57results about How to "Improve target detection accuracy" patented technology

The invention belongs to the technical field of radar image processing, and particularly relates to a ship target detecting and discriminating method in an SAR image with a complicated background. The method comprises the following main steps of (1), fine sea-and-land dividing; (2), performing high-efficiency detection on the ship target, wherein the step comprises large-scale CFAR and small-scale iteration CFAR, wherein a synthetic aperture radar image clutter statistics distribution model based on generalized Gamma distribution is used; and (3), performing nearshore target false alarm discrimination, wherein the step comprises a false alarm discriminating algorithm based on a maximal likelihood and a false alarm discriminating algorithm based on polarization information. The method can efficiently and accurately detect the ship target in complicated backgrounds such as nearshore and harbor, and furthermore can utilize the false alarm discriminating algorithm based on maximal likelihood and the polarization information for discriminating a false alarm target, thereby improving ship target detecting accuracy. The ship detecting algorithm provided by the invention is suitable for a random SAR image background and furthermore has advantages of high robustness, high real-time performance and good popularization prospect.

The invention discloses a high-resolution remote sensing image weak target detection method based on deep learning. For a remote sensing image with low resolution, a small target size and fuzzy quality, the method comprises the following steps: firstly, improving the resolution of an image by adopting a WGAN-based super-resolution reconstruction method; inputting the image with the enhanced quality into a target detection framework; carrying out deep feature extraction on the image by using a residual network; fusing the extracted low-level features with the extracted high-level features; it is ensured that the fused multi-layer feature map has rich detail information and also contains high-level semantic information; and carrying out region-of-interest coarse extraction on the feature mapby using the fused multi-layer features and the region suggestion network, mapping the extracted region to the same dimension by using a region-of-interest alignment method, and carrying out subsequent target accurate classification and position refinement to obtain a final target detection result. According to the method, the weak and small target detection precision and recall rate under the conditions of low remote sensing image resolution and complex background are effectively improved.

The invention discloses a high precision direction measurement method for linear arrays, comprising: forming frequency domain traditional wave beam for the two-dimensional space-time signals received by a linear array; scanning in the wave beam domain to obtain a target angle measured roughly; dividing the linear array into a plurality of sub arrays; in the target angle range measured roughly, processing frequency domain traditional wave beam formation and focusing on the space-time two dimension signals received by each sub array to form frequency domain-wave number data; in the target angle range measured roughly, processing the frequency domain-wave number data according to a wideband focusing minimum variance distortion-less method, to obtain the direction of target signal. The invention has high calculation speed, can realize real-time processing, has high algorithm robustness, and has high target detection precision.

The invention discloses an unmanned aerial vehicle video small-object detecting method based on super-resolution reconstruction. The unmanned aerial vehicle video small-object detecting method comprises the steps of selecting one input image as a reference frame, and selecting three consecutive image frames for performing sub-pixel displacement estimation on the reference frame; then placing the displacement estimation result of the four image frames into a high-resolution image grid; and estimating pixels which are lost in the high-resolution image grid, thereby obtaining an objective image with a relatively high resolution. Afterwards, an objective template is extracted from the objective image, and the characteristic of the objective image is calculated. Then the reconstructed objective image is divided for obtaining a plurality of objective area blocks. Characteristic extraction and characteristic identification are successively performed on all objective area blocks, thereby finishing preliminary detection for the object. Afterwards, false object elimination is performed, thereby obtaining a final detection result.

The invention discloses a traffic flow visual inspection method. Traffic scene image sequences are acquired through electronic monitoring in real time, a traffic scene weather condition environment is sensed by an illumination intensity sensor, a temperature sensor and a humidity sensor, daytime/night, fine day/rainy day/normal day and other weather conditions are judged, and illumination and shadow pretreatment is performed correspondingly. Running distances are restrained according to traffic rules, meanwhile lane departure phenomenon caused by overtaking and passing-by in the actual vehicle running is considered, double virtual lines are arranged at the same horizontal positions of all of lanes within a monitoring range in images, vehicle positions are quickly detected and positioned by utilizing double-template matching convolution in a double virtual line detecting region, 'one-to-more' and 'more-to-one' phenomena are eliminated, and wrong detection and wrong judgment are decreased. Intervals of vehicles in the horizontal direction and the vertical direction are judged and identified, vehicle target positions are restrained according to the horizontal and vertical position information of the vehicles and correctly positioned, normally running vehicles are counted, traffic flow statistic is performed, and the problem of inaccurate traffic flow counting is solved. The traffic flow visual inspection method has high detecting accuracy and good anti-interference performance and real-timeliness.

