36253results about How to "Improve detection accuracy" patented technology

The invention discloses a small target detection method based on feature fusion and depth learning, which solves the problems of poor detection accuracy and real-time performance for small targets. The implementation scheme is as follows: extracting high-resolution feature map through deeper and better network model of ResNet 101; extracting Five successively reduced low resolution feature maps from the auxiliary convolution layer to expand the scale of feature maps. Obtaining The multi-scale feature map by the feature pyramid network. In the structure of feature pyramid network, adopting deconvolution to fuse the feature map information of high-level semantic layer and the feature map information of shallow layer; performing Target prediction using feature maps with different scales and fusion characteristics; adopting A non-maximum value to suppress the scores of multiple predicted borders and categories, so as to obtain the border position and category information of the final target. The invention has the advantages of ensuring high precision of small target detection under the requirement of ensuring real-time detection, can quickly and accurately detect small targets in images, and can be used for real-time detection of targets in aerial photographs of unmanned aerial vehicles.

This invention is intended to control the amount of power to be supplied to a fusing heater below a maximum applicable current value. The engine controller supplies electricity to both of two heating bodies at the same fixed duty D1. At a phase angle α1 corresponding to the fixed duty D1, pulse signals ON1 and ON2 are issued in response to a ZEROX signal as a trigger. A current value I1 is detected based on a HCRRT signal from the current detection circuit. The engine controller calculates an upper limit of applicable power duty Dlimit based on the detected current value I1, the fixed duty D1 and the preset applicable current value Ilimit. Then, a PI temperature control is performed at a duty below the upper limit duty Dlimit.

The invention relates to an external skeleton robot, in particular to a human-simulated external skeleton robot assisting the lower limbs. The human-simulated external skeleton robot assisting the lower limbs aims to solve the problems that an existing external skeleton robot is low in coupling degree of motion space and poor in wearing comfort, reliability and adaptation, and power needed by a motor is large. The human-simulated external skeleton robot assisting the lower limbs comprises an upper body back part, a left leg and a right leg. The left leg and the right leg respectively comprise a hip drive system, a knee drive system and a foot wearing system. A rear side connection board of the waist is in rotating connection with a load installation board. Each hip joint supporting board is provided with a first motor and a first reducer, wherein the first motor is provided with an encoder, and the output end of the first motor provided with the encoder is connected with the input end of the first reducer. Each hip joint connecting board can rotate in the vertical plane. Each thigh stretching board is in detachable connection with the corresponding hip joint connecting board. The output end of a main drive mechanism is connected with each crus connecting board. The lower surfaces of elastic boards are bonded with the upper surfaces of the rubber soles of the feet. The human-simulated external skeleton robot assisting the lower limbs can assist in walking.

An enterprise based workflow server based system to detect and classify images as potentially of pornographic nature. The system is based on two independent stages. The first stage is an automatic stage which eliminates most images that are positively non pornographic, including an image analysis engine with a plurality of sub engines, each dedicated to different attributes of an image. The second stage describes an additional embodiment of a manual review stage. An alternative embodiment describes a similar server solution to detect images suspected as copyrighted material. The images can be individually classified, or analyzed as a cluster or a collection of images such as an album or motion-picture segment, sampled into key still frames. The system optimizes time saving by elimination of images, and assures accuracy by generating a workflow reviewing solution which can involve a human decision-making stage.

Gaming chips 30 have an IC tag 34 which is embedded therein, and a special bar code 101 invisible to the naked eye, which is printed thereon. A server calculates the number of gaming chips 30 based on individual information of the IC tag 34 that was read out by a reader 38, based on individual information of the special bar codes 101 read out by a bar code scanner 102, based on weight data acquired at a weighing device 40 and based on image data processed in a camera 103. Further, the server checks a partial number or the total number of gaming chips 30 thus calculated, and outputs the check results from a display or speaker.

