps: obtaining image data and space data of a to-be-monitored region, combining the image data and the space data with a DEM, and obtaining a DEM image; realizing projection of geographic coordinates of the pipeline to image pixel coordinates to obtain a plane image of the pipeline; constructing and deploying a lightweight target detection network DRF-YOLOv8n, performing reasoning on a plane image of the pipeline based on the lightweight target detection network DRF-YOLOv8n, outputting a candidate bounding box, and performing screening and key frame judgment; a Promp-guide SAM pixel level segmentation module is used to realize pixel level segmentation of the key frame, and a candidate pixel mask is generated; and judging whether occupation occurs or not through the intersection of the pipeline projection line and the candidate pixel mask, and outputting alarm information with geographic coordinates.

The invention discloses a three-dimensional target detection method, device and equipment based on an improved Pilllar structure and a medium, and relates to the field of three-dimensional point cloud processing and target detection. Dividing the original LiDAR point cloud data into a plurality of Pill units through a height perception double-branch coding module of the three-dimensional target detection model, and carrying out parallel branch feature coding to obtain a pseudo image feature map; global maximum and average pooling is carried out through a space attention feature extraction module to generate a space attention weight map, then the space attention weight map is multiplied by a pseudo image feature map point by point to obtain an enhanced weighted feature map, and the robustness of the model in a complex background and other scenes is enhanced; multi-scale context feature extraction fusion and receptive field expansion are carried out through the multi-scale feature extraction module, details can be seen clearly, the global context can be understood, and the problem that a long-distance target is difficult to detect due to sparse point clouds is solved; target prediction is carried out through the SSD detection head module, and the real-time requirements of applications such as automatic driving are met.

The invention discloses a distributed photovoltaic panel anomaly detection method under a high-altitude view angle based on a YOLOv8n model, and belongs to the technical field of target detection. The model comprises the following steps: acquiring a distributed photovoltaic panel image data set under a high-altitude view angle, and performing data enhancement and labeling; an improved YOLOv8n model is constructed, and the improvement comprises the steps that a C2fAT module is introduced into a backbone network, and the small target feature extraction capacity is enhanced; an SPPF module is replaced by an SPPF-LSKA module, and complex background interference is suppressed; an EMA attention mechanism is introduced into the neck network, and the multi-scale adaptive capacity is improved; optimizing a training process by adopting a WIoU v3 loss function; and training and optimizing the model by using the training set, and finally outputting a detection result through the test set. According to the invention, the problems of false detection and missing detection of the distributed photovoltaic panel in a high-altitude view angle are effectively solved, the detection precision of the distributed photovoltaic panel is further improved, and inspection personnel are helped to troubleshoot the photovoltaic panel in an abnormal state in the high-altitude view angle.

The invention discloses a blue-green algae identification and quantification method and device based on unmanned aerial vehicle remote sensing images and deep learning and a medium, and relates to the technical field of information data processing. Combining meteorological data and water quality monitoring data to construct an adaptive dynamic environment algorithm to extract EXIF metadata including camera parameters and attitude angle information, and establishing a geometric projection model to perform coarse orthographic correction on an original aerial image; the method comprises the following steps: extracting a multi-scale cyanobacterial bloom image feature map by stages based on a ResNet architecture and in combination with a feature pyramid network FPN fused with an attention mechanism, and obtaining an instance mask of cyanobacterial bloom through an anchor-free region proposal network, ROI Align and a head network; based on an improved GIS space projection and deep learning algorithm, carrying out high-precision area measurement and calculation on a binary mask image converted from the instance mask; according to the method, the influence of irrelevant interference on blue-green algae identification can be reduced, and the accuracy and comparability of blue-green algae identification and area calculation are improved.

According to the invention, the current detection precision of the current sensor is improved. A current sensor (3) is provided with: a fluxgate (32); an annular first core (31) made of a magnetic body and having an accommodating space (314) for accommodating the fluxgate (32); and a signal generation circuit (50) that generates a signal based on the magnetic flux detected by the fluxgate (32), and the current sensor (3) detects a current flowing through a measurement object (5) in a state where the measurement object is inserted into the first core. In a current sensor (3), a fluxgate (32) has: a second core (321) made of a magnetic body and extending in an accommodation space (314); and a coil (323) wound around the second core (321) in the direction in which the second core (321) extends. One end of the second core (321) in the direction of extension is in contact with a part of the first core (31), and the other end of the second core (321) in the direction of extension is in contact with the other part of the first core (31).

