329results about How to "Guaranteed detection efficiency" patented technology

The invention discloses a mobile ad hoc network intrusion detection method and device based on deep learning, relating to the field of wireless network safety. The device comprises a data acquisition module, a data fusion module, a preprocessing module, a storage module, an intrusion detection module and a response warning module. After fusion and redundancy elimination of captured wireless data packages, network behavior features are extracted and stored; after deep learning of the network behavior features, a deep neural network model expressing network behaviors is established; and to-be-detected network data is input into the deep neural network model, after intrusion is judged and recognized, response and warning are performed. According to the method, network behavior feature vectors which are detected and are considered to be abnormal are stored and are used for training the deep neutral network. When occurring again, the intrusion types can be detected and recognized. While the model training and detection efficiency are guaranteed, the detection accuracy is improved, and the safety of the mobile ad hoc network is further improved.

The invention provides a topic detection method and device based on big data. By means of the topic detection method and device, accuracy and timeliness of detection effects can be guaranteed at the same time under the condition that a large amount of web page text is updated quickly in the internet environment. The method includes the following steps: extracting hot web pages according to user network behavior data; collecting contents of the hot web pages; extracting and acquiring web page feature vectors of the hot web pages according to the contents of the hot web pages; clustering the hot web pages according to the web page feature vectors of the hot web pages to acquire corresponding potential hot topic classes; carrying out incremental clustering on newly-added web pages with the potential hot topic classes as seed classes, wherein the newly-added web pages comprise on-line web pages; and judging whether the potential hot topic classes after the incremental clustering are hot topic classes through analyzing corresponding user attention degree parameters of the potential hot topic classes.

The invention relates to an island detection method of a photovoltaic grid-connected power generating system, which comprises the following steps of: detecting load power in a photovoltaic grid-connected power generating system; judging whether the load power is matched with the output power of a preset inverter or not, if so, selecting an active detection mode for carrying out island detection; and otherwise, selecting a passive detection mode for carrying out island detection. Correspondingly, the invention also relates to an island detection device and a photovoltaic grid-connected power generating system. By combining active island detection with passive island detection, the invention not only reduces a detecting blind area and effectively ensures island detection efficiency but alsoimproves electric energy quality.

The invention provides a face detection method and device, and the method comprises the steps: inputting a to-be-detected image into a first convolution neural network, detecting a face region in theto-be-detected image through the first convolution neural network, and marking a first face candidate window in the detected face region; adjusting the scale of an (M-1) face candidate window to be the scale of an M-th convolution neural network, inputting the scale of the M-th convolution neural network into the M-th convolution neural network, detecting a face region of the (M-1) face candidatewindow through the M-th convolution neural network, and carrying out the marking of an M-th face candidate window in the detected face region, wherein M=2, 3, ..., N, and N >=3; merging an N-th face candidate window and an (N-1)-th face candidate window into a face well-chosen window through a global non-maximum suppression method, wherein the first convolution neural network, the second convolution neural network, ..., the N-th convolution neural network are cascaded according to a scale sequence of convolution neural networks from the small to the big. The invention also provides a face detection method and device. The method and device can detect the face region precisely and efficiently.

The invention provides RFID label detection equipment used for carrying out detection upon RFID labels. The equipment comprises a frame, and an unreeling unit and a reeling unit respectively arranged on the two ends of the frame. A label conduction detection unit, a vision detection unit and a label changing platform are successively arranged between the unreeling unit and the reeling unit. Paper-guiding rollers are arranged between the units. The label conduction detection unit is arranged between the unreeling unit and the vision detection unit. According to the RFID label detection equipment provided by the invention, the printing quality of the labels, including variable information in the printed information, is detected by using the vision detection unit. The conduction performance of the labels is detected by using the label conduction detection unit. The unqualified labels are replaced by using the label changing platform. Therefore, efficiency and accuracy of the detection, and quality of the products can be ensured.

