50results about How to "Solve the detection accuracy is not high" patented technology

The invention relates to a video flame detection method based on a dual-stream convolution neural network. The main process comprises the following steps: S1, preparing data, inputting a training dataset into the dual-stream convolution neural network for training, and obtaining a trained network model; 2, pretreating that video to be measured by cascade motion feature detection and color featuredetection to obtain a suspected flame region; S3, inputting the obtained suspected flame region into the trained two-stream convolution neural network for classification and identification; S4: identifying The region as the flame by the two-stream convolution neural network is output as the final detection result. The invention efficiently processes most of the non-flame regions in the video by extracting the suspected flame regions, and reduces the time complexity. By using two-stream convolution neural network, combining the video flame motion information, and adding the spatio-temporal pyramid pooling layer to provide more robust spatio-temporal characteristics, the detection accuracy is improved.

In large-scale high speed network, the invasion detecting system adopts always hierarchical distributive structure including dispersive acquisition, distributive processing and concentrated management. In the hierarchical modular structure, deputy technology is always adopted and the coordinating module in the coordinating management layer assigns and controls the deputies. In abnormal case that the static deputy in the lower layer can not settle, some request is sent to the coordinating module, and the coordinating module assigns movable deputy with specific function to the static deputy for complete the task before the cancel request is sent and the cancel command is sent separately to finish the assignment mechanism. The mechanism can balance the system load, respond timely, complete specific data processing function and make the movable deputy controllable.

The invention discloses a multi-spectral remote sensing image change detection method based on a spectral reflectivity neighborhood difference chart and neighborhood probability fusion. The method mainly comprises the steps that 1, a registered two-time-phase multi-spectral image set at the same area is input and is subjected to Wiener filtering noise reduction and normalization processing; 2, the processed image set is converted into a relative ground object spectral reflectivity image set; 3, the spectral reflectivity change module value is calculated, and a spectral reflectivity change module value chart is obtained; 4, the spectral reflectivity angle value is calculated, and a spectral reflectivity angle value chart is obtained; 5, a neighborhood difference value of the spectral reflectivity is calculated and the neighborhood difference chart is obtained; 6, clustering is carried out on the spectral reflectivity change module value chart, the spectral reflectivity angle chart and the neighborhood difference chart, and a binary image is obtained; 7, the obtained binary image is fused based on the neighborhood probability, and a change detection result chart is obtained. Manual participation is not needed, the detection precision is high, and the method can be used for the fields such as urban expansion monitoring, forest and vegetation change monitoring and the like.

The invention discloses an optical fiber mini-sized hydrophone belonging to a sound pressure sensor technology field. The optical fiber mini-sized hydrophone comprises a four-layer structure. A first layer is a diaphragm, which is a PDMS flat diaphragm or a PET convolution diaphragm. A second layer is a first supporting layer having a circular through hole in a center, and a third layer is the second supporting layer having the circular through hole in the center. The fourth layer is the third supporting layer having a circular through hole in the center and an air vent hole in an edge. The center of the diaphragm, the center of the first supporting layer, the center of the second supporting layer, and the center of the third supporting layer are sequentially aligned with each other, and are bonded together. The diameter of the circular through hole of the first supporting layer, the diameter of the circular through hole of the second supporting layer, and the diameter of the circular through hole of the third supporting layer are reduced sequentially, and the above mentioned circular through holes of the supporting layers are in a ladder-shaped arrangement. An inserting core is disposed in the circular through hole of the third supporting layer, and an optical fiber is inserted in the inserting core. An F-P cavity is arranged from the diaphragm to the end surface of the optical fiber. The optical fiber mini-sized hydrophone is used to solve technical problems caused by the materials or the shapes of the diaphragms.

