52 results about "Automatic target detection" patented technology

The invention discloses a method for extracting target closed contour based on shape prior, belonging to the field of computer application technology. From the point of view of simitating the information of human vision, the extraction method inhibiting salient edges of noise and texture is adopted, the shape prior is blended on the basis, and a new target closed contour extraction algorithm can be provided by utilizing curve evolution technology based on differential geometry, so that the important technology is provided for automatic target detection. The experiments on a plurality of imagelibraries prove that when the image background is very disordered and has partial shading, and part of the target boundary is indistinct caused by weaker light, the method can obtain the whole targetclosed contour. Simultaneously, when only a few of templates of a certain category of objects are provided, the method can extract the contour of the category of objects with different postures, namely, the method has invariance property for elastic deformation to a certain extent, so as to be directly applied to intelligent image processing. The method can accurately extract the target closed contour and has wide applicability.

The invention relates to a remote sensing image target detection method and device. The method comprises the following steps: firstly, establishing a depth full-convolution neural network; the depth full-convolution neural network comprises a main network used for feature extraction and a sub-network used for target detection; training the main network through an image classification database; after the main network is trained, adding the sub-network; fine-adjusting the network parameters by means of the remote sensing target detection database to obtain a trained depth full-convolution neuralnetwork; and finally inputting the remote sensing image to be detected into the depth full-convolution neural network to obtain a final target detection result. Compared with a conventional way of directly using a remote sensing target detection database to train the established depth all-convolution neural network, the generalization ability and accuracy can be improved, and the rapid, high-precision and high-robustness automatic target detection can be realized in a complex background.

Methods and apparatus are provided for recognizing particular objects of interest in a captured image. One or more salient features that are correlative to an object of interest are detected within a captured image. The captured image is segmented into one or more regions of interest that include a detected salient feature. A covariance appearance model is generated for each of the one or more regions of interest, and first and second comparisons are conducted. The first comparisons comprise comparing each of the generated covariance appearance models to a plurality of stored covariance appearance models, and the second comparisons comprise comparing each of the generated covariance appearance models to each of the other generated covariance appearance model. Based on the first and second comparisons, a determination is made as to whether each of the one or more detected salient features is a particular object of interest.

The invention discloses an ultrasonic targeted electric flower and fruit thinning machine, which comprises a working frame, a rotating shaft, a flower and fruit thinning adhesive tape set, an ultrasonic sensor set, a flower and fruit thinning motor, an angle adjusting mechanism, a telescopic rod and a control mechanism, wherein the working frame is in a frame structure with one open side; the open side of the working frame is provided with the rotating shaft; the flower and fruit thinning adhesive tape set is arranged on the rotating shaft; one end of the rotating shaft is provided with the flower and fruit thinning motor; the working frame is provided with the ultrasonic sensor set; the angle adjusting mechanism and the telescopic rod are arranged on the end face below the open side of the working frame respectively; and one end of the telescopic rod is connected with the working frame, the other end of the telescopic rod is a handheld end, and the handheld end is provided with the control mechanism. The flower and fruit thinning machine can realize automatic target detection, and has wide operation range, humanized design, high operation efficiency and strong fruit tree applicability.

The invention discloses an automatic video-target detecting and tracking method based on motion information and sparse projection. The automatic video-target detecting and tracking method is used for solving the technical problem that an existing automatic target detecting and positioning tracking method based on forward-backward motion historical images is poor in tracking accuracy. According to the technical scheme, firstly, by means of forward images and the motion information of the forward images, detection of a moving object on the images is finished automatically; secondly, the detected moving object on the images is tracked in real time in a sparse projection mode; finally, by means of statistics of motion trail information, a tracing result of a real moving target is analyzed, and influences from noise targets are removed. Tests show that the detecting and tracing result accuracy reaches more than 85%.

The invention discloses an oil and gas pipeline inspection method based on an unmanned aerial vehicle, and the method comprises the steps: carrying a high-resolution imaging system and a high-precision positioning device by the unmanned aerial vehicle to carry out low-altitude aerial photography, and obtaining continuous oil and gas pipeline images and the position information of the unmanned aerial vehicle; Performing abnormal target detection by adopting a target detection algorithm based on a neural network to obtain an abnormal target; And carrying out abnormal target positioning by adopting an algorithm based on a motion recovery structure. The oil and gas pipeline inspection method provided by the invention is relatively high in intelligent degree, an automatic target detection algorithm is adopted, and target positioning can be realized without adding additional ground control points; The oil and gas pipeline inspection method provided by the invention is relatively high in efficiency; In the abnormal target detection stage, the possible position of an abnormal target is effectively determined through an image segmentation algorithm based on SIFT and Graph Cut, and in the abnormal target positioning stage, under the condition that dense reconstruction, DEM generation and other steps in a traditional algorithm are omitted, high positioning precision can still be obtained.

