40results about How to "Increase the difference" patented technology

The invention relates to an icing thickness detection method and device, relates to the technical field of aircraft icing thickness data processing, solves the problem that the thickness of a thin ice layer is difficult to effectively identify, and realizes high-sensitivity identification in the initial stage of icing and high-precision quantitative detection in the whole process of icing. The method comprises the following steps: receiving a plurality of first reflection signals after ultrasonic waves penetrate through an aluminum layer and an ice layer and second reflection signals after the ultrasonic waves penetrate through the aluminum layer; removing overlapped signals among the plurality of first reflection signals to obtain a plurality of third reflection signals which are independent from each other; and calculating the second reflection signal and the third reflection signal to obtain the icing thickness.

The invention relates to the field of artificial intelligence and finance, and discloses an unsupervised video abnormal behavior detection method and device, electronic equipment and a storage medium, which can be used for carrying out abnormal behavior detection on a monitoring video of the finance industry, and the method comprises the following steps: obtaining an interval frame of a to-be-detected monitoring video, carrying out optical flow estimation on the interval frame, and obtaining an optical flow estimation result of the to-be-detected monitoring video; obtaining an optical flow estimation frame; according to the interval frames and the optical flow estimation frame, monitoring video frames between the interval frames are predicted, and a predicted intermediate frame is obtained; respectively carrying out feature extraction on the predicted intermediate frame and a real intermediate frame between the interval frames to obtain predicted frame features and real frame features; according to the predicted frame features and the real frame features, calculating the similarity between the predicted intermediate frame and the real intermediate frame by using a preset function; and when the similarity is smaller than a preset threshold value, determining that an abnormal behavior exists in the to-be-detected monitoring video. According to the invention, the accuracy of unsupervised video abnormal behavior detection can be improved.

The invention discloses an infrared weak and small target detection method based on multi-frame interactive perception, and the method comprises the steps: 1, carrying out the preprocessing of an infrared sequence image, and avoiding the overfitting in a training process; 2, constructing a time sequence feature extraction module for continuous multi-frame images; 3, constructing a feature interaction perception module with a U-shaped structure; 4, establishing a loss function between a result obtained after the infrared image of the current frame passes through the prediction module and the real binary image; and 5, training the network by using an infrared image, and detecting an infrared weak and small target by using the trained model. According to the method, the spatial-temporal information of the target is obtained by using multiple frames of images, and cross-level information fusion is realized, so that the accuracy of infrared weak and small target detection can be improved, and the capturing capability of a fast moving small target is particularly enhanced.

The invention discloses an adversarial sample detection method and system and electronic equipment, and the method comprises the steps: inputting to-be-detected target images into a classifier, and obtaining classification labels corresponding to the to-be-detected target images; inputting the target images to be detected and the classification labels corresponding to the target images to be detected into a trained sequence condition variation auto-encoder to obtain reconstructed images of the target images to be detected output by the sequence condition variation auto-encoder; and determining whether the to-be-detected target image is an adversarial sample or not according to the reconstructed image. According to the method provided by the embodiment of the invention, the classification label and the to-be-detected target image are synchronously input into the trained sequence condition variation auto-encoder to improve the difference between the adversarial sample and the original sample, so that the adversarial sample can be detected more accurately, the accuracy rate, the precision rate and the recall rate of detection are improved, and the method has wide applicability.

The invention discloses an industrial control time series data anomaly detection method based on deep learning, and relates to the technical field of time series data anomaly detection of an industrial control system.The method comprises the steps that an auto-encoder and a memory enhancement auto-encoder are fused into a double-branch memory enhancement auto-encoder network model; a cutting normalization layer is introduced into an input layer of a network model to cut and screen values exceeding the boundary of the maximum value or the minimum value in a set range, and cosine similarity in a memory module of the network model is replaced by mean square error to expand difference between abnormal data, so that similarity between time series data is obtained, and the time series data is obtained. And meanwhile, a regular term is added into a loss function of an output layer of the network model to measure the similarity between time sequence data and screen abnormal data, so that abnormal information in a complex time sequence can be recognized more clearly, and a detection result closer to an actual condition can be obtained.