The embodiment of the invention discloses a target object detection method and device, computing equipment and a medium. The method comprises the following steps: identifying the scene information ofthe to-be-detected image and the category information of the target object in the to-be-detected image; judging whether the category information and the scene information of the detected target objectare matched or not according to the pre-established corresponding relation between the object and the scene. The problems that when the target object is detected, many false detections exist, and thedetection accuracy is not high are solved. When the image is detected, the technical scheme of the embodiment of the invention improves the detection efficiency of the target object and realizes a more accurate effect of detecting the target object by combining the scene information represented by the whole image while utilizing the information of the object.

The invention provides a refrigeration system and a control method and control device thereof. The control method comprises the steps that detection parameters are obtained; whether the detection parameters meet state setting conditions or not is judged; if yes, a controller judges that the system is in the state and transmits a corresponding instruction. According to the technical scheme, the relationship between key parameters related to the working state of the refrigeration system and detection parameters easy to obtain directly is utilized, the working state of the refrigeration system is judged, and then corresponding control is conducted according to the working state. The refrigeration system and the control method and control device thereof have the advantages that detection elements used for detecting the working states of refrigeration systems in the prior art such as high-voltage switches, low-voltage switches, leeway switches and environmental temperature sensors are omitted, the size of the refrigeration system is reduced to a great extent, and the miniaturization of the refrigeration system is facilitated; compared with the detection conducted by detection elements of the prior art, through the control method, higher accuracy and reliability are achieved, the response speed is higher, and the cost is lower.

The invention discloses a data security protection system and method based on the Internet of Things. The system comprises an Internet of Things terminal identity authentication module, a terminal network traffic encryption module, a background traffic decryption module, a background identity authentication module, a traffic protocol analysis module, a sensitive data detection module and a physical encryption machine control module. According to the invention, multiple data security software and hardware are combined, so that more transparent sensitive data protection and automatic data desensitization which are non-inductive to a user under the scene of the Internet of Things are realized, and the effect of data security protection is achieved.

A fiber optic F-P sound pressure sensor based on metal film belongs to the technical field of the sound pressure sensor. The sound pressure sensor comprises a metal corrugated film, a cavity casing, a ferrule and an optical fiber. The metal corrugated film is used as a sound pressure sensitive element, and is bonded with one end face of the cavity sleeve with a cured adhesive, and a ferrule with the inner wall of the cavity casing is inserted from the other end face of the cavity sleeve, and the center of the ferrule is inserted with an end-polished fiber. The cavity sleeve, the ferrule and the fiber set co-axially, and a cure adhesive is used to fix the cavity sleeve and the ferrule, and the cure adhesive is also used to fix the ferrule and the fiber. The fiber optic F-P sound pressure sensor based on metal film is capable of solving the technical problems caused by the diaphragm material or the diaphragm shape.

The embodiment of the invention discloses a live poultry detection method based on machine learning. The live poultry detection method comprises the steps: obtaining target detection data corresponding to a to-be-detected region, and enabling the target detection data to comprise at least two of a target infrared image, a target RGB image and target sound data; inputting the target detection datainto a preset machine learning model, and obtaining an output result of the machine learning model; and determining target live poultry information corresponding to the to-be-detected area according to the output result. According to the live poultry detection method, the efficiency and accuracy of live poultry detection are improved.

The invention discloses a hair nourishing detecting comb. The hair nourishing detecting comb is characterized in that a comb body is provided with a sensing comb 10.1, wherein detecting subcircuits are arranged at two sides of the sensing comb 10.1 and used for sensing the frictional force, generated by combining hair of a user to be detected, on the sensing comb 10.1; a frictional force signal isinput to an amplifying circuit to be amplified and then is input to an external control performing module. According to the hair nourishing detecting comb, the circuits capable of detecting the frictional force are directly arranged on a comb body 10, so that the frictional force data can be obtained in real time, and as a result, the data foundation is provided for detecting the hair nourishingeffect; the abovementioned circuit structure and a mechanical structure are combined, so that the detecting efficiency can be improved; the detecting subcircuits which are the same in circuit structure are arranged at two sides of the sensing comb 0.1; the quantity of strain gage resistors can be increased as requirement; the accuracy of the data can be improved as demand, so that the detecting accuracy can be improved; related leads are arranged in a junction box 10.3, and thus the mutual disturbance of the circuit structures in the comb body 10 can be avoided, and as a result, the detectingaccuracy is improved.