The invention belongs to the technical field of photoelectric detection, in particular to a distance measuring device and a method thereof. The method comprises the following steps of: assembling a large field by using a 2*3 array of multispectral cameras, obtaining background images and labeling characteristic reference points by using real-time image arrangement, and calculating an object distance by using an image convergence shooting algorithm. Because the invention adopts a large-field angle, observed objects are more easily found, and judging whether the observed objects are the same one is more accurate; and because the characteristic reference points are selected after image arrangement, the data volume of object calculation is smaller, and calculation is faster.

The invention relates to a towed linear array device based on attitude real-time measurement vector hydrophones. The towed linear array device comprises a switching section, a first vibration damping section, an acoustic array section, a second vibration damping section, and a tail rope section which are arranged successively. The switching section is connected with the towing cable of a mother ship. The towed linear array device is characterized in that the acoustic array section comprises attitude real-time measurement vector hydrophones, submarine acoustic signal amplifying and filtering modules, 16-channel acquisition and transmission modules, depth sensors, and a data transmission control module; and the attitude real-time measurement vector hydrophones comprise attitude sensors. The towed linear array device uses the attitude real-time measurement vector hydrophones, prevents a digital section of a conventional towed linear array, and is simple in structure, capable of distinguishing the larboard and the starboard of a target in real time, and really improving target detection precision by means of acoustic signal attitude correction.

The invention discloses a weak and small object detection method in a visible image of an unmanned plane, and mainly solves a problem that the scaling and rotation of shape information in the prior art causes the difficulty in target detection. The method comprises the following steps: (1) inputting an unmanned plane image frame containing a plurality of man objects, extracting two image blocks from the image as target templates, wherein each image block contains one man object; (2) calculating mean values mu, standard deviations sigma and entropies H of the target templates; (3) carrying out the enhancement of the inputted image, segmenting the enhanced image according to the color information, and obtaining a plurality of super-pixel blocks; (4) sequentially carrying out the feature extraction and recognition of all unmanned plane, and completing the primary detection of targets; (5) carrying out the removing of a false target for the image after primary detection, and obtaining the final detection result. The method effectively improves the recognition accuracy of weak and small objects in the visible image of the unmanned plane, and can be used for the visible image or video of the unmanned plane.

The invention discloses a small-size intelligent oceanic earthquake and electromagnetic data acquisition system. The small-size intelligent oceanic earthquake and electromagnetic data acquisition system comprises an intelligent acquisition station arranged underwater and two pairs of electric field detection devices installed at the exterior of the intelligent acquisition station; an ultra-short baseline transponder, a computer control system, a four-component earthquake data sensor unit, a magnetic field sensor and a three-component attitude sensor are arranged in the intelligent acquisition station; the output end of the four-component earthquake data sensor unit is connected with the four-component earthquake data acquisition unit; the output end of the magnetic field sensor is connected with a magnetic field data acquisition unit; and an electric field data acquisition device, the magnetic field data acquisition unit, the three-component attitude sensor, the ultra-short baseline transponder and the four-component earthquake data acquisition unit are all connected with the computer control system. With the small-size intelligent oceanic earthquake electromagnetic data acquisition system of the invention adopted, oceanic earthquake data, oceanic magnetotelluric data and oceanic controllable source electromagnetic data can be acquired simultaneously, and the data acquisition quantity of one-time construction can be improved manyfold, and target detection accuracy can be improved effectively.

The invention discloses a target detection method, a target detection device, equipment and a storage medium. The method comprises the steps: obtaining video data, carrying out the preprocessing of afirst image sequence of the video data, obtaining a second image sequence with a background image removed, inputting the second image sequence into a trained detection model for target detection, andobtaining a target detection result. On one hand, only a foreground target is reserved for an image with a background removed, interference of other background images is avoided. A detection model pays more attention to the foreground target during learning and reasoning. Therefore, the target detection accuracy can be improved. On the other hand, since the background pixels of the input image areremoved, only the foreground pixels are seen by the detection model and are not influenced by a video or picture sequence scene at all, so that the scene migration performance of target detection isimproved.

