53results about How to "Avoid false detection" patented technology

The invention discloses a rice seedling leaf age intelligent identification system based on a depth camera, and belongs to the technical field of agricultural intelligent equipment and machine vision. Comprising a monitoring point gridding path planning module, a mechanical arm cooperative positioning data acquisition module, an RGB-D data preprocessing module, a dynamic foreground seedling accurate extraction module, a leaf age recognition confidence evaluation module, a leaf age data visualization presentation module and an integrated control strategy feedback module. RGB color features and connected domain analysis are combined to optimize a mask, a front single seedling is accurately segmented, background interference is eliminated, the confidence coefficient is dynamically calibrated through a deep learning model and image quality features, texture definition and edge continuity features are combined to generate calibrated confidence coefficient, confidence coefficient evaluation is directly associated with an agricultural decision, and the accuracy of the agricultural decision is improved. The leaf age result is generated only based on reliable data, the personnel rechecking demand is reduced, and the decision-making efficiency is improved.

The application belongs to the technical field of optical lens detection, and particularly relates to an optical lens appearance defect detection device, which comprises a supporting assembly, a feeding device is installed on the supporting assembly, a first discharging slide plate is installed at the discharging end of the feeding device, a mounting seat is fixedly connected to the supporting assembly below the first discharging slide plate, supporting rollers are rotationally connected to both ends of the mounting seat, and the cooperation of the structures such as the belt ring, the first visual defect detection camera, the pushing cylinder, the guide plate and the driving roller is achieved, so that after the optical lens is accurately positioned in the positioning groove formed in the belt ring, the first visual defect detection camera arranged on both sides is used to detect the end of the optical lens, then the pushing cylinder is used to push the optical lens to the driving roller, the driving roller is rotated to drive the optical lens to slowly rotate, the second visual defect detection camera is used to detect the circumferential side of the optical lens, and the mechanical hand is not needed to be used for carrying and aligning, so that the detection efficiency is greatly improved.

The invention relates to a steel bar reverse bending test device, and belongs to the field of steel bar detection devices. The device comprises a rack and a workbench rotationally arranged on the rack. A supporting shaft is arranged on the rack; the supporting shaft penetrates through the workbench upwards and is coaxial with the workbench. A plurality of mounting holes are formed in the workbench; a bending block is inserted into any mounting hole; a working plate is arranged on the rack in the vertical direction in a sliding mode. A sliding rail is arranged on the working plate; scale marks are arranged in the length direction of the sliding rail. A first supporting piece is slidably arranged on the sliding rail. A first through hole is formed in the first supporting piece; a second supporting piece is arranged on the working plate; a second through hole is formed in the second supporting piece; the steel bars sequentially penetrate through the first through holes and the second through holes. A first image sensor is arranged on the rack; the method has the technical effects of avoiding manual calculation of the steel bar bending angle and improving the steel bar detection precision.

The application discloses a cable surface sheath image preprocessing method, device and equipment and a medium, and relates to the technical field of image processing. The method comprises the following steps: collecting a cable surface sheath image and dividing the image into multiple regions; calculating the average gray value and standard deviation of all pixels in each region, and determining a deviation range based on the 3σ criterion; determining pixels with a gray deviation not within the deviation range as abnormal pixels; determining whether there are other abnormal pixels in the eight-neighbor domain of each abnormal pixel, filtering the abnormal pixel if there are no other abnormal pixels, and constructing a connected domain composed of abnormal pixels if there are other abnormal pixels; filtering connected domains with an area less than a threshold value, and retaining connected domains with an area greater than or equal to the threshold value, to obtain a preprocessed image. Through local adaptive statistics, connected domain analysis and selective processing based on the area, the application overcomes uneven illumination and reflection interference caused by a cable curved surface, and helps to reduce the false positive rate and the missed detection rate of subsequent defect detection.

The application discloses a converter automatic slag splashing control method and system based on deep learning, relates to the technical field of converter steelmaking, and acquires a current converter opening image in a slag splashing process in real time, then inputs the current converter opening image into a trained deep learning model, judges whether the current slag splashing type is in a non-slagging state or a slagging state, returns to step S1 if the current slag splashing type is in the non-slagging state, and determines the current slag amount according to the current converter opening image if the current slag splashing type is in the slagging state, and finally controls the oxygen lance to perform corresponding processing actions according to the current slag amount, so that the problem of slag block misjudgment caused by light and smoke interference in the non-slagging state is avoided, and the accuracy of slag splashing control is improved.