The invention discloses an underwater target detection method based on differential attention. The method comprises the following steps: acquiring an underwater image to be detected; based on the underwater image data set, constructing a lightweight underwater target detection network model based on a differential attention mechanism; constructing a joint loss function, and performing end-to-end training on the lightweight underwater target detection network model to obtain a trained lightweight underwater target detection network model; performing mathematical equivalence fusion on the lightweight underwater target detection network model to generate a lightweight reasoning model; and inputting the obtained underwater image to be detected into the lightweight reasoning model to realize target detection of the underwater image. According to the method, a differential attention feature interaction mechanism is introduced, common-mode noise is stripped through differential calculation, and the detection precision and real-time performance of an underwater fuzzy scene and a tiny target are remarkably improved in combination with a feature pyramid network integrating dynamic scale perception and semantic gating.

The invention discloses a multi-layer semantic perception, distillation and semi-supervised cooperative training target detection method, and the method specifically comprises the following steps: constructing a data set: extracting a first part of images from image data, marking the first part of images to construct a supervised target detection data set, and taking the remaining images as an unmarked image data set, the data volume of the unlabeled image data set is greater than that of the supervised target detection data set; teacher model optimization: performing supervision training on the teacher model on the supervised target detection data set; constructing a teacher model, and executing self-distillation training on the teacher model; pseudo labels are generated for the unlabeled images through a teacher model, and adaptive screening is carried out based on confidence distribution; mapping a pseudo label to a strong enhanced sample through enhanced geometric transformation, carrying out semi-supervised training by using the strong enhanced sample and the pseudo label, and introducing a feature layer distillation constraint at the same time; optimizing a student model; and outputting the target detection model obtained through training.

The invention discloses an outer wall hollowing detection method and device based on thermal infrared and facade point cloud fusion and a storage medium, relates to the technical field of building structure safety detection and intelligent sensing, and discloses the outer wall hollowing detection method based on thermal infrared and facade point cloud fusion. Comprising the steps that a thermal infrared image collected by a thermal infrared camera for a to-be-detected outer wall and a vertical face laser point cloud collected by a laser radar for the to-be-detected outer wall are acquired; determining a temperature abnormal area in the thermal infrared image and an outer wall point cloud subset corresponding to the temperature abnormal area in the facade laser point cloud; performing reference surface fitting on the outer wall point cloud subset data, constructing an outer wall normal deformation model reflecting concave-convex characteristics of a wall surface based on a fitting result, and extracting a local deformation abnormal region from the outer wall normal deformation model; and fusing the detection results of the temperature abnormal region and the local deformation abnormal region, and judging a target outer wall hollowing detection result. The technical effect of improving the outer wall hollowing detection accuracy is achieved.

The invention discloses a life body feature separation method based on millimeter-wave radar signal enhancement, and the method comprises the following steps: obtaining an original radar signal, collected by a millimeter-wave radar, of a life body monitoring target, and carrying out the preprocessing of the original radar signal; six types of features are extracted from the preprocessed signals, and the preprocessed signals are divided into four scene types based on the six types of features; according to the divided scene categories, calling a corresponding signal processing model to enhance or reconstruct the preprocessed signal to obtain an enhanced radar signal; and performing signal quality evaluation on the enhanced radar signal, inputting the enhanced signal meeting the quality evaluation requirement into a variational mode decomposition module, and separating a respiration signal and a heartbeat signal of the life body monitoring target. According to the method, the millimeter wave radar vital sign signals in a complex scene are subjected to scene classification and adaptive calling of the model to enhance the signals, breathing and heartbeat are separated through variational mode decomposition, and the vital sign detection precision in a complex environment is improved.

