25results about How to "Recognition stability" patented technology

The invention discloses an intelligent detection method and system for garment quality inspection, and relates to the technical field of computer vision and intelligent manufacturing. The method comprises the following steps: carrying out pixel-level feature analysis on two-dimensional image data, identifying a sewing line and a printing area, and mapping the sewing line and the printing area to a unified space coordinate system; reconstructing the three-dimensional surface form of the garment based on a stereoscopic vision algorithm, and calculating the dimensional deviation of key parts; converting the spectral data into a CIELAB color space, and carrying out optical distortion correction on local chromatic aberration in combination with a three-dimensional curvature; and under a unified space coordinate system, performing correlation analysis on the stitches, the dimensional deviation and the corrected space distribution, identifying a composite defect which cannot be judged by a single defect, generating a correlation defect identifier, and finally forming a quality report. The technical problems that in the prior art, local chromatic aberration is easily influenced by three-dimensional curved surface deformation and optical distortion of ready-made clothes, information such as dimensional deviation, sewing and printing defects and the local chromatic aberration is isolated, comprehensive correlation analysis is lacked, and consequently the detection accuracy is insufficient are solved.

This invention discloses a method for identifying the connection location of underground dissimilar pipelines based on 3D image reconstruction, belonging to the field of location recognition technology. It addresses the problems of distorted spatial structure representation and insufficient accuracy in determining connection locations. By segmenting the underground pipeline, fusing ground-penetrating radar reflection data and electromagnetic detection results, abrupt gradient coefficients are calculated to mark potential material abrupt change segments. 3D scan data of the marked segments are retrieved, and combined with scan density and segment structure information, connection candidates are generated. Enhanced scanning is performed on the candidates, and the pit landing point is determined through spatial coarseness features. The existence of a connection interface is determined based on local point clouds, regional point clouds are extracted, and axis fitting is performed on the regional point clouds to obtain 3D misalignment data of the pipelines on both sides. The construction area is divided based on the soil obstacle density around the connection interface, and the connection center point is output based on misalignment constraints. This significantly improves construction accuracy and reduces the risk of blind excavation.

The application belongs to the technical field of vehicle information anonymization detection, and particularly relates to a multi-type image anonymization annotation dataset construction and target coverage rate determination method. The method is based on a face and license plate image dataset with double annotation of theoretical anonymization region and anonymization features, introduces an anonymization feature extraction branch in an improved YOLOv5-L model and performs cross-modal feature fusion, combines a plurality of loss functions with theoretical anonymization region positioning loss as the core and a two-stage progressive training and difficult example mining mechanism, and realizes precise learning of the anonymization features. After normalizing the input anonymization image, the model inference obtains the theoretical anonymization region coordinates and maps them back to the original size, and through non-maximum suppression and matching of the IoU threshold, the region coverage rate is calculated to determine the positive detection, missed detection and statistical false detection rate. The application effectively overcomes the feature dependency failure and model robustness problem, and realizes high-precision, low-misjudgment anonymization detection and evaluation under various anonymization conditions.

This invention relates to a method and system for cross-domain open set person re-identification based on unsupervised domain adaptation. The method includes: constructing a cross-domain feature adversarial alignment module including a feature extractor and a cross-domain discriminator; constructing an integrated binary classification bias module including multiple binary classification bias modules; alternately optimizing the cross-domain feature adversarial alignment module and the integrated binary classification bias module in an adversarial manner by combining multi-class cross-entropy loss, difference loss, and bias loss in the source and target domains; after the model converges, determining whether the person in the current test image belongs to a known object or an unknown object based on the re-identification result and bias score output by the model; this invention uses a joint training method to simultaneously perform cross-domain feature distribution adversarial alignment and unknown object identification, ensuring efficient re-identification of known objects while also identifying unknown objects with high accuracy.

The invention provides a self-adaptive grid generation method and device, electronic equipment and a storage medium. The method comprises the steps that a water depth prediction field is obtained on a coarse grid through model deduction; the water depth gradient of the coarse grid is calculated, and an encryption index is weighted in combination with the distance to the dike and the prediction uncertainty of the model; performing local encryption on a to-be-encrypted parent unit determined according to the encryption index to generate a subunit; the water depth on the parent unit is transmitted to the subunits to serve as initial water depth, and local fine water depth is obtained on the subunits through model deduction; at the junction, the local fine water depth and the water depth of the coarse mesh are fused; transmitting the water depth of the fused subunits back to the parent unit according to area weighted average, and correcting the water depth of the subunits according to the water depth of the parent unit; and when the parent unit does not change any more, outputting the grid structure at the moment. According to the method, the problems that a grid encryption area is unstable and engineering reusable grids are difficult to output are solved.

