68results about How to "Implement object detection" patented technology

The invention discloses a control method of a monitoring ball machine. The method comprises the following steps of (1) vertically and horizontally dividing a to-be-monitored space into a plurality of small partitions, and setting a shooting focal distance for each small partition; (2) setting corresponding preset positions for a central point position and an edge point position of each small partition, and storing horizontal position information, vertical position information and shooting focal distance information of the ball machine; (3) reading a video frame, and performing target detection on the video frame; (4) according to direction information of a detected target in a monitoring scene, mapping the target to the corresponding preset position; (5) calling the preset position by the ball machine to acquire a monitored image. The defects that the automation degree of control is not high, the real-time property and the flexibility are not enough, and human manual interference is required in a ball machine of the traditional video monitoring system are overcome; the control method is convenient in operation, high in automation degree of control and good in instantaneity, and is particularly good in capture effect on the monitored image of a quickly moving target.

The invention discloses a method for realizing non-standard parking behavior identification by using a monitoring video. A high-point camera is combined with a GIS and a computer vision technology torealize rapid, accurate and real-time identification of various non-standard parking behaviors according to parking postures. The method mainly comprises the steps of constructing a non-standard parking behavior rule base, extracting spatialized parking space poses of a video scene based on an improved Vibe algorithm, an SSD convolutional neural network model and a CSRT tracking algorithm, and finally realizing accurate detection of non-standard parking behaviors in combination with a geometrical relationship between parking spaces and parking features. The invention further provides a devicebased on the method, and the invention is high in recognition precision and high in real-time performance.

The invention provides an anti-unmanned aerial vehicle detection tracking interference system and a photoelectric tracking system working method. The anti-unmanned aerial vehicle detection tracking interference system comprises a radar, a photoelectric tracking system, an unmanned aerial vehicle interference unit and a holder; the photoelectric tracking system comprises a motion detection module,a correlation filtering target tracking module, a deep learning target detection module and a deep learning target tracking module; the radar is in communication connection with the photoelectric tracking system; and the photoelectric tracking system is in communication connection with the holder. According to the anti-unmanned aerial vehicle detection tracking interference system, when target distance is farther, the deep learning target detection module cannot extract target characteristics, and target detection is carried out by using the motion detection module; when the target distance isfarther, under the condition that the deep learning target tracking module cannot extract target characteristics, target tracking is carried out by using the correlation filtering target tracking module; and the problem that the deep learning target tracking module cannot provide degree of confidence is solved by using data of the correlation filtering target tracking module.

The invention discloses a DOA estimation method in a co-prime array based on iteration sparse reconstruction. A receiving antenna array uses a nonlinear co-prime array, through vectorized processing on a second-order statistical characteristic covariance matrix of a received signal, and a difference array in larger aperture length can be determined, so as to improve detection capability. Dispersing processing is performed on the angle domain where targets are in, targets can be regarded as sparsely distributed on grid points or near grid points, and sparse signal reconstruction problems on logarithm and forms are established. Using convex compact upper bounds of logarithm and a function, an original sparse problem is reestablished, to dynamically adjust and update discrete points of the angle domain in an iterative manner, so approach the actual arrival angle of the target.

The invention discloses a ground penetrating radar target detection method based on a full convolution network, which comprises the steps of building a three-layer full convolution network to train aground penetrating radar data set, scaling an image to obtain different scales, then inputting into the network for convolution operation, outputting a heat characteristic map, performing mapping calculation on the heat map, and positioning the location of a target so as to complete the target detection. The network does not need to use a data set marked by a location box when being trained, can accept input pictures of any size, detects targets of different sizes and is high in speed. In the case of a small data volume of the ground penetrating radar, ground penetrating radar target detectionbased on the full convolution network is realized through data expansion. The algorithm has the advantages of high speed, high detection accuracy and the like.