The invention provides a radar point cloud-based foreign object intrusion detection method and system in a working area, and the method comprises the steps: setting a detection area range according to the working space of a mechanical connecting rod, and dynamically generating a direction bounding box (OBB) of the connecting rod through obtaining the posture state information and size data of the working connecting rod in real time; spatial screening is carried out on original radar point cloud data, only point cloud data which are located in a detection area and are not located in an OBB are reserved, and background point clouds outside the detection area and point clouds generated by a connecting rod are filtered out; and performing multi-dimensional anomaly detection on the filtered point cloud data, and comprehensively judging whether a continuously existing foreign object intrusion event exists or not. Through a linkage mechanism of point cloud filtering, background modeling and clustering verification, high-precision, low-false-alarm and real-time industrial intrusion detection is realized, and the problems of poor adaptability, high false alarm rate and the like of a traditional method in a complex environment are solved.

The invention provides a color feature extraction and correction method and system for drug sensitive plate MIC value interpretation and a computer readable storage medium. The method comprises the following steps: acquiring a to-be-processed drug sensitive plate image; performing visual preprocessing such as exposure, contrast and white balance on the image according to the configuration file; high-precision alignment correction of the image is realized by using ORB feature detection and a BFMatcher matching algorithm; generating a transparent mask with an Alpha channel based on predefined hole position contour data, and performing non-destructive target segmentation on the aligned image; and converting the target region to an HSV color space, extracting H-channel hue features, executing cyclic correction, eliminating the problem of numerical value jump of the red region, and obtaining stable color feature data. The method has the advantages of being high in image registration precision, accurate in segmentation, stable in color feature, high in robustness and the like, and automatic and standardized interpretation of the MIC value of the drug sensitive plate can be achieved under different shooting conditions.

The invention relates to a target detector training method and device, computer equipment and a storage medium, and the method comprises the steps: obtaining a plurality of pre-constructed training batches; each training batch comprises a plurality of style image samples of at least one vehicle external target instance; through a pre-constructed student detector and a pre-trained teacher network, respectively extracting a first feature vector and a second feature vector of a target annotation box in each style image sample; the teacher network is used for extracting feature vectors with invariable styles; constructing an instance consistency loss function according to the first feature vector; constructing a distillation loss function according to the first feature vector and the second feature vector; and training the student detector based on the preset detection loss function, the instance consistency loss function and the distillation loss function to obtain the target detector for detecting the external target of the vehicle, thereby solving the problem of insufficient cross-domain generalization ability of the target detector, and improving the cross-domain generalization ability of the target detector.

This application discloses a display device, comprising a display panel and a driving module. The driving module is connected to the display panel and is configured to drive the display panel in a first driving mode when it detects that the grayscale to be displayed on the display panel is less than or equal to a first preset grayscale. In the first driving mode, when the driving module detects that the grayscale to be displayed is greater than a second preset grayscale, it switches the first driving mode to a second driving mode, where the second preset grayscale is greater than the first preset grayscale. This application can reduce false detections caused by grayscale fluctuations and avoid display problems such as screen flickering.

The utility model belongs to the technical field of detection equipment provides a kind of vice instrument panel detection tool, vice instrument panel is provided with peg, peg is made of metal material, vice instrument panel detection tool includes: pedestal;Positioning seat, it is set in the top of pedestal, positioning seat is provided with the detection through-hole compatible with for vice instrument panel, and positioning seat and pedestal between there is installation space;Compression assembly, it is set in pedestal, compression assembly has the compression operating mode and loosening operating mode that can be switched to each other;Wherein, when compression assembly is in compression operating mode, compression assembly compresses vice instrument panel;When compression assembly is in loosening operating mode, compression assembly loosens vice instrument panel;Detection component, it is set in pedestal, and detection component is used to detect the peg on target vice instrument panel.The vice instrument panel detection tool provided by the utility model is simple in structure, can automatically detect the peg on vice instrument panel, and is higher in efficiency.