The invention discloses a WebShell detection method and an electronic device. The method comprises the steps of acquiring an access log that meets a first preset condition, wherein the first preset condition shows that the access log aims at a dynamic page; extracting the information of a universal resource identifier (URI) out of the access log; parsing the information of the URI to obtain a parsed result based on the first application; judging whether the parsed result contains preset characteristic parameters or not; on the condition that the parsed result contains preset characteristic parameters, determining that the information of the URI points to a WebShell, and conducting the alarming treatment on the information of the URI.

The invention discloses a different-source image fusion method based on salient target region extraction and NSST (Non-subsampled Shearlet Transform). The method comprises the following steps that: firstly, extracting the salient target region of an infrared image; carrying out registration on infrared and visible images to obtain an affine transform matrix; through the affine transform matrix, obtaining the position of the salient target region of the infrared image in the visible light image; carrying out NSST on the infrared and visible light image; utilizing different methods to carry outfusion on the NSST results of a target area and a non-target area of two images; and finally, through NSST inverse transform, obtaining a final fusion result. By use of the method, two different-source images of the infrared light and the visible light can be effectively fused, a fusion result can better highlight and keep the target, and the background detail information of the non-target area inthe visible light image can be more favorably kept.

The invention relates to the technical field of face detection, and provides a face detection model training method and device. The method comprises the following steps: extracting a feature map of atraining image by using a feature extraction network and determining a training sample; taking i as 1 to M, inputting the feature map to the i-th personal face detection network, obtaining a prediction box corresponding to the training sample to obtain a position and a probability of the existence of a face in the prediction box; Calculating the loss of the i-th personal face detection network according to the classification label of the training sample; When i is less than M, determining a classification label of the training sample when calculating the loss of the (i+1)th human face detection network according to the positional relationship between the prediction frame and the real frame corresponding to the training sample; When i is M, the total loss of M personal face detection network is obtained and the parameters of face detection model are updated; Continue to train the face detection model until the end of the training conditions are met. The precision of face detection modeltrained by this method is better than that of the existing single-stage face detection model.

The application relates to a computer technology, and especially relates to a method and a device for detecting horizontal permission vulnerabilities. The method and the device are used to improve the detection accuracy of horizontal permission vulnerabilities and improve the detection efficiency. The method includes the following steps: replacing original user identity information carried in a target request message with corresponding test user identity information; accessing a corresponding target address based on the test user identity information carried in the target request message; and determining whether a service function corresponding to the target request message has horizontal permission vulnerabilities according to whether a received response message contains sensitive information corresponding to the original user identity information. Thus, no additional operation processing burden is placed on the system, and the efficiency of vulnerability detection is guaranteed. Meanwhile, as vulnerability detection is carried out based on the nature of horizontal permission vulnerabilities during operation, the accuracy and credibility of detection results are ensured. The false detection rate and missing report rate are reduced, and the detection performance is improved.

The invention discloses an automatic traveling detecting device suitable for an air train track and belongs to the field of air train track detection. An air train trolley capable of automatically operating on the track and convenient to dismount and place is arranged on the air train track, and various information collecting sensors, a navigation communication system and an obstacle cleaning mechanism are arranged on the air train track, so that the operation state of the air train track is automatically detected, the specific positions of damaged parts are fed back in real time, and obstacles in the air train track are cleared. The automatic traveling detecting device suitable for the air train track is simple in structure, easy to operate and control and capable of greatly improving theoverhauling efficiency of the air train track, reducing the overhauling cost of the air train track, and ensuing normal and stable operation of the air train track.

The invention provides an assembling device and method of crystal arrays of a PET (Polyethylene Terephthalate) detector. The assembling device comprises a square frame, two pushing and blocking components and two spring clamps, wherein the two pushing and blocking components and the two spring clamps are arranged in the square frame; and one ends of the two spring clamps are respectively connected with the two pushing and blocking components, the other ends of the two spring clamps are respectively connected with the two deviated ends on the two adjacent edges of the square frame, and an elastic extruding and setting space of basic crystal array units to be assembled is formed between the other two adjacent edges of the square frames and the two pushing and blocking components. According to the invention, not only can the basic crystal array units bonded by single crystals be conveniently extruded into crystal arrays which are distributed uniformly and have good alignment in the two-dimensional direction so as to guarantee the detecting efficiency of the PET detector in the later-stage manufacture, but also the structure is simple, the cost is low, the operation is convenient and the assembling efficiency of the crystal arrays can be greatly improved.