The application relates to the technical field of liquid chromatography tandem mass spectrometry all-in-one machines, in particular to a high-performance liquid chromatography tandem mass spectrometry all-in-one machine which comprises a bottle box, the top of the bottle box is used for placing a solvent bottle, a fixing clamping plate for limiting the solvent bottle is arranged on the bottle box, a side edge buckle slot is formed in the side edge of the bottle box, a refrigeration box is arranged in the side edge cooling box, an intelligent control screen is mounted on the surface of the refrigeration box, a pull-out box is connected to the side edge cooling box through a pull-out slot, and a box handle is arranged on the pull-out box. On the sample management, the refrigeration box in the side edge cooling box cooperates with the intelligent control screen to realize intelligent temperature control, the influence of manual operation on the refrigeration environment is avoided, the arc-shaped clamping groove, the rubber pad and the compression spring structure in the pull-out box can stably clamp a test tube, the sample pollution risk is reduced, the detection precision is improved, and meanwhile, the detachable connection design of the fixing clamping plate enhances the use flexibility of the equipment.

The utility model discloses an automatic probe mechanism component of a numerical control machine tool, which relates to the technical field of automatic detection of machine tools and comprises a machine tool ram, one end of the machine tool ram is detachably connected with an L-shaped fixing plate, the lower end of the L-shaped fixing plate is detachably connected with a marble base, and the lower end of the marble base is detachably connected with a linear module. Circular double guide rails are arranged in the linear module, a movable sliding plate is arranged between the two circular double guide rails, a probe seat is detachably connected to the lower end of the movable sliding plate, a probe is arranged on one side of the probe seat, and a headless screw used for fixing the probe is arranged at the lower end of the probe seat. According to the automatic probe mechanism component of the numerical control machine tool, the probe base is converted to the horizontal plane through the L-shaped fixing plate, and meanwhile, the stability of a probe mounting track is effectively improved under the cooperation of the marble base station; the telescopic design effectively protects the waterproofness and safety of the probe, and further improves the detection precision of the probe.

This utility model provides a condenser tube sheet sealing testing device, relating to the field of non-destructive testing technology. It includes a movable plate, a housing, and a rotating component. Four sets of casters are installed at the lower end of the movable plate, and a fixing component is installed at the upper rear side of the movable plate. The fixing component has a circular hole on its inner side, and a gas storage tank is fitted inside the circular hole. A movable component is fitted through a groove on a partition plate. The movable component has a circular hole at its center end, through which the front end of a thin flexible tube passes. This design utilizes the circular hole of the movable component to provide support and positioning for the thin flexible tube, preventing excessive bending, twisting, or compression during use, thus extending the service life of the thin flexible tube. It also ensures the normal working condition of the thin flexible tube and prevents damage to the tube from affecting the test results. This solves the problem of excessive bending and twisting of the gas tube during use in existing condenser tube sheet sealing testing devices.

This invention discloses a self-learning method and apparatus for the characteristic curve of a Venturi differential pressure sensor, comprising: acquiring vehicle operating conditions; determining the standard flow rate of an EGR valve when the vehicle operating conditions meet preset operating conditions; acquiring a first differential pressure value detected by the Venturi differential pressure sensor and a second differential pressure value based on the first differential pressure value after self-learning; determining a first flow rate of the EGR valve based on the first differential pressure value, determining a second flow rate of the EGR valve based on the second differential pressure value, and determining a first difference between the standard flow rate and the first flow rate, and a second difference between the standard flow rate and the second flow rate; and determining the characteristic curve of the Venturi differential pressure sensor based on the first difference and the second difference. This application compares the calculated standard flow rate of the EGR valve under set operating conditions with the EGR valve flow rate corresponding to the actual differential pressure value measured by the Venturi differential pressure sensor and the EGR valve flow rate corresponding to the self-learned differential pressure value. Based on the comparison results, it can determine whether to apply the self-learned characteristic curve, thereby improving the detection accuracy of the Venturi differential pressure sensor.

The application relates to the technical field of remote sensing image detection, and discloses a hyperspectral remote sensing image anomaly detection method, a system, equipment and a storage medium. The method comprises the following steps: acquiring a hyperspectral remote sensing image to be detected; inputting the hyperspectral remote sensing image to be detected into a pre-trained anomaly detection model to perform anomaly detection, and obtaining a detection result of the hyperspectral remote sensing image to be detected; and the anomaly detection model comprises a first convolution module, a second convolution module, a first Transform module, a first Resize layer and a first full connection layer which are sequentially connected. The application detects background and abnormal samples through clustering inspiration, effectively reduces the dependence of a network model on artificial labeling samples, and not only distinguishes abnormal information and background information during the training of the anomaly detection model, but also utilizes a reconstructed image of the hyperspectral image to improve the training effect of the model, so that the whole model has a local and global receptive field, and the detection precision of the anomaly detection model is effectively improved.