The invention relates to the technical field of image data processing, in particular to a fabric defect detection method for producing protective clothing, and the method comprises the steps: obtaining a grey-scale map of a fabric for producing the protective clothing, extracting the texture features of the grey-scale map through a local binary pattern algorithm, and obtaining the feature value of each pixel point in the grey-scale map; the method comprises the following steps: continuously selecting a plurality of pixel points in a current row by taking any pixel point as a starting point to obtain a continuous pixel set of any pixel point, calculating the feature prominence degree of any pixel point in any row, and correcting the gray value of the pixel point in the current row by utilizing the index difference of the pixel point with the maximum feature prominence degree in the current row and the adjacent row to obtain a gray value of the pixel point in the current row. And obtaining the similarity between the corrected grey-scale map and a pre-stored intact grey-scale map, calculating the defect degree of the fabric, and carrying out defect detection on the fabric based on a comparison result of the defect degree and a preset threshold value. According to the invention, the fabric defect detection accuracy can be improved.

The application provides a pre-processing method of to-be-compensated data, a picture compensation method and related equipment. The pre-processing method of to-be-compensated data comprises the following steps: acquiring first luminance data of a plurality of sub-pixels in a display panel when a to-be-compensated picture with a display gray scale less than or equal to a preset gray scale; determining first part sub-pixels and second part sub-pixels from the plurality of sub-pixels, adjusting the first luminance data of the first part sub-pixels to second luminance data, and adjusting the first luminance data of the second part sub-pixels to third luminance data; wherein the second luminance data is greater than the first luminance data, and the third luminance data is less than the first luminance data. By adopting the scheme of the application, more reasonable compensation data can be obtained when Mura compensation is performed, and a better compensation effect can be achieved.

An image processing apparatus and air conditioner include: an acquisition unit configured to acquire low-light and infrared images; and a processing unit comprising: a low-light image enhancement algorithm model, which includes: a gradient reconstruction module comprising: a frequency domain processing branch configured to transform the gradient map of the low-light image from the spatial domain to the frequency domain to extract the phase spectrum and amplitude spectrum; and a spatial domain processing branch configured to fuse the gradient map of the low-light image with the infrared image in the channel dimension; the gradient reconstruction module adds the residual of the spatial domain processing branch to the amplitude spectrum to obtain an enhanced amplitude spectrum; an amplitude-enhanced gradient map of the low-light image is obtained through an inverse Fourier transform between the enhanced amplitude spectrum and the original phase spectrum; and an enhanced visible light image is generated through an encoder-decoder architecture based on the gradient features of the amplitude-enhanced gradient map of the low-light image and the image features of the low-light image; this application has high accuracy and is suitable for use in smart home appliances.

This invention belongs to the field of multimedia analysis technology and discloses a multimodal media tampering detection method, system, device, and medium based on multi-view contrastive learning. The method includes: acquiring a training dataset, which includes training image-training text pairs and corresponding tampering category labels; introducing a cross-encoder based on a visual-language model, setting several multilayer perceptron head structures, and designing three contrastive learning methods: noise enhancement, prototype-based, and multi-label tampering classification, to obtain an initial multi-view contrastive learning framework; training the initial multi-view contrastive learning framework based on the training dataset to obtain a trained multi-view contrastive learning framework; and performing a tampering detection task on the image-text pair data to be detected based on the trained multi-view contrastive learning framework. The technical solution of this invention can improve the accuracy and robustness of multi-label classification.

The invention discloses a control method, device and equipment of a rod mill feeding system and a medium, and belongs to the technical field of metallurgy. The feeding system of the rod mill comprises a fan-shaped valve, the fan-shaped valve is used for adjusting the feeding amount of the rod mill, and the method comprises the steps that the actual feeding amount and the target feeding amount of the rod mill are obtained; if the difference value obtained by subtracting the target feeding amount from the actual feeding amount is larger than a preset first difference value threshold value, the fan-shaped valve is controlled to execute the closing action of a preset first duration; if the difference value is smaller than a preset second difference value threshold value, the fan-shaped valve is controlled to execute the opening action of a preset second duration; wherein the first difference threshold value is larger than 0, and the second difference threshold value is smaller than 0. According to the method, the opening degree of the fan-shaped valve can be finely adjusted every time, the precision of adjustment every time is guaranteed, then the final deviation is small or even zero, and the rod mill can be kept in an efficient working state.