The invention relates to a method for detecting inclusions and air holes in an aluminum alloy and belongs to the technical field of inclusion and air hole detection. The method comprises the following steps of sample preparation, rough grinding, fine grinding and measurement. A cast aluminum alloy sample is made into a lath shape or a round rod shape of a certain specification; the prepared sample is ground on a sand paper until the surface of the sample has no scratches, and flatness of the sample is kept as much as possible; the section area of the sample is computed; and electrical resistivity of the sample is measured through an alloy electrical resistivity measuring device, and inclusion contents of the sample are computed by contrasting. The method has the advantages of solving the problem of low accuracy of the inclusion and the air hole detection in the aluminum alloy, and quantitatively computing the inclusion contents. The method has simple and convenient operation, and high measuring accuracy.

The invention discloses a pressure sensor and electronic equipment, relates to the technical field of touch control, and is used for reducing the influence of temperature on the detection accuracy of the pressure sensor. The pressure sensor comprises a Wheatstone bridge and a thermal emission diode; the Wheatstone bridge comprises a first resistor, a second resistor, a third resistor and a fourth resistor; the first resistor, the second resistor, the third resistor and the fourth resistor are coupled to form a loop; the first resistor, the second resistor and the third resistor are fixed value resistors, and the fourth resistor is a piezoresistor. The first output end of the Wheatstone bridge is coupled to the first end of the thermal emission diode, and the second output end of the Wheatstone bridge and the second end of the thermal emission diode are used for outputting electric signals; when the temperature rises, the equivalent resistance value of the thermal emission diode is reduced, and the resistance value of the fourth resistor is increased; when the temperature is reduced, the equivalent resistance value of the thermal emission diode is increased, and the resistance value of the fourth resistor is reduced.

The invention discloses a mode recognition method and system based on function integration association, and the system comprises a filtering system data module, a mode recognition calculation module and a detection mode control module, the filtering system data module is used for collecting and recording the wind speed flow and pressure data related to a filter screen; the mode recognition calculation module is used for recognizing the filth blockage mode of the filter screen according to the recorded data and judging whether further accurate detection is needed or not, and the detection mode control module is used for regulating and controlling the detection mode in real time according to the judgment result and carrying out early warning reminding on the filth blockage condition of the filter screen. The filter system data module is electrically connected with the mode recognition and calculation module, the mode recognition and calculation module is electrically connected with the detection mode control module, and the filter system applies different detection modes to a filter screen, calculates the filth blockage degree coefficient of the filter screen and carries out real-time reminding and early warning according to a detection result. The method has the characteristics that the dirty mode of the filter screen can be identified and the detection method is adjustable.

The invention relates to the technical field of water flow velocity detection, and discloses a water flow velocity intelligent detection method and device, and the method comprises the steps: obtaining to-be-detected water, adding a thermal tracer agent into the to-be-detected water, obtaining a thermal infrared video, calculating the flow conversion value of the to-be-detected water and the thermal tracer agent, extracting a thermal infrared image based on the flow conversion value, and determining the flow velocity of the to-be-detected water according to the thermal infrared image. Constructing a target and a coordinate system, combining the target and the coordinate system with the thermal infrared image to obtain a region of interest of the thermal infrared image, carrying out temperature testing on the region of interest to obtain a thermal tracer region image, carrying out image processing on the thermal tracer region image to obtain a pure region image of the thermal tracer, determining a front edge point and a mass center point of the pure region image, and obtaining a thermal tracer image; instantaneous speeds of the front edge point and the mass center point are calculated respectively, the front edge instantaneous speed and the mass center instantaneous speed are obtained and subjected to arithmetic averaging, the average flow velocity of the water to be detected is obtained, and water flow velocity measurement is completed. The problem that the accuracy of water flow velocity detection is not high can be intelligently solved.