The invention discloses a few-sample target detection method based on meta-feature and weight adjustment and a network model. The method comprises the following steps: S1, constructing a detection network model and preprocessing an image; s2, extracting meta-features and weight vectors of the base class images; s3, combining the extracted meta-features and weight vectors to obtain a multi-dimensional feature map, and inputting the multi-dimensional feature map into a classification regression module to calculate a loss function; s4, adjusting network parameters according to the loss function and the gradient descent, and realizing training of a detection network model by the base class image; s5, extracting meta-features and weight vectors of the base class and new class joint images; s6,repeating the step S3 and the step S4, and training of the new class and base class combined image on the detection network model is completed; and S7, detecting the test image by using the trained detection network model. According to training of the detection network model, meta-features are extracted by using samples of a large amount of data, and fine adjustment is performed by means of few sample data, so that the target detection accuracy of a small amount of marked samples is improved.

A point target movement velocity detection method based on multiple linear moveout scanning, extending and sampling comprises the following steps: (1) establishing a multiple linear moveout scanning detection device; (2) using each linear detector to detect point targets in an extending and sampling manner; (3) processing Nt groups of image data acquired by each linear detector to obtain sub-pixel images; (4) performing non-uniformity correction and sub-pixel match on two sub-pixel images processed by the two adjacent processed linear detectors, then performing difference computation to complete the background subtraction; (5) performing threshold filtering on difference images processed by the two adjacent processed linear detectors, extracting positive and negative point pairs in the difference images by adopting the neighborhood constraint criterion, so as to complete the target detection and extraction of moving points in once scanning process; (6) marking regions of paired positive and negative points extracted in the step (5) respectively, calculating movement velocity and movement directions of the targets according to the position relation of the marked regions of the positive and negative points; (7) averaging the movement velocity and the movement directions obtained in the step (6) to obtain the velocity and the direction of a detected target.

The invention discloses a target detection method and device, electronic equipment and a storage medium, and the specific implementation scheme comprises the steps: responding to collection processing, and obtaining a to-be-processed picture containing a target text; inputting the to-be-processed picture into a pre-trained target detection network to obtain at least two candidate detection boxes containing the target text and a part of text in the target text, the part of text being a text region completely or partially overlapped with the target text and containing at least one row of text sequences; and performing operation according to the overlapping region of the at least two candidate detection boxes to obtain a classification index, and performing classification screening on the atleast two candidate detection boxes according to the classification index to remove the candidate detection boxes containing part of the text in the target text to obtain a target detection box. According to the invention, the accuracy of target detection can be improved.

The invention is suitable for the technical field of computer vision, and provides an image-based target detection method and apparatus. The method includes the following steps: generating a classifier of a cascade structure formed by multiple binary trees, the binary trees taking pixel intensity contrasts as features; conducting image traversal through sliding windows, and obtaining multiple window images; inputting the pixel intensity contrast features of the window images into the classifier; if there is one window image that passes the classifier and an output classification result is not less than a preset threshold value, determining that the window image includes a detection target; and if the classification result of the window image at any grade of the classifier is smaller than the preset threshold value, determining that the window image does not include a detection target. According to the invention, the detection rate of target detection for images is effectively increased, and the false detection rate is reduced.

The invention discloses a multi-aircraft cooperative detection and guidance integrated method and system. The method comprises the steps of obtaining the initial lateral distance between an interceptor and a maneuvering target, the initial lateral relative speed between the interceptor and the maneuvering target, the initial acceleration of the maneuvering target and the initial acceleration of the interceptor; determining an initial state vector according to the initial lateral distance, the lateral relative speed, the initial acceleration of the maneuvering target and the initial acceleration of the interceptor; determining measurement information according to the initial state vector and interceptor noise; based on the measurement information, obtaining an estimated state vector throughinteractive multi-model filtering; determining optimal control input according to the estimated state vector; and controlling the sight angles of the two interceptors according to the optimal controlinput to realize tracking interception of the target. Interactive multi-model filtering is introduced to recognize the switching time and state of target maneuvering, the sight separation angle of the two interceptors is modulated, the target detection precision is enhanced, and target tracking and interception are achieved.