The application belongs to the technical field of image detection, and particularly relates to a visual detection method for intelligent segmentation of components on a circuit board. The method comprises the following steps: collecting original images of the circuit board and performing manual labeling and correction to construct a component segmentation training data set; training a semantic segmentation model based on a deep convolutional neural network, adopting a cross-entropy and Dice mixed loss and dynamically adjusting the class weight to alleviate the class imbalance; inputting a to-be-detected image into the model after preprocessing to obtain a semantic segmentation mask; performing morphological post-processing on the segmentation mask to eliminate false areas and regularize the contour; and finally outputting the class, contour and position information of each component. Through standardized data set construction, preprocessing enhancement, mixed loss function and post-processing optimization, the application effectively improves the precision and stability of component segmentation under complex circuit board images, and especially improves the segmentation effect of low-frequency key components.

The present application discloses a display panel and a display device. The display panel includes a fingerprint recognition area. The display panel includes a cover plate and an optically transparent layer. The optically transparent layer at least covers a portion of the cover plate located in the fingerprint recognition area. A refractive index of the optically transparent layer is greater than a refractive index of air and less than a refractive index of a finger. A thickness of the optically transparent layer is an even multiple of a half-wavelength of a fingerprint recognition light ray. Half-wave loss and interference of light are used to improve fingerprint recognition performance.

The present application discloses a method for detecting abnormal facilities based on deep learning, a device for detecting abnormal facilities based on deep learning, an electronic device and a computer-readable storage medium. The detection method is implemented based on a pre-trained facility detection model. The facility detection model sets an MSDSF module after the feature extraction module corresponding to the last scale of the backbone network, and fuses local target information and non-local background information at different scales, thereby enriching the features of small targets and enhancing the distinction between small targets and backgrounds. An ARFA submodule is set in the feature extraction module corresponding to each scale, and different regions of the input features at a single scale are fused based on different weights; the attention level of different regions of the input features is dynamically adjusted based on the weight learning mechanism, so that the input features can be processed more finely and adaptively, the expression ability of key features is improved, and the detection accuracy of the facility detection model for different abnormal urban facilities under complex and changing backgrounds is improved.

The present invention provides a vertical semiconductor structure of an integrated sampling structure and a manufacturing method therefor. The vertical semiconductor structure of an integrated sampling structure comprises: a vertical semiconductor structure primitive cell, a sampling primitive cell, a control electrode, a first electrode, a second electrode, and a sampling electrode. According to embodiments of the present invention, first, the sampling electrode can sample a voltage difference between the first electrode and the second electrode in real time; a PN junction is formed between a first / second P-type diffusion region and a second N-type base region of the sampling primitive cell, and the PN junction forms a barrier for blocking electron emission of a sampling electrode terminal. Therefore, a voltage signal of the sampling electrode is input to a protection circuit, and the protection circuit can safely and quickly detect whether the vertical semiconductor structure is in a desaturation state while determining that the vertical semiconductor structure primitive cell is in an open state. Secondly, a sampling resistor is connected in series between the sampling electrode and the first electrode, the positive voltage of the sampling electrode increases the height of the barrier, and the stable operation of the sampling primitive cell can be ensured.

The invention provides a control method based on an intelligent pressure switch, the intelligent pressure switch and a related device, and the method comprises the steps: controlling an air compressor to work when the current temperature is greater than the target temperature; the pressure intensity value of the intelligent refrigerator is obtained every first time interval, multiple pressure intensity values are obtained, and each pressure intensity value corresponds to one sampling moment; straight line fitting is carried out according to the multiple pressure values and the corresponding sampling moments, a target fitting straight line is obtained, the transverse axis of the target fitting straight line is time, and the longitudinal axis of the target fitting straight line is the pressure value; determining a prediction moment corresponding to the first pressure intensity value through the target fitting straight line; obtaining a target slope of the target fitting straight line; acquiring a first moment before the prediction moment according to the target slope; the pressure value of the intelligent refrigerator is obtained every second time interval, and a second pressure value is obtained; and when the absolute value of the difference value between the second pressure intensity value and the first pressure intensity value is smaller than a preset threshold value, the air compressor is controlled to stop working through the intelligent pressure switch.