The invention relates to the technical field of microlens detection, in particular to an automatic microlens detection method and system, and the method comprises the steps: obtaining a stripe image of a microlens under the illumination of a stripe structure; preprocessing the stripe image to obtain a preprocessed image with a block area as a foreground; determining a rotation angle according to the boundary of the reference block region, and rotating the corrected stripe image to align the stripe with the row direction; segmenting the rotating image, identifying a connected block, solving an external frame to generate a block mask and a complementary set interval mask thereof, and obtaining a block image and an interval image; performing segmentation on the block image, taking a median value according to rows to replace and generate a block reference image, performing differential binarization to obtain a block defect candidate, and performing independent binarization and merging on an abnormal region; overlapping the interval image and the block mask, performing closed operation to suppress line noise, and then performing binarization inversion to obtain an interval defect candidate; and superposing the two candidate output defect mark images. According to the invention, the defect can be automatically identified and marked and output by the microlens under the stripe background.

The invention discloses a self-adaptive control method for intelligent mixing of roadbed solidified soil at an open cut tunnel section of a tunnel, relates to the technical field of road roadbed engineering construction, and solves the technical problems that an existing roadbed solidified soil mixing effect evaluation method lags behind time and highly depends on manpower. According to the method, the mixing state of the road mixer can be quickly and effectively judged by utilizing the characteristic that the gray difference between the curing agent and the soil matrix is large, calculating the global gray value of the image of the mixing area and preliminarily evaluating the mixing effect, and meanwhile, the gray-scale map of the mixing area is divided into the curing agent area and the soil matrix area by adopting an OSTU binarization segmentation method, so that the mixing effect of the road mixer is improved. Based on the distribution characteristic of the image gray scale, the curing agent can be effectively separated from the soil matrix. The image is divided into a plurality of windows, the proportion of the curing agent in each window is counted, and the variance of the proportion of the curing agent in all the windows is calculated, so that the distribution uniformity of the curing agent among the windows can be effectively judged, and the mixing effect can be effectively evaluated.

This invention relates to the field of indoor geomagnetic positioning technology, specifically to a method for dynamic maintenance of geomagnetic fingerprint maps based on mobile robots. The method includes multimodal data acquisition and spatiotemporal alignment to establish a correspondence between sensor data and global map coordinates; constructing a cross-confidence mechanism using geomagnetic anomaly factors and visual verification factors, and accurately triggering map updates using a time decay accumulation model; identifying newly added magnetic objects and extracting their parameters, calculating the radius of influence of the disturbance field using a magnetic field disturbance model; and employing a region-growing local incremental update method, introducing a boundary smoothing factor to ensure map continuity. This invention achieves autonomous and efficient maintenance of geomagnetic fingerprint maps without requiring global rescanning or additional hardware, improving the stability of geomagnetic positioning for mobile robots. Simultaneously, the accurate triggering of updates through a dual-modal cross-confidence evaluation mechanism enhances the accuracy of geomagnetic fingerprint map updates.

This invention discloses a dense microchip instance segmentation method using a category-mixed mask fusion convolutional neural network, belonging to the field of chip instance segmentation and detection. First, a chip wafer optical image dataset is constructed, divided into a test set and a training set. Then, an initial model of the category-mixed mask fusion convolutional neural network is constructed. This initial model is trained using the established training set to obtain a final model. Finally, the final model is used to segment the constructed test set, completing the instance segmentation of the chip images. This dense microchip instance segmentation method using a category-mixed mask fusion convolutional neural network solves the problem of insufficient detection accuracy of existing detection technologies for industrial-grade dense microchip images. By strengthening the local features of each chip, it significantly improves the model's performance in category classification and fine localization.