The invention is suitable for the radar signal processing and array signal processing technology field and provides a space-time adaptive processing method based on a co-prime pulse recurrence interval and an apparatus thereof. The method comprises the following steps of step S1, receiving pulses emitted by using a co-prime way and using time delay among the received pulses to construct a virtual pulse Ps; step S2, according to the constructed virtual pulse, constructing a virtual snapshot xv; step S3, using the constructed virtual snapshot to estimate a covariance matrix RCPRI-SMI of clutters and noises; and step S4, using the acquired covariance matrix to design a space-time filter so as to carry out clutter suppression. By using the method and the apparatus of the invention, a distance and doppler ambiguity can be solved to some extent, a capability of electronic interference resistance can be increased and simultaneously flexibility of a system can be increased too.

The invention discloses a moving target tracking method based on multi-target characteristics and an improved correlation filter. The method comprises the following steps: inputting position information of a tracked target in a tracking video sequence and an initial frame; extracting multi-channel features of the target to achieve comprehensive information representation of the target; constructing a pixel reliability graph to perform constraint optimization on a correlation filter, and limiting the correlation filter in an image area suitable for tracking; reducing the number of parameters inthe model by using a linear dimension reduction operator, and training a compact sample classification model; performing secondary optimization on the correlation filter through a Gauss-Newton methodand a conjugate gradient method to obtain an optimal correlation filter; responding to the improved correlation filter and the extracted target features of the target search area, and determining theposition of a target tracking box; jointly updating the filter model and the pixel reliability diagram; and outputting a tracking result map. According to the method, moving targets in most scenes can be effectively tracked, and the method has good tracking precision and real-time performance.

The invention relates to a self-evolutionary radar target detection algorithm based on deep learning, and relates to radar signal processing. The self-evolutionary radar target detection algorithm based on deep learning comprises the steps of: performing feature processing of radar data; designing a radar target detection base model; and applying a two-view cooperative training algorithm to modelself-evolution. By adoption of a deep learning method, radar target detection is realized; simultaneously, self-evolution of the capability of the radar target detection model is realized through theprovided two-view cooperative training algorithm; the algorithm is established based on deep learning and machine learning; therefore, the radar identification capability can be improved; the practicability is high; the portability is high; reliable improvement of the detection capability is realized by sufficient utilization of the radar data flow; and requirements of semi-supervised learning lowin human loss can be realized.

The invention provides a constant false alarm detection method and device for radar detection, electronic equipment and a storage medium, and belongs to the technical field of radar signal processing, and the method comprises the steps: obtaining the sample data of at least one radar frame, and generating a corresponding distance-Doppler scattering center energy matrix; for each radar frame, based on the corresponding distance-Doppler scattering center energy matrix, determining a truth value of noise energy estimation of each distance unit, calculating a target noise energy estimation coefficient of each distance unit and a target detection threshold coefficient of each distance unit, and calculating a reference detection threshold value of each distance unit; and performing constant false alarm detection on the Doppler dimension scattering center of the radar detection data based on the reference detection threshold value of each distance unit. According to the invention, target detection under a non-uniform clutter distribution condition can be realized, and ideal detection performance can be achieved.

The invention relates to a system and method for establishing target division remote damage assessment of different vehicle types based an artificial intelligence radial basis function neural network method and belongs to the vehicle damage assessment field. The objective of the invention is to solve problems in target detection of collision vehicles after a vehicle collision. According to the technical schemes of the invention, a target detection subsystem is adopted to judge collision objects in the vehicle collision; the target detection subsystem learns target training data so as to generate a target model, wherein the target model is built by adopting the radial basis function neural network method. With the system and method provided by the technical schemes of the invention adopted, target detection in the vehicle collision can be realized; and a machine learning method is used in the remote damage assessment technical field, so that the accuracy of judgment in a damage assessment process can be improved.