This application discloses a semiconductor device, a memory, a printhead, and a consumable cartridge. The semiconductor device includes a semiconductor substrate and a transistor, an antifuse element, a resistor element, a first terminal, and a second terminal disposed on the semiconductor substrate. The transistor is connected to the first terminal. The antifuse element is connected between the transistor and the second terminal via a first wiring and a third wiring. The resistor element is connected in parallel with the antifuse element between the transistor and the second terminal via a second wiring and a fourth wiring. The first and third wirings are both referred to as antifuse wirings, and the second and fourth wirings are both referred to as resistor wirings. At least one antifuse wiring is arranged in a first stacked structure, and / or, at least one resistor wiring and its corresponding antifuse wiring are arranged in a second stacked structure. This application enables a larger wiring space for the resistor wirings, thereby increasing the resistance of the resistor wirings and improving data reading accuracy.

The utility model provides a kitchen waste back-mixing water washing device with a material blocking early warning function. The kitchen waste back-mixing water washing device comprises a flow meter arranged on a conveying pipeline and used for monitoring the flow of oil and water in the conveying pipeline; a first weight sensor, a second weight sensor and a supervision analysis unit connected with the flow meter, the first weight sensor and the second weight sensor are arranged at the joint of the first supporting seat and the back-mixing washing device and the joint of the second supporting seat and the back-mixing washing device respectively. The flow meter on the conveying pipeline is used for monitoring the flow of oil and water in the conveying pipeline; a first weight sensor and a second weight sensor are arranged at the joint of the first supporting seat and the back-mixing washing device and the joint of the second supporting seat and the back-mixing washing device respectively, the sensors are connected with a flow meter and a supervision and analysis unit, and when a filter screen is blocked, an alarm is given through signals transmitted by the sensors and the flow meter. The problem that in the prior art, blockage of a filter screen caused by flaky residues retained in a back-mixing washing device cannot be found in time is solved.

The invention discloses an OCTA retinal vessel segmentation method based on Gabor modulation coding improved U-Net, and the method comprises the steps: carrying out the modeling of a primary visual cortex simple cell through employing a Gabor filter, forming a Gabor convolution pathway fusing edge features, assisting the convolution pathway to complete the feature extraction, and carrying out the segmentation of a retinal vessel on the same Gabor scale. The method comprises the following steps: firstly, performing contour information extraction on a feature map by using Gabor convolution kernels in four directions to obtain texture features of different levels so as to compensate for edge features of a blood vessel joint and a fovea centrally area; carrying out integration processing on the features, namely stacking according to dimensions; repeatedly extracting the integrated features, namely concentrating the edge features by using 1 * 1 convolution; and finally, realizing mutual association with a common convolutional coding path, namely realizing the purpose of enhancing mutual association between feature maps with different scales through maximum pooling operation, supplementing edge features and amplifying the difference between blood vessels and backgrounds.

The present invention relates to a drainage pipe defect recognition model training method and a drainage pipe defect recognition method. The technical solution of the present invention is a drainage pipe defect recognition model training method, which is characterized by: obtaining a training sample, the training sample has a drainage pipe image and a defect type corresponding to the image, and there are n defect types in the training sample; preprocessing the drainage pipe image, including graying the drainage pipe image; extracting image features corresponding to each defect type from the preprocessed drainage pipe image corresponding to each defect type; training the initial binary classification model based on the image features corresponding to the i-th defect type and the image features corresponding to the j-th defect type among the n defect types, and obtaining defect recognition models for the i-th and j-th defect types, and obtaining defect recognition models in total. The present invention is applicable to the field of image recognition technology.