The invention provides a hyperspectral image target detection method and system based on random forest measure learning. The method comprises the steps that firstly, n training samples are selected from a hyperspectral remote sensing image X to be detected; t then a mapping function is formed, the formed mapping function serves as input of a decision tree of the random forest to train the random forest, then the trained random forest is formed, average operation is conducted on all the decision trees, and a final measurement distance function is found; a measurement distance detection statistical value between all pixels on the X and a certain target pixel in the training sample are calculated by adopting the measurement distance; and finally, target detection is achieved according to thedetection statistical value. The method has the beneficial effects that the technical scheme provided by the invention has relatively good robustness, the high-dimensional data problem can be well solved, the target detection precision is improved, and the separation of the target pixel and the background pixel is better realized; and higher target detection precision is achieved.

The embodiment of the invention discloses a target detection method and device, electronic equipment and a storage medium. The method comprises the following steps: respectively acquiring first target data acquired by a millimeter wave radar and second target data acquired by an image sensor; performing time alignment and space alignment on the first target data and the second target data; carrying out target matching on the first target data and the second target data which are aligned in time and space; and determining a fused target detection result according to the target matching result. According to the technical scheme, by fusing the radar image of the millimeter-wave radar and the visual image of the image sensor, the omission ratio and the false alarm rate of the millimeter-wave radar are reduced, the position information, the target type and the like of the target are obtained, and the target detection accuracy is improved.

The invention discloses a commercial place cross-camera pedestrian trajectory tracking method. The method comprises the following steps: (1) target detection; (2) target feature extraction; and (3) cascade matching. According to the commercial place cross-camera pedestrian trajectory tracking method, GPU parallel computing characteristics are fully utilized, an appropriate data structure is organized, and the computing speed is effectively increased; priori information of a specific scene is fully utilized, a reasonable ID adding and deleting scheme is formulated, and the target detection accuracy is effectively improved. A discrimination model and a generation model are effectively unified into one framework, advantages of the two methods are complemented, and the multi-target tracking accuracy is improved.

The invention provides a seabed seismic data acquisition system and acquisition method based on distributed optical fiber sensing. An air gun source excitation ship is provided with a plurality of pull-type air gun sources; a plurality of boxes are distributed on the data acquisition armored optical cable at equal intervals, a surrounding optical fiber ring is arranged in each box, and an optical fiber attitude sensor is arranged at the top of each box; and the distributed optical fiber acoustic wave sensing modulation-demodulation instrument is arranged in a buoy on the sea surface, is connected to one end or two ends of the data acquisition armored optical cable, and is used for receiving optical fiber earthquake signals distributed along the data acquisition armored optical cable and signals of the optical fiber attitude sensors. The production and manufacturing cost of the seabed seismic data acquisition system is greatly reduced, the seabed seismic data acquisition system is convenient to use and maintain in offshore production, the seabed seismic data acquisition system can be made longer than a conventional active seabed seismic data acquisition cable, more seismic sensors can be arranged on each cable, and high-density seabed seismic data can be acquired more efficiently.

The invention provides a power transmission channel target object identification method and system. The method comprises the steps: acquiring an image containing a target object needing to be identified as a training set image; labeling different target objects on the training set images; constructing an image identification deep learning network, inputting the labeled images into the network, andsearching for optimal network parameters according to different target objects to complete image training; and transmitting the to-be-detected image to the trained image identification deep learningnetwork, and identifying a target object in the image. And the number of network layers is increased under the condition of maintaining a small calculation amount, so that the calculation speed is increased.

The embodiment of the invention provides a target tracking method. The method comprises the following steps: extracting target counting information, target detection frame information and target prediction frame information of each frame of image in a to-be-processed image sequence; calculating a first tracking trajectory of each target in the to-be-processed image sequence; judging whether a first missing detection condition exists or not; if the first missing detection condition exists, judging whether a second missing detection condition exists or not according to the target prediction frame information corresponding to the nth frame of image and the first missing detection target point corresponding to the (n + 1) th frame of image; if the second missing detection condition does not exist, determining first missing detection frame information according to the first missing detection target point; if the second leak detection condition exists, determining second leak detection frame information according to the second leak detection target point; and obtaining a target tracking trajectory based on the first tracking trajectory, the first missing detection frame information and/or the second missing detection frame information. The accuracy of multi-target tracking can be improved.