The invention provides embedded face recognition tracking device and method, belonging to the technical field of face recognition tracking. The embedded face recognition tracking device is composed ofa power module, an embedded processing module and a camera interface module, wherein the output terminal of the power module is connected with and supplies power to the embedded processing module andthe camera interface module respectively; and the embedded processing module comprises a storage module, a first partitioning module, a detection module, a second partitioning module and a recognition module. Based on research and development in an embedded environment, a control algorithm is realized through software programming, only procedure codes need to be modified when the algorithm needsto be modified, and a hardware circuit does not need to be modified, so that flexibility is high, once an algorithm procedure is fixed in a microprocessor, and algorithm structure and performance cannot be changed; meanwhile, a target detection, recognition and tracking algorithm can be conveniently realized in the microprocessor, so that system reliability is greatly improved.

The invention provides a radar target detection method, which comprises the steps of generating a training sample: the training sample comprising a plurality of sample pictures, each sample picture being a labeled radar channel amplitude and phase diagram, and the radar channel amplitude and phase diagram being formed by mapping a single-frame radar echo; training the model by using the training sample to generate a training model; processing the radar data to generate a to-be-detected picture; and inputting a to-be-detected picture to the training model, and detecting a target position and a target classification through the training model. According to the method, the training samples can be synchronously and continuously supplemented, new sample training is carried out on the trained model through transfer learning, continuous iterative updating of the model is realized, and thus the target detection capability of the radar can be continuously improved.

The invention relates to a system for detecting a sea target by using a robust intelligent radar. The system comprises a radar, a database and an upper computer, which are orderly connected, wherein the radar is used for irradiating a sea area to be detected and storing radar sea clutter data in the database; and the upper computer comprises a data preprocessing module, a robust forecast model modeling module, an improved intelligent optimizing module, a target detecting module, a model updating module, and a result display module. The invention also provides a method for detecting the sea target by using the robust intelligent radar. The invention provides the system and method for detecting the sea target by using the robust intelligent radar which has advantages of robustness, capability of avoiding artificial factors, and high intelligence.

The embodiment of the invention relates to the technical field of image processing, and provides a monocular camera ranging method and device, and the method comprises the steps: collecting original image data of a target scene through a target camera; performing target detection on the original image data to obtain a target object in the original image data; performing target tracking on the target object, and obtaining a motion trail of the target object on the original image data; and converting the motion trail of the target object on the original image data into an aerial view, and obtaining the actual distance between the target object and the target camera according to a preset conversion ratio of aerial view pixels to the actual distance. According to the scheme, the cost is low, target detection, target tracking and target distance measurement can be achieved only through the camera at the intersection, and due to the fact that the target camera is static at the target intersection, the accurate monocular distance measurement effect can be achieved through simple calibration.

The invention discloses a signal processing method of a variable-frequency compressed sensing radar, which comprises the steps of distinguishing pulses with the same remainder from pulses with different carrier frequencies according to a transmission sequence, and obtaining a plurality of radar echo signals with a single carrier frequency; performing sparse recovery processing on each group of echo signals with the same carrier frequency; carrying out post-processing on reconstruction processing results of the echo signals with different carrier frequencies; and voting by taking the data in the interval near the maximum value to obtain a final restored signal. According to the method, a random modulation frequency conversion radar target detection model based on compressed sensing is established by combining the thought of constant false alarm detection and compressed sensing radar signal reconstruction of a traditional radar. Meanwhile, a weighting algorithm of multiple paths of signals is provided by utilizing the random characteristics of the transmitted signals, so that a final restored signal is obtained, and the target detection of the variable-frequency compressed sensing radar under multiple targets can be realized.