The invention discloses a method for extracting a suspected flame region based on infrared light and visible light images, and relates to the field of computer vision. In this method, brightness threshold segmentation is performed on a collected infrared image to obtain a suspected high-temperature flame region; at the same time, the infrared image and a visible light image are image-registered toobtain an affine transformation matrix, and the position of the suspected high-temperature flame region on the visible light image is obtained by combining the suspected high-temperature flame regiontherewith; and finally, a suspected flame region is obtained by flame segmentation based on the YCrCb color space, morphological processing, and connected component labeling. The method can effectively extract the more accurate suspected flame region, can guarantee the detection of the suspected flame region in the night environment at the same time, and is conducive to the all-weather flame detection.

A wearable device which diagnoses personal cardiac health condition by monitoring and analyzing heartbeat includes a motion sensor detecting whether the user is in a motion state; a heart rhythm sensor detecting the user's signal of heartbeat interval; a memory storing algorithm program and database for data process, comparison, and analysis; a microprocessor calculating, filtering and judging the signal of heartbeat interval; and a feedback module displaying or alarming a judgment result, wherein the wearable device continuously detects the user's motion state and heartbeat condition, and detects heart rhythm when the user is not in the motion state to further judge the user's cardiac health condition, such as atrial fibrillation.

The present invention provides a contourlet transform-based image splicing detection method. Contourlet transform can greatly describe contour and directional texture information in an image. According to the method, contourlet transform is carried out on training images, and then, Markov features are extracted; a support vector machine-based recursive feature elimination (SVM-RFE) method is used to carry out dimensionality reduction on a feature set, so that detection efficiency and accuracy can be improved; optimal parameters are found by using the feature set which has been subjected to dimensionality reduction, so that an SVM classification model can be obtained; corresponding feature vectors are extracted from a test image; and the obtained classification model is adopted to carry out classification prediction on the feature vectors of the test image, so that the result of judgment on whether the test image has been subjected to splicing operation can be obtained. According to the method, feature extraction is carried out in a new transform domain, so that the method has high detection efficiency and accuracy.

The invention provides a railroad platform invasion alarm system by aiming at the problems in the prior art. The railroad platform invasion alarm system adopts the structure that field equipment realizes acquisition and analysis of invasion alarm information through a detection sensor unit and a video acquisition and processing unit for an embedded main control unit to perform alarm analysis, theembedded main control unit transmits the alarm information to a back end server side unit through a communication unit, and the embedded main control unit controls an alarm voice playing unit to sendout live alarm voice. For long detention invasion personnel, communicating information is transmitted to a field digital network broadcast playing unit through a digital network broadcast acquisitionunit and by virtue of the communication unit, so that the aim of frightening the invasion personnel is fulfilled.

The invention discloses rail transit tunnel geological radar mobile intelligent detection equipment and an operation method. The working state of each rotary telescopic device is controlled by a console, the tail end of each rotary telescopic device is provided with a radar antenna detection device through a telescopic slideway device, the rotary telescopic device rotates under the control of theconsole, the working angle and the working pose of the radar antenna detection device installed on the telescopic slideway device are adjusted, and the fixed distance, which is required by detection,between the radar antenna and the tunnel lining wall is kept; each rotary telescopic device is further provided with a laser detection device, and obstacles on the surface of the tunnel lining are monitored in real time in the moving process of an engineering vehicle so as to avoid the obstacles and adjust the detection route. According to the invention, the laser detection result can be received,the detection path can be set, the working posture of the detection system can be intelligently adjusted, and intelligent detection is achieved.