The invention discloses a flow battery energy storage balancing system, and belongs to the technical field of electrochemical energy storage. The flow battery energy storage balancing system comprises a main system used for providing power supply voltage and energy storage capacity, the main system is connected with a balancing system used for balancing the capacity and the voltage of the main system, and the flow battery energy storage balancing system further comprises an electrical control system connected with the main system and the balancing system. The invention also discloses a flow battery energy storage balancing method. The energy storage balancing system of the flow battery is suitable for engineering application, is simple and convenient to operate, and can actively balance the inconsistency of charge states among multiple sets of process systems of the flow battery and reduce the difference among the systems, so that the utilization rate of the electrolyte is improved, the system efficiency is improved, and long-term operation of the battery and efficient utilization of the electrolyte can be realized.

The present application relates to the technical field of oil drilling risk monitoring, well control safety early warning and artificial intelligence identification, and specifically discloses an early overflow intelligent identification method and system based on knowledge guidance and residual error enhancement, which comprises the following steps: first, obtaining original drilling monitoring data and completing preprocessing; then, performing feature enhancement through a knowledge-guided self-attention module, obtaining a knowledge bias term by combining a standard self-attention score with a trend consistency index, a pool volume correction factor and an inflection point index, and generating an enhanced feature matrix; subsequently, utilizing a prototype learning residual error enhancement module to construct a normal working condition feature prototype, matching an optimal prototype and calculating a feature difference value to obtain a residual error feature matrix; finally, inputting the residual error feature matrix into a classification layer to output an early overflow prediction probability; the present application can focus on weak but mechanism-compliant abnormal periods and reduce the overlap interference between normal working conditions and early overflows, thereby improving the identification accuracy and early warning timeliness.

The invention discloses a radar target high-resolution one-dimensional range profile recognition method, and relates to the field of radar signal processing. Comprising the following steps: S1, constructing a target high-resolution one-dimensional range profile training sample set X, a label matrix H and a label block diagonal matrix Q; s2, extracting features from the training sample set X and the tag block diagonal matrix Q, and optimizing to obtain a feature dictionary matrix D and a feature representation matrix Z; s3, constructing a linear classifier C by using the feature representation matrix Z and the label matrix H; s4, constructing a feature representation matrix Zt of the test sample set Xt according to the feature dictionary matrix D; and S5, performing target classification by using the feature representation matrix Zt of the test sample set Xt and the linear classifier C. The problem that a traditional target recognition method is poor in training effect is solved.

The application is suitable for the field of image processing, and provides a no-reference color image quality evaluation method based on color and structure distortion, comprising the following steps: performing a blocking operation on a color image to form a color distortion image block data set, and forming a training set and a test set; a group of two-dimensional Scharr detection operators are constructed, and convolution processing is performed on the input image block by using the two-dimensional Scharr detection operators to obtain gradient feature information of the distortion image block; the obtained image block data set is converted into an HSV color space; a no-reference image quality evaluation convolutional neural network model based on color and structure distortion is constructed according to the existing image block data set; and the divided training set is input into the constructed no-reference image quality evaluation convolutional neural network model to obtain a trained no-reference image quality evaluation model. The application has reasonable design, the simulated quality score and the real quality score have strong correlation and small error, and is worth promoting.

The invention is suitable for the technical field of micro projection display, and discloses an optical machine, an optical machine control method and a near-to-eye display device.The optical machine comprises an illumination system, an imaging module and an optical relay system, and the illumination system comprises a plurality of illumination areas arranged on a two-dimensional surface in an array mode; the imaging module comprises a beam splitter prism and a micro display panel, the lighting system and the micro display panel are symmetrically arranged around the beam splitter prism, and a plurality of image areas are formed on the micro display panel; the optical relay system comprises a first lens and a second lens, the first lens is located between the illumination system and the beam splitter prism, the second lens is located between the micro display panel and the beam splitter prism, and the first lens and the second lens are symmetrically arranged around the beam splitter prism, so that the plurality of image areas are in one-to-one correspondence with the plurality of illumination areas; and each illumination area is used for emitting illumination light when the corresponding image area displays a picture so as to illuminate the corresponding image area. The embodiment can improve the contrast of an optical machine and reduce the power consumption of the optical machine.