The invention discloses a superpixel local information measurement-based polarized SAR ship target detection method, and mainly aims at solving the problem that the target detection rate lower complicated scenes is low. The method comprises the following steps of: 1, carrying out superpixel segmentation on an original image so as to obtain superpixel segmentation results under different scales; 2,for the segmented results, calculating three superpixel level-based difference measurements by utilizing a sliding window model; 3, converting the superpixel level-based difference measurements intopixel level-based difference measurements; 4, mapping pixel-level difference measurement vectors into different measurement values by utilizing kernel fisher discrimination, so as to a difference measurement value of each pixel point; and 5, classifying the difference measurement values of the pixel points by utilizing a linear SVM classifier, determining a category of each pixel and carrying outautomatic target detection. The method is capable of enhancing the target detection performance under complicated scenes an realizing the automatic detection process, and can be used for subsequent ship target discrimination, recognition and classification.

The invention mainly belongs to the technical field of target invasion detection and particularly relates to an unmanned aerial vehicle image-based region invasion target detection method. The method is used for target detection of an unmanned aerial vehicle. The method comprises the steps of performing Gaussian pyramid layering on an original image in an original video obtained by a vehicle-mounted camera of the unmanned aerial vehicle to lower calculation complexity of feature point extraction; and then extracting SIFT feature points of the image to perform image registration, capturing motion information in the image by adopting LK sparse optical flow of a pyramid to realize target point motion calculation, performing motion point clustering and false target removal, and finally performing target judgment to realize target detection. According to the method, the search range of inter-frame feature points of the unmanned aerial vehicle video can be reduced, the motion large-displacement problem in the unmanned aerial vehicle image is solved, and the detection capability is improved; and therefore, the manual detection intensity of region invasion is reduced and the automatic perception capability of the unmanned aerial vehicle is improved.

It is morphology filter automatic object detection method, which comprises the following steps: first to collect training samples to form heredity arithmetic for optimization train, wherein, the arithmetic adopts new range discretization code and self-adapting major/minor cross and dissociation operator; to use collected sample to train the structure element value by heredity arithmetic to form morphology filter based on Top-Hat operator; to filter the infrared object image; finally to do the division based on self=adapting threshold to detected multiple weak spots; to detect the object spots with high signal to noise rate by use of fixed threshold division.

The invention discloses a motorized station automatic target detection and spatial positioning method based on YOLO-SIOCTL. The method comprises the following steps that: a mobile station acquires hardware configuration; The carrier platform of the vehicle-mounted mobile measurement system carries out image acquisition in the process of driving at a normal speed; Realizing geographic object automatic target detection based on a YOLO algorithm; back calculation of a geographic pixel frame of the image pixel frame is cahieved based on an SIOCTL algorithm; and a single-object multi-space positionpoint is fitted with an optimal position point. The method can be widely applied to the field of patrol supervision, the workload of manual target checking is effectively reduced, the patrol efficiency is improved, the consumption of manpower and financial resources is reduced, and the method has a good application value in practical application.

The invention provides a line-of-sight artillery target-scoring system and method. The system comprises a laser detection automatic target detection module, a control module, a target-scoring module and a wireless communication module. The laser detection automatic target detection module is used for forming a laser detection fan. According to the line-of-sight artillery target-scoring system andmethod, a fan light curtain is formed through two laser beams swinging at a high speed, and when pellets penetrate through the fan light curtain, the pellets can be irradiated by the laser beams and reflection waves can be formed. A laser receiving probe set is arranged below a target surface; and when the laser receiving probe set catches the pellet reflection waves, the system can calculate theair positions of the pellets according to rotation angles, detected by angle sensors connected with two laser emission devices, of the laser beams and the relative distance between two laser devices,and the aim of line-of-sight artillery live-ammunition firing precise target scoring is achieved.

The invention discloses a THz-QoMIMO framework suitable for a Terahertz security check instrument, comprising signal processing and target recognition digital hardware, Terahertz MIMO array front end high frequency hardware, and one-dimensional quasi-optics focusing and scanning, and employing an MIMO array layout. By utilizing MIMO synthetic focusing, a system frame rate is increased, and sending and receiving channels are reduced so as to reduce system costs; in addition, the THz-QoMIMO framework can simplify a system whole machine structure, improve system work reliability, perform repeated multi-angle measurement, reduce coherent speckle influence, and facilitate automatic target detection and recognition; besides, the THz-QoMIMO framework employs an image classification recognition method with distance information z as the primary and intensity information I as the complement, and is especially suitable for an active Terahertz rapid security check instrument.