The invention discloses an optimization method of a Tiny-YOLO network for detecting a ship target on a satellite. The method comprises the following steps: employing a sample set of a ship image, carrying out the training of an original Tiny-YOLO network, and obtaining the parameters of a convolution kernel in each convolution layer in the network; determining a Tiny-YOLO network for ship target detection according to the original Tiny-YOLO network and the parameters of the convolution kernel in each convolution layer in the network; according to the method, the Tiny-YOLO network is sparsifiedby reducing convolution kernels, and transfer learning is carried out according to the position of each convolution layer of the sparsified Tiny-YOLO network, so that the operation speed and the detection accuracy of the sparsified Tiny-YOLO network on a satellite meet the requirements; and the convolution kernel parameters in the Tiny-YOLO network after transfer learning are converted into integers from floating-point numbers, so that a final Tiny-YOLO network can be obtained, and therefore, the requirement of improving an operation speed by using the improved Tiny-YOLO network on a satellite can be satisfied.

The invention discloses a multi-linear-array scanning and area array staring integrated space optical camera, relates to the field of moving target detection, recognition and tracking, and aims to obtain a plurality of detection images in a short time period to perform target detection and enable that target detection and tracking are highly consistent in the space reference. A multi-linear-arrayscanning capture channel comprises a scanning mirror and single-spectrum-band detector groups of at least two spectrum bands; each single-spectrum detector group comprises more than two linear array detectors with parallel linear arrays; the scanning mirror has a certain scanning speed, so that an object space moving target is imaged on the different linear detectors in sequence, and an object space image is obtained; and the obtained object space image is sent to an information processing module. An area array staring tracking channel comprises a two-dimensional pointing mirror and an area array detector, the two-dimensional pointing mirror points to a target position under control of the information processing module, and the area array detector is used for imaging the target position pointed by the two-dimensional pointing mirror. The information processing module acquires the object space image and performs target capturing, identification and tracking.

The invention provides a heavy haul railway corrugation recognition method, device and equipment based on target detection. The heavy haul railway corrugation recognition method comprises the steps that the time domain waveform of the acceleration of a wheel-rail contact surface when a heavy haul train passes through a heavy haul railway is sampled; coloring the time domain waveform according to an acceleration threshold to obtain a colored image; importing the colored image into a corrugation recognizer, and determining corrugation of the heavy haul railway; wherein the corrugation recognizer is obtained by training a Faster-RCNN model, coloring of a time domain waveform is realized through an acceleration threshold value, the colored image can be caused to the corrugation recognizer, corrugation on the heavy haul railway is recognized through the color of the colored image, target detection different from a deep learning convolutional network is realized, and the detection accuracy is improved. And the corrugation identification accuracy is improved.

The embodiment of the invention provides an intelligent detection method and system for cracks around a transverse hole. The method comprises the following steps: acquiring a to-be-detected traverse hole ultrasonic image; inputting the ultrasonic image of the to-be-detected transverse hole into a pre-trained crack detection model to obtain a crack detection result output by the crack detection model; wherein the crack detection model generates an anchor frame through a K-Means + + algorithm and a K-Mediods clustering algorithm for the transverse hole crack data set, and the category information and the position information of the transverse hole crack data set are obtained through training and testing of a YOLOV3 algorithm. According to the embodiment of the invention, the method achievesthe quick and accurate recognition of the crack defects, achieves the higher accuracy while achieving the target detection, meets the requirements of real-time detection, and is more suitable for an application environment of ultrasonic nondestructive detection.

The invention discloses a 3D radar life detecting and positioning device with high resolution and high penetrability. The device comprises an emission subsystem, a reception subsystem, a multichanneldata collection system, a master control unit, a wireless display control terminal and a power supply module; the nanosecond level high-amplitude pulse emission subsystem is composed of a marx pulse generation circuit, three avalanche triodes and a resonant trigger circuit; echoes are received by the reception subsystem, processed by a noise reduction unit, a filtering unit and a signal amplification unit, and sent to the multichannel data collection system for sampling, a sampling result is input and output to the master control unit in which a Linux system is operated by a DSP data processing module and a WiFi communication in parallel, and bidirectional data communication of the wireless display control terminal is realized by 2.4GWiFi. Thus, the device can realize 3D life detection andpositioning, pulses are prevented from distortion, 3D simulated images are clearer, and received weak signals are stronger.