Embodiments of this application disclose an RC oscillator. The RC oscillator may amplify a difference between a first voltage and a second voltage by using a first amplifier and a second amplifier. The first amplifier may include a first amplification circuit and a second amplification circuit. The first amplification circuit and the second amplification circuit may share a same voltage-current conversion circuit. The RC oscillator disclosed in the embodiments of this application can not only avoid noise introduced by the first amplifier, but also reduce internal noise of the RC oscillator. This reduces a jitter of a clock signal.

The application discloses a kind of based on optimization Makishima-Mackenzie formula's fiber glass modulus prediction method.The application first establishes an oxide dissociation energy optimization model based on differentiating weight coefficient, by introducing independent optimization coefficient, the difference of different oxides in glass network structure to modulus contribution can be optimized, compared with the prediction error of traditional MM formula on high modulus glass modulus is greatly reduced;At the same time, the clear physical image based on dissociation energy and component is retained, the internal relationship between glass microstructure and macro modulus can be revealed, and the "black box" problem of pure machine learning method is avoided.In addition, the modulus prediction can be completed only by oxide dissociation energy, molar volume and other basic parameters easy to obtain, the calculation efficiency is improved by more than 100 times, and high-performance computing resources are not needed, which significantly reduces the time cost and technical threshold of high-performance glass fiber research and development.

A gas inlet assembly of a process chamber of a semiconductor process apparatus, a gas inlet device, and a semiconductor processing apparatus are provided. The gas inlet assembly is configured to transport the process gas to the gas inlet pipeline that communicates with the process chamber. The gas inlet assembly includes a plurality of mixing chambers arranged sequentially along the gas inlet direction. Any two neighboring mixing chambers communicate with each other. The gas inlet assembly includes a gas inlet connector. The gas inlet connector communicates with the mixing chamber at the most upstream in the gas inlet direction. The mixing chamber at the most downstream in the gas inlet direction communicates with the gas inlet pipeline. The gas outlet direction of the gas inlet assembly is the same as the gas inlet direction. The gas inlet assembly, the gas inlet device, and the semiconductor processing apparatus are configured to sufficiently mix and homogenize the process gas. Thus, the distribution uniformity of the components of the process gas can be improved, and the differences in the gas flow speeds between different areas of the gas inlet pipeline can be reduced. Thus, the process stability can be improved.

The invention provides a video processing method and apparatus, a computer device and a storage medium. The method comprises the steps of obtaining a to-be-processed video; determining a to-be-processed video frame in the to-be-processed video, and determining a main body area and a non-main body area of the to-be-processed video frame; respectively carrying out stylization processing on the main body area and the non-main body area of the to-be-processed video frame to obtain a stylized video frame; wherein the stylization parameters of the main body area and the non-main body area are different when the stylization processing is carried out on the main body area and the non-main body area; and generating a stylized video corresponding to the to-be-processed video based on the stylized video frame. Through the method, the video processing speed can be increased, and the video stylization processing quality can be improved.

A modular ultrasonic flow meter comprises: a connected substantially waterproof meter housing; a measurement printed circuit board including a measurement circuit in communication with one or more ultrasonic transducers disposed in the meter housing for transmitting and receiving ultrasonic signals; and a self-contained power supply. The ultrasonic flow meter further comprises a control module comprising: a module housing connected to a main printed circuit board including a central processing unit, a memory circuit and a communication circuit; and power supply and communication connections are provided between the main printed circuit board and the measurement printed circuit board of the control module, whereby the self-contained power supply can power the main printed circuit board and the measurement printed circuit board.

The present application discloses a method, system, electronic device and medium for establishing an entity recognition model. The method for establishing an entity recognition model includes: performing language masking on a first entity in an input representation vector, obtaining a first language masked text, selecting non-entity words from the word segmentation result of the input representation vector and performing language masking on the words, and obtaining a second language masked text; inputting the first language masked text and the second language masked text into the encoding layer of an initial Transformer model, and obtaining a first hidden layer vector and a third hidden layer vector from the output result; calculating a first training matrix and the first hidden layer vector, and obtaining a probability distribution vector, and then calculating a second training matrix and the third hidden layer vector, and obtaining a classification probability distribution vector; performing back propagation calculation on the initial Transformer model using the probability distribution vector and the classification probability distribution vector, and updating the encoding layer to obtain a final Transformer model.