460results about How to "Realize accurate detection" patented technology

The invention discloses a complex scene-based human body key point detection system and method. The method comprises the following steps of: inputting monitor video information to obtain a single-frame static map and a multi-frame optical flow graph; extracting features of the single-frame static map through a convolution operation so as to obtain a feature map, and in order to solve influences, on personnel target detection, of interference targets under complex scenes, judging a practical confidence coefficient and a preset confidence coefficient of the feature map by adoption of a personneltarget detection algorithm so as to obtain a discretized personnel target surrounding box; and carrying out optical flow overlapping on the multi-frame optical flow graph to form a two-dimensional vector field; extracting features in the discretized personnel target surrounding box to obtain a feature map, obtaining key points and association degrees of parts, generating a part confidence map foreach part of a human body by utilizing a predictor, and accurately detecting human body key points through the part confidence map and the two-dimensional vector field. The system and method are usedfor human body key point detection under complex scenes so as to realize accurate detection of personnel target key points.

The invention discloses a light path structure and a method for detecting large-dynamic range liquid turbidity. In the light path structure, an infrared emission light path, a first scattered light path, a second scattered light path, a long vessel transmission light path and a short vessel transmission light path are arranged in a groove-shaped concave cavity respectively; after being subjected to beam expanding by a convex lens in the front, light emitted from a detecting light source is transmitted in liquid in the concave cavity and is received by light detectors respectively, and an electrical signal outputted by each light detector is inputted to an amplification demodulator circuit to be converted into a direct-current voltage signal for calculating liquid turbidity; and the detecting light source realizes pulse modulation by the demodulator circuit. By the structure and the method, the requirement of the accurate detection of the turbidity of various water bodies with different concentrations is met fully, so the structure can be used for detecting water turbidity quickly on line by environmental monitoring departments and also can be used for performing long-term continuous detection on the water turbidity of fixed monitoring points by being arranged on a buoy or an automatic monitoring station, and has the excellent market prospect.

The invention discloses a curve detection device having multi-point curve detection loading manners and a curve detection method. The curve detection device comprises a base, an upper pressing block, curve pressure heads, a horizontal position adjustment device, at least two support blocks and two movement seats, wherein one ends of the support blocks are mounted on the upper pressing block in a sliding manner; the curve pressure heads are mounted at the other ends of the support blocks; the movement seats are mounted on the upper surface of the base in a sliding manner through the horizontal position adjustment device; backup rollers are arranged on the movement seats. The detection method comprises the following steps: (1) mounting the detection device; (2) mounting a test piece, and adjusting the positions of the support blocks and the movement seats; (3) starting a testing machine for detection. With adoption of the detection device and method, the detection device can meet the loading manners of various curve detection, avoids the situation that different curve loading manners adopt different loading devices, is simple in structure and excellent in applicability and reduces the cost; with adoption of the detection device, the operating processes are simplified, the operation is simple, and the reliability is excellent.

The invention discloses a stacking glass bottle detection method based on machine vision and a transporter based on the method. The detection method comprises the following steps: firstly, obtaining a bottle mouth image sent into a bottle collecting zone via a conveyor belt; then, performing image processing on the obtained bottle mouth graph so as to indentify whether the glass bottle in the current bottle collecting zone has abnormal situation or not; and if yes, giving an alarm. The transporter comprises a glass bottle conveyor belt, a travelling crane provided with a gasbag-type grappling fixture lifting gear and a fastening frame, a camera and a stand column rack, wherein the bottle collecting zone is directly below the fastening frame and is connected with the conveyor belt; a longitudinal guide rail and a horizontal guide rail which can move along the longitudinal guide rail are arranged in the bottle collecting zone; the camera used for collecting glass bottle images in the bottle collecting zone is arranged on the horizontal guide rail; and the lens of the camera is downwards vertical to the bottle collecting zone and is connected with a computer for processing images. The invention is easy to apply, has high degree of automation, high efficiency and wide industrial application prospect.

The invention discloses a biological sensor and a manufacturing method and a detection method thereof. The biosensor has a field effect transistor structure, which comprises a flexible substrate, printed silver source drain electrode arranged on the flexible substrate and a printing functionalized nano carbon material layer connected with the source drain electrode. A specific receptor is connected with the functionalized nano carbon material layer as a sensitive material for detecting minute quantity of a specific target material. The source drain electrode is connected with a set of signal detecting system, which cooperates with signal regulation of a suspended grid in a detection liquid to realize the rapid, accurate and sensitive biological diagnosis; therefore, the biological sensor has the advantages of integration, compatibility and portability.

The invention discloses a fatigue detection method for fixed-post personnel based on information fusion, which solves the problems of fatigue feature detection and decision-level information fusion under single-path video. Firstly, the regression tree method is used to calibrate the face of the input image to achieve high-precision face detection. Then, the feature points of eyes and mouths are detected accurately by using cascade convolutional neural network. The PERCLOS parameters, blink frequency and yawn frequency are calculated by analyzing the feature points of multi-frame images of thevideo. Then, the head pose is estimated by ASM-based local localization and facial feature triangles, and the nodding frequency is calculated. Finally, the information fusion algorithm based on roughset theory is used to realize the fatigue detection of fixed-post personnel.