The invention relates to a visual detection positioning method for a continuous casting bale collector nozzle, and belongs to the technical field of metallurgy intellectualization, and the method comprises the following steps: S1, driving a visual detection module to photograph and scan the continuous casting bale collector nozzle through a moving module, and collecting a depth image and corresponding point cloud data; s2, performing visual processing on the acquired depth image; s3, when an effective circle is not detected, performing circle reinspection by adopting an RANSAC algorithm; s4, determining an ROI (Region of Interest) of the point cloud according to an image detection result, and processing the ROI point cloud; and S5, the coordinates of the circle center are converted from the camera coordinate system to the robot base coordinate system through the hand-eye calibration matrix, the attitude angle of the target in the robot base coordinate system is calculated and communicated to the long nozzle four-axis robot, and automatic long nozzle installation operation is completed. The invention further provides a visual detection positioning system for the continuous casting bale collector nozzle.

The application relates to a defective product detection paper shell paste box machine and relates to the technical field of paper shell detection. The machine comprises a paper drawing mechanism, a paper conveying mechanism and a detection mechanism which are sequentially connected. The paper drawing mechanism and the paper conveying mechanism are arranged on a workbench. The paper drawing mechanism is used for sequentially conveying paper shells to the paper conveying mechanism. The paper conveying mechanism is used for conveying the paper shells to the detection mechanism. The detection mechanism comprises a detection assembly and a kick-off assembly which are arranged on the workbench. The detection assembly is used for detecting the quality of the paper shells and conveying the paper shells to the kick-off assembly. The kick-off assembly is used for separating the unqualified paper shells. The application has the effects of automatically separating defective products and improving production efficiency.

The application discloses a kind of vision-based industrial robot intelligent discharging device, system and method, first discharging line is arranged in the discharge end of the transmission unit, including conveying unit, separating mechanism, visual detection mechanism and pushing mechanism;Conveying unit is arranged in the discharge end of the transmission unit, for conveying material;Separating mechanism is arranged in the first half of the conveying unit, for separating multiple materials;Visual detection mechanism is arranged in the middle section of the conveying unit;Pushing mechanism is arranged in the tail end of the conveying unit, for pushing out the material that detection is not up to standard.The present application avoids missing detection and wrong detection, makes the detection result more accurate, improves product quality consistency, greatly improves detection and discharging efficiency, avoids channel blockage and jam caused by mixed conveying of different height materials, ensures smooth conveying, avoids fatigue error caused by long-time operation, has the beneficial effects of low manufacturing cost, simple operation and not easy to damage, etc.

The embodiment of the invention provides a personnel behavior determination method, system, device and equipment and a storage medium, and relates to the technical field of artificial intelligence, and the method comprises the steps: carrying out the human shape detection of a video frame in a to-be-processed video, and obtaining a first image region where a person is located in the video frame; determining a target person in the video frame based on the face feature of the person in the first image area and a pre-stored template face feature of the target person; and inputting a to-be-processed video marked with the target person in the video frame, a first prompt word used for scene filtering and a second prompt word used for behavior filtering into the target visual language model to obtain an output behavior of the target person. Behavior detection is carried out in combination with the to-be-processed video providing the visual features, the first prompt word providing the scene semantics and the second prompt word providing the behavior semantics, and the accuracy of behavior detection is improved.

The application discloses a large language model hallucination detection post-processing method and system based on adaptive order statistics aggregation, equipment and medium. In view of the fact that hallucination signals in long text responses in a retrieval augmented generation (RAG) scene have significant sparsity and locality, which leads to the technical problem that a conventional global pooling strategy has signal dilution or noise sensitivity, an adaptive sparse signal extraction and decision framework is provided. The method comprises the following steps: acquiring a time sequence feature matrix of a response sequence; constructing an adaptive sampling window in a non-linear mapping relationship with the response length; performing order statistics filtering and mean aggregation on each feature dimension based on the window to extract local significant abnormal signals; constructing a response level feature vector containing multiple order statistics moments; performing non-linear mapping by using a gradient boosting tree model, and outputting a final decision by combining a dynamic boundary calibration strategy based on F1 score maximization.

The invention relates to a method and system for detecting same-color pollutants in a snowfield area near a garbage can, and the method comprises the steps: obtaining the image data of the garbage can, and extracting the position information of the garbage can through the image data; determining a plurality of candidate detection areas based on the position information, and performing snowfield judgment on the plurality of candidate detection areas through a snowfield feature recognition algorithm to obtain a snowfield area identification result; carrying out grid division on the candidate detection area which is identified as a snowfield according to the snowfield area identification result to obtain a plurality of sub-areas; and extracting color feature parameters and texture feature parameters of the plurality of sub-regions, and performing same-color pollutant judgment on the plurality of sub-regions through a feature combination scoring algorithm to obtain a detection result. The detection accuracy of the same-color pollutants in the snowfield environment is improved.