The invention provides a customer signature detection method and device in double-record data quality inspection, which can be used in the financial field and other fields. The method comprises the following steps: acquiring double-recording data and electronic signature information, and decomposing the double-recording data to obtain video data; analyzing the video data to obtain a signature picture group, and identifying the signature picture group to obtain a signature identification result; according to a preset validity rule, performing validity detection on the electronic signature information to obtain a validity detection result; and generating a client signature detection result according to the signature recognition result and the validity detection result. According to the method, the video data and the electronic signature information are identified and detected, the detection accuracy of a customer signature link during double-recording quality inspection is improved, the signature picture group is identified and detected, the problem that the detection accuracy is not high due to the fact that detection is carried out purely depending on the electronic signature information is solved; therefore, the double-recording quality inspection success rate is improved.

The invention discloses a colorimetric blood culture interpretation method and device. The method comprises the following steps: judging whether bacterial culture duration reaches a first preset value or not; if yes, acquiring an RGB value of the bottle bottom of a blood culture bottle, and judging whether a condition for converting the RGB value into an HSV value is met or not; if yes, converting the RGB value into an HSV value; acquiring a first change critical point and a second change critical point corresponding to the H value; acquiring H values corresponding to the first change critical point and the second change critical point, and judging whether a difference value of the H values corresponding to the first change critical point and the second change critical point is greater than a second preset value or not; if yes, judging the current moment as a positive judgment moment, and judging whether the S value is greater than or equal to the H value; and if yes, judging that a blood culture result is a positive result. Compared with the prior art, the detection accuracy of the blood culture result is improved, and the detection time is shortened, so that the detection efficiency is improved.

The invention provides a customer confirmation detection method and device in double-record data quality inspection, which can be used in the financial field or other fields. The method comprises the following steps: acquiring double-recording data, and decomposing the double-recording data into video data and audio data; performing confirmation action detection on the video data to obtain a confirmation action detection result; performing voice recognition on the audio data to obtain voice text data, and performing text rule matching on the voice text data to obtain a text matching result; and generating a customer confirmation detection result according to the confirmation action detection result and the text matching result. The video data and the audio data are identified and detected, so that the detection accuracy of a customer confirmation link during double-recording quality inspection is improved, meanwhile, the confirmation action of the customer is identified and detected, the problem of low detection accuracy caused by detection simply relying on voice-to-text is solved, and thus, the double-recording quality inspection success rate is improved.

The invention provides a nanometer magnetic microsphere, a preparation method thereof and application of the nanomemter magnetic microsphere in detection of drugs and metabolites thereof, and the preparation method comprises the following steps: (1) mixing iron salt with a chelating agent and an alkaline precipitator for reaction to obtain ferroferric oxide nanospheres; (2) mixing the ferroferric oxide nanospheres obtained in the step (1) with a solvent for ultrasonic treatment, and then conducting mixing with an alkaline regulator, tetraethyl orthosilicate and methacrylic acid-3-(trimethoxysilyl)propyl ester for reaction to obtain double-bond modified magnetic silicon dioxide microspheres; and (3) mixing the double-bond modified magnetic silicon dioxide microspheres obtained in the step (2) with a hydrophilic compound, a lipophilic compound and an initiator for reaction to obtain the nanometer magnetic microspheres. The nanometer magnetic microspheres provided by the invention are simple to prepare, can quickly complete detection when being applied to drug detection, and consumes less time.

A WLAN wireless network data encryption system comprises a packet capture processing module, a data analysis module and a management console module, wherein the packet capture processing module comprises a packet capture unit, a data preprocessing unit and a response unit, and the data analysis module comprises a data processing unit, a network performance detection unit, a security detection decision unit, an intrusion detection engine unit and a storage unit. The WLAN wireless network data encryption system provided by the invention can effectively improve the security of data in the WLAN wireless network.