The invention discloses a single photon avalanche diode detector array used for laser range finding. The array comprises a single point detection surface, shaping circuits, an ''And'' processing circuit and an ''Or'' processing circuit. The single point detection surface is integrated by a plurality of detection single tubes in an n*n (n is greater than or equal to 2) mode. Each detection single tube includes a single photon avalanche diode and a quenching circuit. An output signal of each detection single tube is shaped to a pulse square wave signal by the respective shaping circuit. Each twopulse square wave signals form one group and are input into the ''And'' processing circuit. All the signals which are processed by the ''And'' processing circuit are processed by the ''Or'' processing circuit so as to generate a final output signal. By using the detector array, false triggering caused by dark counting and a stray light noise can be effectively reduced; detection efficiency is guaranteed and simultaneously range finding accuracy is ensured; and laser range finding with a high sensitivity can be realized.

The invention belongs to the technical field of pedestrian detection and relates to a small target pedestrian detection method under complex scenes. Firstly, a neural network is trained. The sample data is normalized in size; The shared feature extraction network is used to extract the shared feature, and the feature map is obtained. Classification feature extraction network is used to extract classification features from that feature map to obtain a classification feature map; the rough regression operation is carried out, the result of the rough regression is filtered by the confidence degree to obtain the candidate region; Classification of pedestrian target or not is carried out according to the classification feature map and the corresponding classification feature extracted from thecandidate area. Quadratic regression was performed on the candidate region, and the final regression positionis obtained by combining the position obtained by rough regression operation. SoftNMS is carried out by using the final regression position, and the duplicate box is removed to obtain the final target position and the corresponding classification result. The invention can effectively solvethe problems of high false detection rate and imprecise positioning of small target detection under complex scenes.

The invention relates to an EDID (extended display identification data)-integrated display frame deviation detecting system and a using method. The EDID-integrated display frame deviation detecting system comprises a PC (personal computer), a signal generator, a CCD (charge coupled device) camera and a detected display. The PC is connected with a video interface of the detected display and used for reading EDID information of the detected display and judging and analyzing the EDID information and is further connected with the input end controlling the signal generator so as to control the signal generator to output proper detection images, the output end of the signal generator is connected with the detected display and outputs the detection images to be displayed on the detected display, the CCD camera is arranged near the detected display and aligned to the same for capturing the displayed detection images; the output end of the CCD camera is connected with the PC, and the PC reads the pictured detection images and is used for judging whether the images offset or not. Via EDID-integrated detection, the detection images in the optimal resolution rate are automatically output for different displays, and image offset is judged by machine vision technology instead of manual detection.

The invention discloses a method for old man falling-down detection based on an attractor propagation algorithm. The method comprises a step 1) of collecting sample information of each sensor; a step 2) of processing the sample information and performing normalization, and constructing a sample state weighted sum; a step 3) of inputting the sample state weighted sum into an AP cluster; a step 4) of utilizing clustering results to construct a balanced binary tree, and marking the clustering results; a step 5) of collecting information of each sensor; a step 6) of processing the sensor information and performing normalization, and constructing a state weighted sum; a step 7) of searching the balanced binary tree according to a weighted sum result, and outputting a predicted result. According to the method, through the AP cluster implemented through the attractor propagation algorithm and the balanced binary tree constructed through the state weighted sum, the old man falling-down detection accuracy rate is effectively improved, and thus the problem that an existing falling-down detection method is not high in accuracy rate and large misjudgment occurs is solved.

The embodiment of the invention discloses training and using methods of a driving detection model, training and using devices of the driving detection model, equipment and a medium. The method comprises: acquiring a training sample, wherein the training sample comprises a driving front vehicle image, and a vehicle marking result and a lane line marking result in the driving front vehicle image; inputting the training sample into a shared extraction sub-model of a driving detection model to obtain shared image features; supervising and training a vehicle detection sub-model and a shared extraction sub-model of the driving detection model by adopting shared image features; and supervising and training the lane line detection sub-model and the shared extraction sub-model of the driving detection model by adopting the shared image features, so as to train the driving detection model. By adopting the scheme of the invention, the vehicle detection sub-model and the lane line detection sub-model are integrated in the same model for supervised training, so that vehicle detection and lane line detection can be realized simultaneously when the trained model is used subsequently.