The invention discloses a road marking identifying and spraying method and system based on image processing, and the method comprises the steps: collecting a lane line image through a camera, and converting the lane line image into an aerial view; establishing rigid calibration of the pose relation between the nozzle and the camera to realize coordinate registration; converting the aerial view from an RGB space to an HSV space, introducing a self-adaptive threshold adjustment mechanism on this basis, and obtaining a lane marking area image from the lane line image according to self-adaptive segmentation; a lane image is converted from a pixel coordinate space to a parameter space, and automatic detection of lane edge lines and marked lines is realized through Hough transform; in combination with the detected lane marking area, the detection area is classified through a gray threshold value or a brightness feature, and the offset of a marking center line is calculated; the transverse position of the spraying mechanism is adjusted in real time according to the offset of the center line of the marked line, dynamic opening and closing control over the spraying head is achieved in combination with the vehicle running speed and the recognition signal state, and it can be guaranteed that the spraying track and the recognition center line are kept consistent.

This application discloses an obstacle recognition method and an automatic cleaning device. After acquiring multiple initial point cloud data frames from a ranging sensor, this application filters each initial point cloud data frame to obtain a target point cloud data frame corresponding to the obstacle. Then, based on the point cloud features of the point cloud dataset within a preset distance range in each target point cloud data frame, the obstacle category of each point cloud dataset is determined. Based on the obstacle categories of all point cloud datasets, the target obstacle category corresponding to the area where each point cloud dataset is located is determined. Finally, the automatic cleaning device is controlled to avoid obstacles based on the target obstacle category. The automatic cleaning device of this application can utilize the laser emitted by the ranging sensor to acquire environmental information over a large area to identify obstacles, eliminating the need for multiple retraction actions to scan complete obstacle information, thus improving the efficiency of obstacle recognition and overall work efficiency.

The present application provides a kind of termite indicator identification method and system based on optical and infrared image data, it is related to image recognition technical field, the present application obtains the gray optical image of current dam, and the spectral analysis of each pixel in it is carried out, to obtain the optimal direction and optimal micro-period texture response of the pixel, filter out the pixel of optimal micro-period texture response qualified as candidate pixel, based on the path expansion rule of optimal direction setting, to form multiple indicator direction chain with candidate pixel;Setting extension pixel judgment rule, based on the corresponding indicator chain of setting extension pixel judgment rule connection, based on the abutment relationship between indicator direction chain, indicator direction chain is merged, and based on the size of merged area, the optical image of dam is divided into indicator region, indicator pending area and other regions, obtain infrared image data, and analysis is carried out, to further identify indicator pending area.

The invention discloses a same-supplier power grid equipment operation reliability prediction method and system based on deep learning, and relates to the field of power grid technology and artificial intelligence. The prediction method comprises the following steps: acquiring and preprocessing multi-source heterogeneous data covering full life cycles of equipment design, production, factory test, online operation and the like so as to construct a standardized equipment feature matrix; performing multi-dimensional feature screening and dimension reduction on the feature matrix to extract a core feature subset with the highest prediction value; a deep learning reliability prediction model is constructed and trained, and the model takes the core feature subset as input and objective performance indexes reflecting the historical batch operation stability of the equipment as labels for supervised learning so as to autonomously learn a nonlinear mapping relation between the core features and the operation reliability of the equipment; and finally, according to the trained model, carrying out reliability prediction on new to-be-predicted equipment, and outputting a reliability prediction index of long-term operation of the new to-be-predicted equipment.

The application relates to the technical field of body sense interaction and virtual character control, and discloses a surfboard body sense input and character linkage system, which comprises a board body, a tension detection unit, a posture detection unit and a central processing unit. The system collects cable tension through the tension detection unit, collects an original pitch angle and a roll angle through the posture detection unit, and a decoupling module is built in the central processing unit. A mechanical coupling pitch angle is calculated according to the cable tension and a preset mechanical coupling coefficient, the original pitch angle is subtracted by the coupling pitch angle to obtain a net pitch angle, so that the physical torque interference is eliminated and a speed instruction is generated. Meanwhile, the system dynamically adjusts a roll dead zone threshold according to the tension intensity by using a dead zone module, and identifies a drift state based on a motion time window and a multi-dimensional parameter threshold through a judgment module. The application effectively solves the interference problem of the cable tension on posture recognition, and realizes accurate judgment and stable control of complex combined actions.