The application discloses an unmanned aerial vehicle reconnaissance video image target automatic detection method, and belongs to the technical field of unmanned aerial vehicle image processing. The method comprises the following steps: preprocessing and frame extraction are performed on an unmanned aerial vehicle reconnaissance video image, a video frame sequence is constructed, multi-scale feature extraction is performed on the video frame sequence, multi-scale features extracted based on semantic guidance are fused, a time sequence memory model is constructed, the multi-scale features are corrected and enhanced in a time sequence context, a weak and small target detection branch is generated based on multi-dimensional feature perception, the multi-scale features corrected and enhanced are detected in the weak and small target detection branch, and a detection result is output. Through grouping convolution, the method can efficiently fuse, guarantees the calculation efficiency, and enhances the contour perception ability for fuzzy weak and small targets.

The utility model discloses an inductor surface defect detection device, which relates to the technical field of inductor surface defect detection and comprises a mounting component, the mounting component comprises a bottom plate, side plates are fixedly mounted on two sides of the bottom plate, and a top plate is fixedly mounted at the tops of the side plates. A rotating assembly is fixedly mounted in the middle of the top end of the bottom plate, a mounting hole is formed in one side of the top plate, two universal joint assemblies are fixedly mounted in the mounting hole, camera assemblies are fixedly mounted on one side of the bottom of the rotating assembly and the bottoms of the universal joint assemblies, and each camera assembly comprises a camera. An annular light supplementing lamp is fixedly installed on the outer side of the camera, the light emitting direction of the annular light supplementing lamp is consistent with the camera shooting direction of the camera, and a control panel is fixedly installed on the top of one end of the top plate. According to the utility model, the illumination can uniformly cover the detection station, the detection error caused by non-uniform light distribution of the traditional detection device is avoided, and the practical value is higher.

The utility model relates to a pull ring detection technical field, concretely relates to a pull ring detection device. The pull ring detection device, include: conveying assembly, conveying assembly is used for conveying jar body, detection assembly, detection assembly sets up on conveying assembly, detection assembly includes image acquisition structure, image acquisition structure has the end with pull ring to jar body orientation, to make image acquisition structure obtain jar body pull ring's image, control system, control system has controller and man -machine interface, controller is connected with image acquisition structure, man -machine interface and conveying assembly signal respectively. Through setting up image acquisition structure and control system, the image that controller can gather through image acquisition structure judges whether the pull ring of jar body that passes through image acquisition structure on conveying assembly has the defect, detects through the controller, effectively has improved the detection efficiency, has reduced the labor intensity.

The utility model discloses a kind of electric energy meter pulse lamp and tripping lamp machine detection equipment, including conveying line, table drag device, optical signal acquisition device, screen turning key pressing mechanism, positioning mechanism and defective product rejection mechanism.Equipment realizes the full-automatic detection of pulse lamp and tripping lamp by automation process, including power-on tripping lamp detection, pull-in and pull-out command verification and pulse lamp flicker detection, solve the problem of low efficiency of artificial detection, high risk of missed detection and misdiagnosis, realize unmanned operation.

The invention provides an image detection method and device, electronic equipment and a storage medium. The method comprises the steps of obtaining a to-be-detected image, performing transformation processing on the to-be-detected image to obtain a multi-scale frequency domain feature map, and performing extraction processing on the to-be-detected image to obtain an initial noise feature map; performing resolution processing on the to-be-detected image to obtain a multi-scale spatial feature map, and performing resolution processing on the initial noise feature map to obtain a multi-scale noise feature map; adding the multi-scale frequency domain feature map into the multi-scale spatial feature map to obtain an enhanced spatial feature map, performing fusion processing on the enhanced spatial feature map and the multi-scale noise feature map to obtain an initial fusion feature map, and performing aggregation processing on the initial fusion feature map to obtain a target fusion feature map; and determining a tampering probability graph based on the target fusion feature graph, taking the tampering probability graph as an image tampering detection result, and generating an image detection report based on the multi-scale frequency domain feature graph, the initial noise feature graph, the initial fusion feature graph and the tampering probability graph.