The invention belongs to the technical field of mobile malicious software detection, and discloses a mobile malicious software detection method and system with privacy protection, a storage medium andan application. In a feature training phase, a semi-trusted part of a malicious software detection service provider MDSP operates software with known attributes in a simulator, a malicious feature set MS and a normal feature set NS are obtained by analyzing data during operation, and a corresponding effective Bloom filter and a corresponding effective Sketch data structure are established for detection and use; the mobile device collects runtime data of the position attribute software, encrypts the runtime data and sends the encrypted runtime data to the MDSP; the MDSP loads the Bloom filter,the Sketch and the received encrypted data to a trusted part of the MDSP and then executes a detection program; collecting software with newly confirmed attributes, adding the software into the training set, and updating MS and NS. Mobile malware detection is allowed to be outsourced to cloud in an efficient mobile user privacy protection mode to achieve malware detection serving as service, andthe method is suitable for malware detection algorithms matched with various features.

The invention discloses a microbial chip detection and supporting device, which comprises a base. Support columns are fixedly connected to both sides of the upper surface of the base. The inside of the groove is slidingly connected with a sliding seat, the inside of the sliding seat is rotatably connected with a rotating shaft, the rear end surface of the rotating shaft is fixedly connected with a hexagonal block, and the rear end surface of the support column is fixedly connected with a tapered groove seat, the hexagonal block is slidably connected to the inner surface of the tapered groove seat, the inner surface of the rotating shaft is fixedly connected with a support seat, and the inner surface of the support seat is provided with a second slide groove, and the second slide A supporting spring is fixedly installed on the inner bottom surface of the groove, and a hexagonal mounting bracket is slidably connected to the inner upper end of the second sliding groove. The microbial chip detection and support device can greatly improve the convenience and stability of use, ensure the detection efficiency, and save the detection cost.

The invention relates to a turbine blade processing and on-line detection system based on a numerical control processing device. The turbine blade processing and on-line detection system mainly comprises an industrial control computer, a circuit switching module, a digital value I / O (input / output) card, a numerical control machine tool, a laser displacement sensor and the like. The circuit switching module realizes switching between a blade processing system and an on-line detection system; two working procedures of processing and detection can be finished by only clamping and positioning a blade once; the blade processing and detection working procedures can be implemented in a numerical control system; the generation algorithms of a cutter trace point and a laser detection point are consistent; furthermore, the speed, the acceleration, the pause time, the position precision and the like of a target detection point under the two states of processing and detection can be matched; deflection, which is caused by mis-match of two system parameters, between an actual laser detection point and a theoretical point to be detected is avoided; the substantial cost for independent adoption of the detection system and system development and maintenance is avoided.

The invention relates to a face detection method based on context information enhancement, and the method comprises the steps: collecting image data comprising face targets in different scenes, and marking the face targets contained in each image; Dividing an image data set; Designing a context information enhancement-based deep convolutional network structure; Designing a context information enhanced sub-network; Designing a multi-path enhanced sub-network; DESIGNING DETECTION SUBNETWORKS.

The invention provides a chemiluminescence detector. The chemiluminescence detector comprises a reagent storage device, a reaction device, a dispensing device, a cleaning device and a measurement device; the reagent storage device is used for storing reagents; the reaction device is used for carrying reaction containers and performing sample addition, reagent addition and incubation operation; thereaction device comprises a reaction outer disk mechanism for carrying the reaction containers and performing reagent addition and mixing operation, and a reaction inner disk mechanism for carrying the reaction containers and performing incubation operation, wherein the reaction outer disk mechanism sleeves the outer side of the reaction inner disk mechanism, and the reaction outer disk mechanismand the reaction inner disk mechanism operate separately; the reagent storage device, the dispensing device, the cleaning device, and the measuring device are disposed around the reaction outer diskmechanism, in this way, operation on the reaction outer disk mechanism and the reaction inner disk mechanism can be performed simultaneously without interfering each other; and therefore, the detection speed of the chemiluminescence detector is improved, and detection efficiency is ensured. The invention also provides a detection method of the chemiluminescence detector.