The invention discloses a ceramic green surface bubble identification method based on machine vision, particularly relates to the field of image processing and surface defect detection, and is used for solving the problem of false detection caused by artifacts generated by bubble brightness overturning and mirror surface reflection of a green glaze surface under different illumination. A first illumination image and a second illumination image are collected, an alignment image group is established, local illumination normalization and peripheral texture compensation are executed to form a highlight texture removal image group, a symmetric residual image and an abnormal response are calculated to obtain a candidate region set, and the candidate region set establishes corresponding association in the alignment image group to obtain a highlight texture removal image group. Extracting a main reflection direction and calculating a slip normalization quantity of the main reflection direction and a slip normalization quantity of a vertical reflection direction, outputting bubble credibility through logistic regression, dividing a bubble region and a reflection artifact region, generating a bubble mask to update a reference texture, updating a highlight compensation model through the reflection artifact region, and performing consistency verification on a compensation result. And outputting a bubble recognition result.

The embodiment of the invention provides a time sequence signal fault prediction method and device, equipment and a storage medium, and relates to the technical field of anomaly prediction. The method comprises the following steps: generating a simulation vibration signal based on the vibration response of a gear, and inputting the simulation vibration signal into a gear fault prediction model for signal prediction to obtain a predicted vibration signal; physical information loss is calculated based on predicted displacement, predicted speed and predicted acceleration in the predicted vibration signals, predicted error loss is calculated based on the simulated vibration signals and the predicted vibration signals, a total loss value is calculated according to the predicted error loss and the physical information loss, parameter adjustment is carried out, and a trained gear fault prediction model is obtained. And inputting a to-be-predicted signal into the trained gear fault prediction model for signal prediction to obtain a reference vibration signal, and obtaining a time sequence abnormal result based on the reference vibration signal and the to-be-predicted signal. The prediction process fits the actual characteristics of the signal and the physical law of gear vibration, and the recognition accuracy and stability are improved.

The invention provides a first arrival identification method based on spatial attribute change. The first arrival identification method based on spatial attribute change comprises the following steps: step 1, inputting a seismic trace set of which first arrival needs to be picked up; step 2, inter-track time difference elimination based on a small smooth reference surface is carried out; step 3, reconstructing the gather by using the spatial correlation of the gather by using a PCA method, extracting a wavelet template on the PCA reconstructed gather, and calculating the spatial attribute of each point of the gather by using a correlation coefficient method; and 4, identifying the first-motion wave according to the spatial attribute change. According to the first arrival recognition method based on the spatial attribute change, the problem that first arrival wave recognition is inaccurate under the complex earth surface condition is solved, statistical correlation in the first arrival signal space is fully utilized, element-mode seismic wavelets of seismic exploration are combined, and stable first arrival wave recognition under the complex earth surface condition is achieved.

This application provides a method and apparatus for identifying HPV integration into newly formed chromatin loops, relating to the field of bioinformatics. The method includes: acquiring interaction sequencing read data and determining HPV reference sequences and host reference sequences; identifying HPV integration breakpoints based on chimeric read characteristics, split read characteristics, and abnormal pairing end characteristics; constructing fusion contigs based on breakpoint directions to obtain a breakpoint-aware extended reference set; comparing the interaction sequencing read data with the breakpoint-aware extended reference set, and obtaining high-confidence cross-genomic anchor pairs based on breakpoint-aware alignment constraints; clustering the anchor pairs and screening target HPV-loops; calculating the newborn score of the target HPV-loop and outputting the chromatin loop identification result. This application solves the problem of low accuracy and identification bias that often occurs in traditional interaction localization systems during the identification of newly formed chromatin loops.