The invention discloses a modified electrode, which belongs to the field of analytical chemistry. The electrode comprises a base electrode, Nafion film coated on the surface of the base electrode and bismuth-film coated on the surface of the Nafion film. The invention also discloses a preparation method of the electrode. The invention also discloses an electrolytic system adopting the electrode as the working electrode and the method for detecting heavy metal in a sample by utilizing the electrolytic system. The invention has the advantages that the response speed is high, the stability is good, the sensitivity is high, various heavy metals can be rapidly detected, the heavy metal in the sample with complex components can be detected; in addition, the environmental protection material is adopted for the electrode, therefore, pollution is avoided.

The invention discloses a defect defecting method based on a wide frequency band ultrasonic phase array, which is applied to an ultrasonic phase array transducer consisting of multiple array elements. The defect defecting method comprising the steps of: exciting by multiple array elements by adopting modulation codes, transmitting modified big-timewidth and big-bandwidth linear frequency modulation signals at a transmitting end, gathering the linear frequency modulation signals at a focal point and reflecting the signals; receiving the reflected signals at a receiving end by each array element and distributing the delayed and corrected signals within a same time window range; and implementing matched-filtering within a frequency domain according to the reflected signals received by each array element, coherently combining the matched-filtered signals within the frequency domain to obtain bandwidth signals within the frequency domain, and implementing inverse fast Fourier transform, thus obtaining time domain pulse compression signals. By virtue of the method, the bandwidth of the transmitted signals can be increased and the pulse compression signals with high resolution can be obtained, so that micro defects can be accurately detected.

The invention discloses an adaptive zero-speed interval detection method for a pedestrian navigation system. According to a fixed threshold method, a zero-speed detection interval has the problems of false detection and missing detection under the walking condition with varying gait frequencies; by an adaptive threshold method, the relationship between a zero-speed interval detection threshold and the gait frequency of a pedestrian can be analyzed, so that the detection accuracy of a zero-speed interval is effectively improved. By carrying out a calibration experiment of a zero-speed point judging threshold under the condition with different gait frequencies, the functional relationship between a zero-speed detection threshold and the gait frequency is established, and the adaptive adjustment of the detection threshold and the accurate detection of a zero-speed point are achieved, so that the detection accuracy of the zero-speed interval is improved; by the adaptive zero-speed interval detection method, adaptive detection of the zero-speed interval can be achieved only by using output data of a gyroscope and three accelerometers in an inertial measurement unit, without needing to add or use other external sensors to assist the detection of the zero-speed interval, so that the detection method is simple and accurate.

The invention provides a rail transit automatic detection system and method. The system comprises a small electric rail car, a rail detection device, an electrical control system and a rail state parameter detection data system. The small electric rail car is used as a detection platform of the rail detection device, the rail detection device is used for collecting the parameters of the steel rails, the tunnels and the contact networks and the data information of the running state of the rail car, the electrical control system is used for controlling the walking of the rail car and supplying power to the detection device, and the track state parameter detection data system is used for storing and processing the data obtained by the track detection device. According to the technical scheme,the problems that at present, the track geometrical state and tunnel deformation detection work is low in efficiency, and the applicability of a detection system is poor, are solved.

The invention discloses an acoustical method and a device for detecting low-concentration sulfur hexafluoride gas. The method comprises the following steps of: utilizing a relation between velocity of ultrasonic wave transmitting in gas and gas molar mass, measuring travel time when the acoustic wave passes through a detection channel with fixed length by a method of amplitude modulation and envelope detection applied to communication, calculating the travel time in air without mixed sulfur hexafluoride under a current environmental condition at the same time, and representing sulfur hexafluoride concentration by use of time difference of ultrasonic wave detection propagation delay and reference propagation delay. The method has a good detection effect and high precision, and problems of insufficient detection precision of the travel time and a cross cycle in a conventional low-concentration sulfur hexafluoride acoustic velocity detection technology are solved, and precise detection of low-concentration sulfur hexafluoride concentration in a single channel manner is realized.

The invention relates to an optical remote sensing image multi-class target detection method based on cross-scale feature fusion. The method comprises the steps: taking training data as the input of aconvolutional neural network to extract image features, obtaining a multi-scale feature map from the output of different convolutional layers, and adding an extrusion-excitation module at the topmostfeature to remodel the channel information of the top feature; performing cross-scale feature fusion operation on the obtained feature maps, training a region suggestion network on the multi-scale feature maps, obtaining a suggestion box for subsequent tasks from the trained region suggestion network, and sending the suggestion box to a classification network and a regression network for training; and finally, realizing accurate detection of multiple types of targets of the optical remote sensing image on the multi-scale feature map through post-processing operations such as non-maximum suppression and the like. By using the method of the invention, various types of targets can be detected from an optical remote sensing image under a complex background. The method has higher detection andidentification precision and higher speed.