In some embodiments, a device may include a laser rangefinder circuit including an emitter configured to direct a focused beam toward a view area and including one or more optical sensors configured to receive light including ambient light and reflections associated with the view area. The device may further include a processor coupled to the laser rangefinder and configured to determine an edge of an object within the view area based on measurements of the reflections.

The invention discloses an intelligent image processing device, which comprises an image sensor, a control circuit, a processor, a power supply and an interface circuit. Image signal transmission between the image sensor, the control circuit, the processor, the power supply and the interface circuit is bidirectional. The processor is used for performing calculation on received image data so as to achieve automatic exposure, automatic focusing, automatic white balance, automatic tracking, automatic target detection and automatic image intensification. Namely, the multiple functional modules of the intelligent image processing device are integrated and the signal transmission relation between the units is fully utilized so as to achieve advanced intelligent judgement about scenes or statuses, such as the functions of target detection, recognition, tracking and the like, accordingly the image processing device is more intelligent, and the intelligent level of the image processing device is raised.

The invention discloses a YOLO-SITCOL-based automatic target detection and spatial location method of a fixed station. The method comprises the following steps: configuring acquisition hardware of thefixed station; acquiring an image; based on a YOLO algorithm, realizing automatic target detection on a geographic object; based on an SITCOL algorithm, realizing reverse calculation on a geographicpixel frame by virtue of an image pixel frame; and fitting the best position point by virtue of multiple spatial positions of a single object. The method disclosed by the invention effectively reducesworkload of manually viewing a target, improves inspection efficiency, reduces consumption of manpower and financial resources and has a better application value in practical application.

The invention discloses an automatic detection and tracking method based on target adaptive projection. The method comprises the following steps: reading to-be-tracked video sequence information, andperforming RGB-to-Gray color space conversion on each frame of image; solving a foreground target mask in the first M frames by utilizing a weighted sliding variance method; adaptively selecting the maximum mask of the foreground target area; carrying out horizontal and vertical projection on the mask; searching the projection result in the vertical and horizontal directions, and determining the center of the optimal target through accumulated pixels and the proportion; performing round-trip detection in the vertical and horizontal directions from the center position, determining the optimal target scale through accumulated pixels and the proportion, and thus determining an ROI (target area); and finally, tracking the target by using a centroid method to finish an automatic target detection and tracking process. The method is easy to implement and clear in process, and the influence of background noise can be effectively avoided in the ROI extraction process through the self-adaptive projection method.

The invention provides an automatic target detection platform comprising a data set management module used for creating and managing a data set, a model training module used for training a to-be-trained model according to the data set to generate a first target detection model, a model verification module used for carrying out online verification on the first target detection model so as to generate a second target detection model, and a model publishing module used for publishing the second target detection model. According to the automatic target detection platform, the technical threshold of a deep learning target detection algorithm can be greatly reduced, and the entry difficulty is reduced.

The invention discloses an automatic targeting detection device for agriculture. The device comprises an unmanned aerial vehicle assembly, a camera assembly, a light source assembly and a control unit; the unmanned aerial vehicle assembly comprises a rack, a connecting rod is arranged at one end of the top of the rack, a supporting rod is connected to one end of the bottom of the rack, and the supporting rod is connected with a protective rod; and the light source assembly comprises a light source generator, one end of the light source generator is connected with a storage battery, the other end of the light source generator is connected with a collecting lens, and the light source assembly is arranged on the inner side of the supporting rod and connected with the rack. The unmanned aerial vehicle is used for carrying out camera shooting detection, higher flexibility and convenience are realized. Compared with the cost other technologies, the cost of the device is lower; on one hand, a view blind area which is difficult to detect by other technologies (such as a remote sensing technology) can be detected, on the other hand, the illumination degree is freely adjusted according to a previous light condition, and the reliability during camera shooting is improved, and the accuracy of an analysis result is improved.

The invention belongs to the technical field of automatic target detection, and discloses an unmanned aerial vehicle target detection system, a target detection method and a computer readable storage medium, and the unmanned aerial vehicle target detection system comprises an unmanned aerial vehicle system which is used for carrying a photoelectric load system to an observation position and hovering at a fixed point; the photoelectric load system is used for executing a fixed-point observation task and performing fixed observation of a target range area; the wireless image transmission system is used for transmitting observation video images by adopting an HHLM type transmission module; the wireless remote control system is used for controlling the unmanned aerial vehicle system and the photoelectric load system; and the target detection comprehensive task station is used for receiving, storing, forwarding and processing video images. The marine target detection system provided by the invention is of great significance in improving the testing technology and testing equipment level and improving the actual combat ability and training level of ship gun shooting of a driving and protecting ship, and meanwhile, has very strong practical significance in guaranteeing the live shell shooting quality of the ship gun and adapting to battlefield requirements.