The invention discloses a tactile stick obstacle intelligent identification method based on real-time image analysis, and relates to the technical field of barrier-free travel assistance, and the method comprises the following steps: obtaining front road surface image data collected by an image collection device arranged at the front end of a tactile stick in real time, carrying out the distortion removal, brightness normalization and scale standardization processing of the image data, and obtaining the image data; and associating the image data with corresponding image acquisition time information to generate normalized pavement image data. According to the method, the image data of the front road surface is normalized and multi-scale road surface feature analysis is introduced, so that pixel-level accurate identification and type judgment of the blind sidewalk area are realized, and the blind sidewalk can still be stably identified under the conditions of complex illumination, material change and environmental interference; the accuracy and robustness of blind sidewalk detection are remarkably improved, the blind sidewalk continuity score and the blind sidewalk prediction area are constructed based on the blind sidewalk center line data and the advancing direction vector data, and time dimension modeling of the blind sidewalk continuous state is achieved.

The application provides a harmful algal bloom area identification method and device based on remote sensing data, and belongs to the field of marine remote sensing and environmental monitoring. The method comprises the following steps: determining a plurality of sensitive bands of harmful algal blooms; acquiring remote sensing data of a to-be-identified area in each sensitive band; determining an identification interference factor of the to-be-identified area; combining each sensitive band according to the relationship between the identification interference factor and each sensitive band; calculating an identification interference factor index corresponding to each combination; correcting the remote sensing data corresponding to the first sensitive band combination and the second sensitive band combination according to the identification interference factor index; calculating a harmful algal bloom index according to the remote sensing data corresponding to the third sensitive band combination after correction; and judging whether the to-be-identified area is a harmful algal bloom area according to the harmful algal bloom index and a dynamic threshold corresponding to the to-be-identified area. The harmful algal bloom area identification method and device based on remote sensing data provided by the application can realize high-precision and high-stability identification of harmful algal bloom areas.

The invention discloses an electric energy meter live and null line positive and negative connection detection system and method based on a PCB capacitor, and the system employs a coupling capacitor formed by reference ground laying copper and the ground to construct a signal backflow path, and outputs high and low level signals in combination with the conduction state change of a triode of a signal processing circuit. Signal isolation transmission is realized through an optocoupler device of the isolation circuit, and finally the MCU judges the connection state of the live line and the null line, thereby effectively detecting the reverse connection state of the live line and the null line of the electric energy meter, avoiding the fire hazard of electric appliances caused by manual missing detection, improving the safety performance of the circuit, and greatly facilitating the application and maintenance of the intelligent electric meter.

The application relates to a plant growth stage monitoring method and device and a computer device. The method comprises the following steps: acquiring multi-modal data of a target plant, wherein the multi-modal data features comprise environmental data features and image data features; inputting the image data features into a plant category recognition model to obtain a target type of the target plant; dynamically adjusting the image data features based on the environmental data features, and generating fusion features of the environmental data features and the dynamically adjusted image data features; and inputting the fusion features into a target growth stage classification network corresponding to the target type, and outputting a current growth stage classification result of the target plant.

A human posture anomaly detection method based on multi-branch spatio-temporal modeling, the method first designs a multi-branch local-global feature fusion module, by constructing a local feature extraction branch for each sub-region of the body, the limbs, torso and other parts of the human body are modeled in detail posture feature, and the dynamic fusion of local and global features is realized by combining the self-adaptive weighting mechanism, forming a multi-level feature representation system, thereby simultaneously improving the detection accuracy of sudden local anomalies and global posture anomalies; At the same time, a time sequence attention enhancement module is proposed, which adopts a time sequence attention mechanism that integrates bidirectional recurrent units, and by parallelly integrating forward and backward time sequence context information, the modeling capability of the model for long time span posture dynamic changes is strengthened, and the recognition robustness for persistent abnormal postures is significantly improved. The method shows superior performance in local sudden anomaly and persistent anomaly detection tasks.

This invention discloses a rapid bridge damping ratio identification method based on a stationary inspection vehicle, belonging to the field of bridge structure vibration testing technology. The method includes: parking the inspection vehicle at the mid-span of the bridge under inspection; exciting the bridge vibration using a passing vehicle and then driving away; collecting the acceleration response of the inspection vehicle in the attenuation segment; establishing a dynamic model of the attenuation segment of the vehicle-bridge coupled system; constructing a two-degree-of-freedom joint state model, internalizing the vehicle-bridge interaction force as a linear combination of system state variables; extracting the effective attenuation segment through signal preprocessing and introducing a dual anti-noise mechanism of adaptive signal-to-noise ratio weights and Huber weights; and employing a profiled two-stage parameter identification strategy, first locking the frequency and modal mass, and then finely scanning and identifying the bridge's first-order damping ratio. This invention requires only a single inspection vehicle and an acceleration sensor, eliminating the need to deploy sensors on the bridge structure. It has strong anti-interference capabilities, high identification accuracy, and is suitable for rapid inspection of small-to-medium span bridges.