A motion detection method, apparatus and system are disclosed in the present invention, which relates to the video image processing field. The present invention can effectively overcome the influence of the background on motion detection and the problem of object “conglutination” to avoid false detection, thereby accomplishing object detection in complex scenes with a high precision. The motion detection method disclosed in embodiments of the present invention comprises: acquiring detection information of the background scene and detection information of the current scene, wherein the current scene is a scene comprising an object(s) to be detected and the same background scene; and calculating the object(s) to be detected according to the detection information of the background scene and the detection information of the current scene. The present invention is applicable to any scenes where moving objects need to be detected, e.g., automatic passenger flow statistical systems in railway, metro and bus sectors, and is particularly applicable to detection and calibration of objects in places where brightness varies greatly.

A display device with a touch screen having display pixels and optical sensors each arranged in a matrix. In the display device, a backlight controller controls a backlight unit to be turned ON in a data display duration and controls the backlight unit to be turned OFF in a sensor detection duration other than the data display duration. Consequently, the influence of backlight on a touch part when the shade of the touch part is to be detected, can be reduced and an accuracy in the detection of an optical sensor can be increased

The invention discloses a cucumber picking robot system in a greenhouse environment. The robot system comprises a binocular stereo vision system, a mechanical arm device and a robot mobile platform; the binocular stereo vision system is used for acquiring cucumber images, processing the images in real time and acquiring the position information of the acquired targets; the mechanical arm device is used for capturing and separating the acquired targets according to the position information of the acquired targets; and the robot mobile platform is used for independently moving in the greenhouse environment; wherein, the binocular stereo vision system comprises two black and white cameras, a dual-channel vision real-time processor, a lighting device and an optical filtering device; the mechanical arm device comprises an actuator, a motion control card and a joint actuator; and the robot mobile platform comprises a running mechanism, a motor actuator, a tripod head camera, a processor and a motion controller. The invention also discloses a cucumber picking method in the greenhouse environment. The method of combining machine vision and agricultural machinery is adopted to construct the cucumber picking robot system which is suitable for the greenhouse environment, thus realizing automatic robot navigation and automatic cucumber reaping, and reducing the human labor intensity.

When starting a brushless motor, if the stop position of the rotor is detected between time t1 and time t2, a start-up excitation pattern in accordance with the rotor stop position is input for an initial energization time Ts1. Afterward, when the energization is stopped, a plurality of signals SL1, SL2, SL3, SL4 are generated in sequence in excitation switch timing signals in accordance with the rotational position of the coasting rotor. From these signals SL1 to SL4, the rotor position is detected using the second and subsequent signals SL2 to SL4 and then the process shifts to ordinary energization switch control. In accordance with the present invention, it is possible to start up a motor in a short time with a simple method so as to obtain a large torque during start-up.

The invention relates to an intelligent traffic signal lamp control device based on the comprehensive computer vision, which comprises a camera device and a microprocessor, wherein the camera device is installed above the middle part of an intersection and used for obtaining a traffic state panorama video image at the whole intersection; the microprocessor is used for detecting the queue length of vehicles waiting for passing through at the intersection and the congestion state of an outlet according to the panorama video data of the camera device and deciding the signal lamp control strategy and controlling the signal lamp switching action; the intelligent control carried out by the signal lamp control strategy of each phase to traffic signal lamps is determined through detecting the vehicle queue length at an inlet of each phase and the congestion state at the outlet, so that the sequence at the intersection is good, therefore, the conflict points where the traffic accidents are possibly caused are reduced or completely eliminated, and the operating indexes at the intersection are optimal.

Disclosed is a sterile adapter. The sterile adapter is interposed between an instrument holder formed at a surgical robot arm and a surgical instrument mounted on the instrument holder, wherein the sterile adapter comprises: a body portion comprising a first surface facing the instrument holder, and a second surface facing the surgical instrument; a first wheel which is elastically supported in the body portion, is exposed to the first surface, and is fitted to a first drive wheel formed at the instrument holder; and a second wheel which is elastically supported in the body portion, is exposed to the second surface, and is fitted to a second drive wheel formed at the instrument.; An elastic structure is applied to the sterile adapter to allow the wheels formed in the sterile adapter to be pushed in both directions, thereby enabling the easy and convenient mounting of a surgical instrument on a robot arm without applying a separate elastic structure to the instrument holder or the instrument.