The application discloses a kind of landing gear inner tube surface scratch weak magnetic flaw evaluation method, its evaluation method includes the following steps: first need to carry out surface cleaning pretreatment to landing gear inner tube to be measured, and in the condition without additional excitation magnetic field, using geomagnetic field as natural excitation source, geomagnetic field environment calibration is carried out to landing gear inner tube surface, establishes background magnetic field baseline, according to the outer diameter size and surface curvature radius R of landing gear inner tube.The application adopts flexible carrier to carry multiple sensors, and good adhesion of sensor and inner tube surface is realized by adaptive layout mode, effectively overcome the influence of landing gear inner tube cylindrical surface curvature on detection signal consistency, improve the repeatability and reliability of detection, adopt the adaptive noise suppression method based on wavelet threshold, can retain weak scratch magnetic leakage signal characteristics while effectively filtering geomagnetic background noise and circuit noise, compared with traditional filtering method has better signal fidelity.

The utility model discloses a photovoltaic cell EL defect sorter, and relates to the technical field of sorters. Comprising a sorter main body, the sorter main body is provided with a photovoltaic cell EL defect detection mechanism, the detection mechanism comprises a moving assembly, the moving assembly comprises an opening groove formed in the top end of the sorter main body, a positioning rod is fixed to the inner wall of the opening groove, and a sliding rail support is slidably connected to the outer wall of the positioning rod; after the camera lens is detected, the protective shell is screwed on the bolt rod on the outer side of the camera lens, the camera lens is protected through the protective shell, external physical collision and scraping can be effectively resisted, tools, parts and other hard objects are prevented from directly impacting the lens under the daily operation, maintenance or unexpected conditions of equipment, and the camera lens is protected. The surface of the lens is prevented from being scratched and damaged, the optical structure integrity of the lens is protected, the definition and accuracy of a detection image are guaranteed, and pollution of impurities and dust is prevented.

This invention discloses a face forgery detection method based on identity decoupling and adaptive cosine embedding loss. Addressing the problems of existing deepfake detection methods being susceptible to interference from identity information and lacking flexibility due to fixed boundary constraints, this invention constructs an identity-independent deepfake detection network. It decouples the input face image into forensic features containing forgery clues and content features containing identity information, eliminating identity interference by ignoring content features. Simultaneously, it designs an adaptive cosine embedding loss based on visual differences, using absolute error and the visual differences between real and fake image pairs as adaptive boundary values ​​to dynamically adjust feature spacing constraints. This invention significantly improves the model's generalization ability when facing unknown forgery types through the synergistic optimization of decoupled representation learning and adaptive metric learning.

The present application belongs to the technical field of concrete preparation detection, and particularly relates to a concrete aggregate water content detection device, which comprises a base and a rack installed above the base. A controller is installed on the side wall of the rack. The device further comprises a through hole opened in the top of the base, a support seat slidably connected inside the through hole, and a weighing mechanism installed inside the through hole for supporting the support seat. A flat hopper is arranged above the base and located on the side of the feed end direction of the support seat. The rack is provided with an adjusting drive mechanism for driving the vertical movement of the flat hopper. A microwave detection unit is arranged on one side of the flat hopper. The present application can perform online detection of the water content of concrete aggregate, improve the detection accuracy by combining weight detection and thickness control, reduce single detection interference error by cross verification of multiple sensing data, and improve data reliability to support subsequent dynamic batching.

The application discloses a visual perception method for assisted driving, comprising the following steps: S100, collecting assisted driving image data, and constructing input X after data set preprocessing; S200, constructing a multi-task neural network model F and substituting the input X into the model to obtain a plurality of branch task-specific outputs f i ; S300, calculating the final loss function L all of the multi-task neural network model F; S400, optimizing the minimum as the target of the loss function L all , and finally obtaining the optimal multi-task neural network model F * . A system using the above visual perception method for assisted driving is also disclosed, specifically comprising a data input module, a shared layer feature extraction module, a task layer structure search module and a model output module. The application has the beneficial effect that a multi-task model is constructed to effectively reduce the network parameter quantity and the calculation quantity, simultaneously reduce the online reasoning time, and is very beneficial to practical application.