The invention relates to an on-tree palm fruit maturity detection method and device based on target segmentation and colorimetric analysis, equipment and a medium, and the method comprises the steps: updating a backbone network of a first maturity detection model to a StarNet framework, updating all C2F modules in the backbone network and a neck network to C2F-Star modules, a detection head network is updated to be a lightweight shared convolution detection head, an original CIOU loss function is updated to be a DIOU loss function to construct a second maturity detection model, and a C2F-Star module is used for updating a Bottleneck structure in an original C2F module to be a Star Block module; and inputting a to-be-detected palm fruit image into the second maturity detection model, outputting a polygon mask corresponding to each target palm fruit, counting a red pixel proportion in a polygon mask area, and determining a maturity category corresponding to each target palm fruit according to the red pixel proportion. According to the method, the target can be stably identified even in the scenes of interlaced branches and overlapped leaves, and the problem of detection failure caused by shielding is solved.

This invention discloses a robot control method, an automatic cleaning device, and a storage medium. In this embodiment, after acquiring multiple frames of point cloud data collected by a line laser sensor, the automatic cleaning device first determines the object category corresponding to each point cloud dataset based on the point cloud features of the point cloud dataset within a preset distance range in each frame. Then, based on the object category of each point cloud dataset in each frame, it determines the target object category corresponding to each environmental object. Finally, it controls the automatic cleaning device to avoid obstacles based on the target object category. This invention, by determining the object category of the point cloud dataset within a preset distance range in each frame and determining the target object category based on the object category of each point cloud dataset in all frames, avoids the situation of identifying obstacles based solely on a single frame, thus improving the accuracy of obstacle identification.

The invention provides a user state monitoring method and device. The method comprises the following steps: collecting a scene image of a target scene through a plurality of cameras; for each frame of scene image collected by each camera, judging whether a face of a target user and / or facial key points of the face are detected in the current frame of scene image collected by the camera; if yes, determining that the target user is in an on-duty state; extracting a region of interest corresponding to the target eye based on the current frame scene image; determining whether the target user is in a long eye closing state based on the region of interest; if the target user is not in the long eye closing state, selecting a specified key point from the face key points; and determining whether the target user is in a focused state based on the specified key point. According to the method, the on-duty state, the long eye closing state and the concentration state of the target user can be comprehensively judged under the constraints of visible light, no wearing, no complex calibration and the like, so that the state of the monitored user can be stably recognized in the natural environment.

The application discloses a kind of fast identification method and system of high-risk local power grid facing extreme disaster driving, method includes: the power grid topology of preset voltage level is as identification reference layer, the power grid in reference layer is carried out connected subgraph identification;When the number of external tie line of connected subgraph satisfies first preset constraint condition, the connected subgraph is determined as preliminary candidate local power grid area;With preliminary candidate local power grid area as reference, respectively to upper level voltage level and lower level voltage level of preset voltage level is topologically traversed, the total connection channel number of preliminary candidate local power grid area between corresponding voltage level and main grid is determined;When total connection channel number satisfies second preset constraint condition, then it is judged as easy-to-isolate local power grid;Risk quantification evaluation index system of easy-to-isolate local power grid under extreme disaster is constructed, and the comprehensive risk value of easy-to-isolate local power grid area is calculated based on risk quantification evaluation index system.

The application provides a dual-purpose item-based controlled article identification method, device, equipment, medium and product. It relates to the technical field of artificial intelligence. The method comprises: obtaining an article to be identified; determining whether the article to be identified belongs to a non-dual-purpose item category based on a preset category library; in the case that the article to be identified belongs to a dual-purpose item category, generating a safety score of the article to be identified according to a preset keyword library and a preset reasoning logic rule; in response to the safety score being greater than or equal to a preset safety threshold, generating a relevance score of the article to be identified based on a vector retrieval result and a preset scoring rule; in the case that the relevance score is equal to a preset relevance threshold, identifying the article to be identified using a preset multi-dimensional judgment strategy to generate an identification result. The method of the application significantly improves the review efficiency, reduces the misjudgment rate, and meets the needs of intelligent, explainability and scalability in the export business scenario while ensuring compliance.