The invention provides a security detection method and system of an Android application program, aiming to solve the problems of an antivirus method of the existing Android platform that the scanning speed is slow and the false positive ratio is high. The method comprises the following steps of: scanning an Android installation package and extracting appointed characteristic information from the Android installation package; finding characteristic records which are matched with the appointed single characteristic information or a combination thereof from a pre-arranged security identification base, wherein the security identification base comprises the characteristic records and security grades corresponding to the characteristic records, and each characteristic record comprises single characteristic information or the combination of the characteristic information; and respectively displaying the security grade corresponding to the found characteristic records in a security detection result of the Android installation package. The security detection method and system of the Android application program disclosed by the invention have the advantages of fast scanning speed and high searching and killing accuracy.

The invention provides a method for detecting an X-ray mammary gland lesion image based on feature pyramid network under transfer learning. The method comprises the steps: 1, establishing a source domain data set and a target domain data set; 2, establishing a deformable convolution residual network layer by a deformable convolution and extended residual network module; 3, establishing a multi-scale feature extraction sub-network based on a feature pyramid structure through a feature map up-sampling and feature fusion method in combination with a deformable convolution residual network layer;4, establishing a deformable pooling sub-network sensitive to the focus position; 5, establishing a post-processing network layer to optimize a prediction result and a loss function; and 6, migratingthe training model to a small sample molybdenum target X-ray mammary gland focus detection task so as to improve the detection precision of the network model on the focus in the small sample image. According to the method, a transfer learning strategy is combined to realize focus image processing in a small sample medical image.

A current detection printed board includes: a board having a penetration hole that penetrates the board; and at least one wire that is formed in a coiled shape having both ends by penetrating the board along the periphery of the penetration hole and alternately connecting a front surface layer and a rear surface layer of the board, wherein, when a conductor, in which an AC current flows, is disposed to pass through the inside of the penetration hole, a current flowing in the wire is output through electromagnetic induction.

The invention provides an automatic vehicle detection method based on deep learning. The automatic vehicle detection method based on deep learning can utilize a convolution neural network to automatically extract the vehicle characteristics in the traffic environment so as to eventually realize automatic detection and identification of a vehicle target. The automatic vehicle detection method based on deep learning can be applied to different environments, different illumination intensities and different weather situations, and can guarantee higher accuracy for detection for a vehicle in a complicated traffic environment, and provides a new real-time reliable preprocessing step for identifying and tracking the vehicle target for the subsequent video monitoring.

To provide a touch sensor device and the like, which can improve the detecting accuracy of a surface-capacitive type touch sensor. A touch sensor device of the present invention includes: a touch panel which forms a capacitance between a human body and itself when the human body makes a contact or approaches thereto; a power supply part which has a battery or the like which outputs a DC voltage, converts the DC voltage to an AC voltage, and outputs the AC voltage from the touch panel; a reference potential part to be a reference potential of the AC voltage; a wrist strap and a cable which are electrically connected to the reference potential part and to which the human body makes a contact; and a current detecting part which detects electric currents flown from the power supply part to the human body.

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 relates to the technical field of traffic image processing, and discloses a driving scene target detection method based on deep learning and multi-layer feature fusion. The method comprises the following steps of 1) collecting a video image to serve as a training data set, and performing preprocessing; 2) building a training network; 3) initializing the training network to obtain a pre-training model; 4) performing training on the training data set by using the pre-training model obtained in the step 3), thereby obtaining a training model; 5) collecting a front image by using a vehicle-mounted camera, and inputting the image into the training model obtained in the step 4), thereby obtaining a detection result. The multi-layer feature fusion method based on a feature pyramid is adopted to enhance semantic information of a low-layer feature graph, so that the feature extraction quality of the network is improved, and higher detection precision is obtained.