The invention provides a 360-degree panorama based moving object detection method. The method comprises the steps that the steps of performing vehicle periphery four-channel blind-area-free video image data acquisition; performing vehicle periphery four-channel video image processing and panoramic stitching so as to form a 360-degree panoramic top view; detecting a moving object at the periphery of a vehicle by adopting a global-local cascading space-time salience model according to the 360-degree panoramic top view; generating a boundary of the periphery of the vehicle according to dynamic data acquired in real time; and measuring the relative distance between the moving object and the boundary of the vehicle, and generating an alarm signal according to alarm rules. According to the invention, the driving state and speed of the vehicle are analyzed by adopting a machine vision technology, detection is performed on a moving object at the periphery of the vehicle according to a video sequence formed by the 360-degree top view, a timely and accurate alarm signal is provided by combining the driving state and the driving speed of the vehicle, a driver is reminded to adjust the driving state timely and drive carefully, and thus occurrence of traffic accidents is reduced or avoided.

The invention discloses a pulsed eddy-current-based metal plate thickness information detection method. The method comprises the following steps of S1, testing the magnetic induction intensity Bair inthe air by means of a probe; S2, measuring the magnetic induction intensity Bref of a standard test piece of the same material as a tested piece and adopting the measured result as a reference signal; S3, detecting the test piece of the same standard material and obtaining a parameter (imgfile = 'DDA 00014861076800000000011.TIF 'wi = '155 'he = '70 '); S4, determining a constant C, and calculating a characteristic parameter KB value; S5, calculating the value of the frequency f of the corresponding KB value by calculating a formula of the characteristic parameter KB value and an additional formula in the step S4; S6, taking the frequency f, the constant C and the parameter (img file = 'DDA 00014861076800000000012. tif'wi = '131 'he = '70 ') into the formula, and calculating to obtain thethickness value h of the tested piece. By adopting the method, the eddy current accurate detection is realized through eddy current signal coupling and eddy current signal decoupling. Meanwhile, the experiment method is simplified, and the research timeliness is improved.

The invention provides a high-torque valve torque monitoring system. A low torque worm couple and a torque amplifying mechanism are additionally arranged between a motor and a high torque worm couple, a method of reducing speed and increasing torque is adopted to drive a high-torque valve, the torque of valve passes through the device, speed increase and torque reduction are reflected by axial movement of the worm shaft of the low torque worm couple, and axial displacement of the worm is detected, and then linear transformation is carried out, so as to calculate the actual torque of the valve, thus the detection method is not only accurate, but also ensures that the cost for detecting high torque is reduced. Besides, the wireless communication way is adopted to exchange data with a personal computer (PC) or notebook computer, remote real-time action control on the valve can be realized, and the dynamic torque, stroke and related alarm information of valve can be monitored in real time. The system carries out real-time quantitative analysis and complete detection analysis on the torque of the valve in the operational process, and complete data basis is provided for improving the valve torque design level and the product quality level and prolonging the service life.

The invention belongs to radar remote sensing or image processing technology, namely, analyzing radar observation information by using image processing technology, and specifically relates to a high resolution SAR image target detection method based on a part model. According to the invention, through performing characteristic compression and screening on each pixel point of original SAR image data, low-level characteristic interest points are extracted, a target part initial unit is generated from interest point local information, calculating initial unit characteristics, generating a target part combination model through an image segmentation method and realizing high resolution SAR image target detection by utilizing description of the target part model. According to the invention, distribution characteristics of a target in the high resolution SAR image are utilized fully and effectively. The target part model obtained through detection is comparatively complete, so that the integral information of a target can be kept well. The method is good in noise robustness, high in algorithm stability, accurate in detection result and can improve the self-adaptive capability of image detection.

The invention discloses a bicycle power generation bodybuilding system, which is characterized by comprising a friction generator for outputting alternating current through friction between the generator and a bicycle wheel, a power supply processing module for providing a power supply and signal processing function for the system, and a display module for displaying the value of calories consumed by a user who rides a bicycle and riding parameters in real time. The system can quickly and accurately detect and display the riding parameters and is high in intelligent degree, and complete parameters are detected.

The invention discloses a method for analyzing cathinone, methcathinone and 4-methylmethcathinone in a biological sample by liquid chromatography-mass spectrometry, which comprises the following steps: (1) preparation of sample solution; (2) preparation of standard working solution; (3) setting of detection conditions; (4) determination by liquid chromatography-mass spectrometry; and (5) experimental result calculation. The biological sample is pretreated to implement efficient and accurate detection on the cathinone, methcathinone and 4-methylmethcathinone, thereby providing technical supports for conviction and sentencing of the judicial department.