The invention relates to an SSR (Simple Sequence Repeat) core primer group and a method thereof for identifying a tea variety, and belongs to the technical field of biology. The method is characterized by comprising the following steps: (1) material selection; (2) genome DNA extraction; (3) PCR (Polymerase Chain Reaction) amplification of a target product; (4) electrophoretic separation of a PCR product; (5) silver staining development; (6) data analysis; and (7) result judgment. The SSR core primer group and the method thereof for identifying the tea variety, which are disclosed by the invention, have the characteristics of easiness and convenience for operation, accuracy in result, high repeatability, high identification capacity, low cost and the like and are suitable for identifying the tea variety.

The invention discloses a multi-sample intermittent rotation detection device, which comprises a base, a placement platform, a rotation placement plate, a placement groove, a wing plate, a second rotation shaft, a cam, a motor, a high-power lens and a camera. The left support plate and the upper mounting plate are provided with brackets, mounting platforms, first rotating shafts, ball bearings, rotating mounting plates, mounting grooves and multiple sets of wing plates on the left side of the upper end of the base. Multiple sets of blind holes, Steel balls and springs, a second ball bearing, a second rotating shaft, a transmission gear and a cam are arranged on the right side of the upper end of the base, a motor and a driving gear are arranged at the upper end of the base, and a high-power lens and a camera are arranged on the upper mounting plate. The invention can drive the wing plate and the rotating placement plate to perform intermittent rotation through the rotation of the cam, and can form a stable intermittent motion with blind holes, steel balls and springs, which is convenient for automatic and stable continuous multi-group sample photography, and is convenient and efficient to use .

The invention discloses automatic detection equipment for producing a battery. The automatic detection equipment comprises a rack, wherein fixed frames are respectively arranged at two ends of the rack in a mutual symmetry manner; a detection rotating shaft is arranged on the inner wall, close to a rack, of the two fixed frames; a detection rotary table is arranged at one end, close to the outer walls of the fixed frames, of the detection rotating shaft; a reinspection rotary table is arranged at one end, close to the outer walls of the fixed frames, of the reinspection rotating shaft; an outlet rotary table is arranged at one end, close to the outer walls of the fixed frames, of an outlet rotating shaft; positioning grooves are formed in the outer wall of the detection rotary table; detection heads are arranged on the bottom surfaces inside the corresponding positioning grooves. The automatic detection equipment disclosed by the invention is convenient to assemble due to an uncomplicated structure and few components; the automatic detection equipment can be arranged at the unbroken position of a production process flow of the battery; a production line for integrated production and detection has high degree of automation; in addition, according to the automatic detection equipment disclosed by the invention, a formed battery is detected by adopting two detection tables, so that the detection efficiency is greatly ensured and leak detection is avoided.

The invention discloses a small-scale equipment part detection method based on weak supervision collaborative learning in an electric power field open scene, and the method comprises the steps: carrying out the fusion of shallow features and deep features through employing a feature pyramid based on the features of a small target of an equipment part, and obtaining richer information. After the extracted multi-scale features are input into the candidate region generation network, candidate regions under different scale features can be generated, and the processing range of the strong and weaksupervised learning network is divided according to the scale sizes of the candidate regions, so that the high performance of the strong supervised sub-network and the collaboration of the weak supervised sub-network are brought into full play. The time cost is reduced to a great extent, and the efficiency and the precision are balanced. Meanwhile, a detection framework different from a classic Faster R-CNN model is used for detecting the target, and the precision and speed of small target detection are improved at the same time.