A welding visual detection method based on a convolutional neural network includes the following steps that firstly, at the training stage, a training sample is input into the convolutional neural network, and the connection weight and offset value of the convolutional neural network are obtained; secondly, at the testing state, a welding picture is read in and preprocessed with the digital picture processing technology, and a region of interest is extracted and then subjected to picture size normalization processing to serve as input of the convolutional neural network. The invention further provides a welding visual detection device based on the convolutional neural network. The welding visual detection device based on the convolutional neural network comprises a crawling mechanism, a power transmission mechanism, visual detection equipment and a weld defect detection and analysis system. By means of the welding visual detection method and device based on the convolutional neural network, the automaton and intelligence level is improved, the detection precision is effectively improved, and the detection speed is effectively increased.

The invention provides a security detection method and system of an Android application program, solving the problems of low scanning speed and high false alarm rate of the traditional virus killing method of an Android platform. The security detection method comprises the steps of: scanning an Android installation package, extracting appointed characteristic information from the Android installation package; uploading the appointed characteristic information to a server, searching a characteristic record matched with appointed single characteristic information or combination from a security identification base preset in the server; and receiving a security detection result returned by the server aiming at the Android installation package, and displaying on a client end user interface, wherein the security detection result includes a security level corresponding to the characteristic record searched by the server. The client end detection is combined with the server detection, thus the security detection method is high in scanning speed and high in killing accuracy rate.

The invention provides an intelligent safety monitor system of an electric power working site. The intelligent safety monitor system of the electric power working site comprises a locating sub-system, an intelligent video analysis sub-system and an intelligent safety monitor sub-system. The locating sub-system comprises a locating server, a locating label and a base station unit. The base station unit obtains the position information of the locating label by detecting UWB signals and MAC data sent by the locating label, and sends the position information to the locating server. The intelligent video analysis sub-system analyzes a video stream of the electric power working site so as to monitor the operation actions of working personnel, and gives an alarm according to the analysis result. The intelligent safety monitor sub-system comprises a locating monitor platform and a video monitor platform. Compared with the prior art, the intelligent safety monitor system of the electric power working site can effectively prevent the personnel from breaking rules, prevent or reduce personal injury of an electric power system, and be widely applied to the electric power safety working and electric power construction working site.

The invention relates to an evaluation system of transformer state based on multisource information integration and an evaluation method thereof, wherein, the evaluation system in the invention is composed of an oil chromatographic analysis subsystem, a partial discharge ultrahigh frequency signal detection subsystem, a winding deformation vibration signal detection subsystem, a current transformer detection subsystem and a computer; the technical point of the evaluation method in the invention lies in evaluating the running state of a detected transformer by fusing the detection results obtained by the detection methods utilized by the four detection subsystems by utilizing a D-S evidence theory fusion evaluation algorithm. The beneficial effects of the invention lie in being capable of timely discovering and accurately judging the malfunction of transformer and comprehensively reflecting the running state of a transformer.

Laser Doppler Perfusion Imaging system including a laser light source (2), an image detector (4), and signal detector (6); the laser light source (2), in use, being arranged for illuminating a selected area of interest (12) on a sample (10) for determining perfusion related data of the sample (10), the image detector, in use, being arranged for capturing an image of the sample (10), and the signal detector (6), in use, being arranged for detecting a laser Doppler signal from the selected area of interest (12) on the sample (10), wherein the laser light source (2) comprises optical means (3) for creating a laser beam comprising a plurality of beams.

High-accuracy magnetic measurement is performed by efficiently using nitrogen-vacancy pairs in all orientations. A magnetic measurement apparatus includes a diamond crystal and an image sensor. The diamond crystal has nitrogen-vacancy pairs. The image sensor detects the intensities of fluorescence generated by an exciting light applied to the diamond crystal by using a plurality of pixels. The nitrogen-vacancy pairs of the diamond crystal are made to one-to-one correspond to the pixels. The fluorescence generated by one nitrogen-vacancy pair is received by one pixel made to correspond to the nitrogen-vacancy pair.