4results about How to "Fix false positives" patented technology

This invention provides a drop-proof detection unit and a photovoltaic panel cleaning robot. The drop-proof detection unit includes a substrate, an ultrasonic sensor, a first laser sensor, and a controller. The ultrasonic sensor and the first laser sensor are both mounted on the substrate, and are spaced apart in a first direction. The controller is mounted on the substrate and electrically connected to the ultrasonic sensor and the first laser sensor. This invention's drop-proof detection unit, by simultaneously utilizing both ultrasonic and laser sensors, achieves comprehensive detection of the edges and gaps of photovoltaic panels, effectively avoiding false alarms that may occur with a single sensor. It also reduces work interruptions caused by signal anomalies. This allows the photovoltaic panel cleaning robot to not only avoid falls but also clean multiple photovoltaic panels across gaps, improving cleaning efficiency.

This invention discloses a lightning potential forecasting method integrating cloud microphysical conditions, belonging to the field of lightning potential forecasting technology. It involves acquiring numerical weather prediction data for a target area, constructing a cloud microphysical dynamic-water vapor coupling index for the target area, and determining the meteorological forecast results for the target area based on the cloud microphysical dynamic-water vapor coupling index. Based on the meteorological forecast results, convection parameters for the target area are extracted, and convection information for the target area is obtained through comprehensive analysis. Based on the convection information, mixed-phase cloud microphysical parameters for the target area are extracted. Based on the mixed-phase cloud microphysical parameters, collision intensity characterization parameters for the target area are determined. The collision intensity characterization parameters are then corrected for charge separation capability based on numerical weather prediction data to obtain an effective collision intensity index for the target area. Based on the effective collision intensity index, the lightning potential index for the target area is determined, and the lightning potential information for the target area is output, thus improving the accuracy of lightning potential forecasting.

The application discloses a kind of based on time-frequency phase feature rolling bearing fault diagnosis and physical positioning method, this method includes: obtaining the vibration time series signal at two different circumferential positions of the monitored bearing seat, and carry out time-domain impact feature enhancement processing;Two enhanced impact signals are analyzed by cross spectrum, and extract coherence function spectrum and cross-phase spectrum;According to the energy peak value of impact signal, coherence function spectrum and cross-phase spectrum set fault diagnosis condition, find out the real fault characteristic frequency that meets the condition, compared with various fault characteristic frequency theoretical value, determine the fault type of rolling bearing;According to the geometric relationship of fault point and two vibration collection points on bearing seat, establish circumferential angle solution model, the stable phase difference and collection point position are substituted into model, and the circumferential position of fault point on bearing seat is calculated.The method can distinguish real bearing fault from external random disturbance from physical principle, greatly improve the reliability and accuracy of diagnosis.

This invention provides a method for segmenting and classifying real-world photo instances, relating to the field of computer vision technology. The method includes: constructing a real-world verification dataset based on the spatial distribution of the vector data to be verified; establishing an interpretation feature library containing texture and geometric features; constructing a cascaded network architecture consisting of an object detection subnetwork and a segmentation subnetwork, using bounding boxes to guide segmentation operations, freezing the segmentation encoder parameters during training, embedding interpretation feature data into the network backbone to obtain a lightweight cascaded instance segmentation model; projecting the vector data onto the image coordinate system using a spatial mapping transformation algorithm to obtain a theoretical projection area; comparing the spatial overlap and attribute consistency between the model segmentation results and the theoretical projection area to generate classification verification correction information. This invention solves the problem of the lack of an effective correlation verification mechanism between heterogeneous data, achieving automation and efficiency in ground feature verification.