The present disclosure relates to processing cell images in a convolutional neural network with contextual data. The present disclosure relates to image processing in machine learning convolutional neural networks. The technology can improve the accuracy and reliability of applications using machine learning convolutional neural networks, particularly in the field of cell microscopy. Contextual data is provided which indicates cell features determined globally for a scene, and this data is then taken into account in the inference of the machine learning convolutional neural network.

The invention provides a frequency band detection method of an OFDM (Orthogonal Frequency Division Multiplexing) communication system and related equipment. The method comprises the following steps: performing sliding discrete Fourier transform on a received signal sampling sequence to obtain sliding discrete Fourier transform output corresponding to each subcarrier index; constructing a frequency band energy sequence of each candidate frequency band based on the sliding discrete Fourier transform output; defining an adaptive threshold value of each candidate frequency band based on the noise statistical parameter of each candidate frequency band in the preset noise interval; constructing a detection window based on the obtained window starting position; based on the frequency band energy sequence of each candidate frequency band, calculating the detection statistics of each candidate frequency band in the detection window and the corresponding confidence; and if the detection statistical magnitude is greater than the adaptive threshold and the confidence coefficient is greater than a preset confidence coefficient, determining each candidate frequency band as an effective communication frequency band. According to the invention, low time delay, low power consumption and accuracy of communication frequency band detection in the OFDM communication system can be considered.

This invention discloses a fully automated intelligent robotic workstation for nucleic acid sampling, detection, and analysis. It comprises a movable prefabricated building with multiple unit compartments arranged sequentially from left to right. These unit compartments, from left to right, are a sampling room, a sample tube loading room, a specimen preparation room, and a detection and specimen processing room. A transfer mechanism is provided between adjacent unit compartments. The sample tube loading room, specimen preparation room, and detection and specimen processing room are all enclosed spaces. Each unit compartment has a corresponding transfer hole on its wall for the transfer mechanism. The transfer mechanism includes a conveyor belt located at the bottom of the transfer hole and extending laterally, and an airtight door located within the transfer hole that divides the conveyor belt into two sections. This system can automatically collect nucleic acid samples and directly perform nucleic acid detection and analysis without needing to transport them to other locations, thus eliminating the risk of infection during transport. It can also receive nucleic acid samples collected from other locations for testing.

This application discloses a method and system for inspecting the surface quality of electroplated finished products. The method includes acquiring information about the electroplated finished product to be tested, including product attribute information and actual electroplating process parameters; determining image acquisition parameters based on the product attribute information and actual electroplating process parameters through a preset parameter mapping table; acquiring an image of the electroplated finished product based on the image acquisition parameters; extracting multi-dimensional optical features from the electroplated finished product surface image; fusing the actual electroplating process parameters with the multi-dimensional optical features to generate process-aware fusion features; and obtaining the surface quality inspection result of the electroplated finished product based on the process-aware fusion features through a preset defect detection model. This application effectively improves the accuracy of electroplated surface quality inspection by using adaptive imaging and fusing electroplating process information.

This invention discloses a machine vision-based method for detecting blockages in coal preparation plant screening tanks, belonging to the field of industrial automation detection technology. The method includes: acquiring continuous video frames; the user defining an irregular polygonal monitoring area and generating a mask; dividing the video frames into equally spaced blocks; performing normalized correlation template matching on each block within the monitoring area using adjacent frames, recording the positional deviation; marking blocks with deviations close to 0 in multiple consecutive frames as continuously stationary blocks; and determining blockage and triggering an alarm when the number of stationary blocks is greater than a threshold. This invention improves robustness to changes in illumination by utilizing normalized correlation matching, effectively filters out instantaneous misjudgments through continuous frame stationary detection and an accumulator mechanism, and solves the problem of water flow interference by combining edge detection preprocessing, thus achieving automatic and accurate detection of blockages in coal preparation plant screening tanks and ensuring production safety.