47 results about "Feature saliency" patented technology

The invention discloses an improved visual attention model-based method of a natural scene object detection, which mainly solves the problems of low detection accuracy rate and high false detection rate in the conventional visual attention model-based object detection. The method comprises the following steps of: (1) inputting an image to be detected, and extracting feature saliency images of brightness, color and direction by using a visual attention model of Itti; (2) extracting a feature saliency image of a spectrum of an original image; (3) performing data sampling and marking on the feature saliency images of the brightness, the color, the direction and the spectrum and an attention image of an experimenter to form a final rough set information table; (4) constructing attribute significance according to the rough set information table, and obtaining the optimal weight value of the feature images by clustering ; and (5) weighing feature sub-images to obtain a saliency image of the original image, wherein a saliency area corresponding to the saliency image is a target position area. The method can more effectively detect a visual attention area in a natural scene and position objects in the visual attention area.

The invention discloses a remote sensing image region of interest detection method based on integer wavelets and visual features, which belongs to the technical field of remote sensing image target identification. The implementing process of the method comprises the following steps: 1, performing color synthesis and filtering and noise reduction preprocessing on a remote sensing image; 2, converting the preprocessed RGB spatial remote sensing image into a CIE Lab color space to obtain a brightness and color feature map, and converting an L component by using integer wavelets to obtain a direction feature map; 3, constructing a Gaussian difference filter for simulating the retina receptive field of a human eye, performing cross-scale combination in combination with a Gaussian pyramid to obtain a brightness and color feature saliency map, and performing wavelet coefficient sieving and cross-scale combination to obtain a direction feature saliency map; 4, synthesizing a main saliency map by using a feature competitive strategy; and 5, partitioning the threshold values of the main saliency map to obtain a region of interest. Due to the adoption of the remote sensing image region of interest detection method, the detection accuracy of a remote sensing image region of interest is increased, and the computation complexity is lowered; and the remote sensing image region of interest detection method can be applied to the fields of environmental monitoring, urban planning, forestry investigation and the like.

The invention relates to a visual underlying feature-based image enhancement method. The method includes the following steps that: 1, the visual underlying features of an image are extracted; 2, weight fusion is performed on the visual underlying features of the image, so that a computing saliency map can be formed; 3, correlation coefficient comparison is performed on an eye movement saliency map measured by an eye tracker and the computing saliency map, so that the optimal weight of the visual underlying features can be determined; and 4, proper image enhancement methods can be selected for different types of images according to weight. According to the visual underlying feature-based image enhancement method of the invention, the eye fixation point-based eye movement saliency map is combined with a traditional image enhancement method for the first time, and an optimized visual underlying feature-based Itti visual attention model is adopted to extract the computing feature saliency map of the image, and therefore, the visual quality of color images re-displayed on a digital display device can be improved; and important information in the images can be highlighted according to different types of images, and less information is weakened, and therefore, a final image enhancement effect can more accords with eye visual perception. The method has a certain application value in image enhancement.

The invention discloses a visual attention mechanism-based method for detecting a target in a nuclear environment. The method includes: extracting brightness, color, direction features collected by an ordinary camera, and so as to obtain three feature saliency images; performing weighing fusion on the abovementioned saliency images to obtain a weighting saliency image; obtaining regions of interest according to the weighing saliency image and performing feature extraction on the region of interests; extracting features of a gamma camera mixed graph; and using an SIFT method to fuse the regions of interest with the mixed graph, and detecting the position of the target. The visual attention mechanism-based method for detecting the target in the nuclear environment uses a bottom-up data drive attention model to extract the plurality of regions of interests, and greatly reduces the calculated amount of a later matching process; then combines the plurality of regions of interest with the up-down task drive attention model to establish a bidirectional visual attention model, detection precision and processing efficiency of target areas in the images can be greatly improved, and a matching process eliminates interference of unrelated regions in a scene, thereby enabling extracted operation target to have better robustness and accuracy.

The invention belongs to the technical field of image processing and unmanned aerial vehicle detection, and relates to a method for detecting a small unmanned aerial vehicle target based on super-pixels and scene prediction. The method mainly comprises the steps of preprocessing, unmanned aerial vehicle target probability estimation and unmanned aerial vehicle detection, and is characterized in that in the step of preprocessing, super-pixel generation and scene classification are performed on an optical image to be detected so as to acquire a super-pixel based scene classification image; in the step of unmanned aerial vehicle target probability estimation, a significance depth value of each scene in the classification image acquired in the step a is respectively estimated, and the probability of existence of an unmanned aerial vehicle of each scene is calculated; and in the step of unmanned aerial vehicle detection, feature of the image to be detected are extracted, feature saliency maps are acquired by adopting an SVD based multilayer pyramid structure, weighting is performed on the different feature saliency maps to acquire a general saliency map, the general saliency map is loaded into the super-pixel classification image acquired in the step a, and a target detection result of an unmanned aerial vehicle is acquired according to a weight of the probability acquired in the step b when being applied to different scene areas by adopting a mechanism of winner-take-all and return inhibition. The method has the beneficial effect that the detection accuracy is higher compared with traditional technologies.

The invention provides a multi-feature-based saliency detection method. The method comprises the steps of firstly, performing calculation by applying three different saliency clues of multi-scale comparison, a center-surround histogram and color space distribution to obtain separate feature saliency maps; secondly, calculating weights of the separate feature saliency maps through Markov model learning, and obtaining an optimal solution of model parameter estimation by adopting a maximum likelihood estimation method; and finally, detecting a test image by utilizing a Markov model.

The invention provides a high-resolution SAR image vehicle target detection method fusing statistical significance, and the high-resolution SAR image vehicle target detection method comprises the steps: reading a high-resolution single-polarization SAR image, and extracting a region with a significant statistical feature in the image through a directed mean difference filter according to the imaging features of a vehicle body region and a shadow region in a vehicle target; fusing the original SAR data with the statistical feature saliency map so as to detect a vehicle target by adopting a deepneural network framework, and facilitating a target detection model to focus attention on an area with a significant statistical feature by fusing the features; training a target detection model, including selecting a training sample, and training to obtain a vehicle target detection model by taking the fusion feature as an input, wherein the target detection model is an improved YOLOv3 model, and the improvement mode is that an original SAR image and a directed mean difference filtering result are extracted at a network input end, and feature fusion is carried out; and performing vehicle target detection in the high-resolution single-polarization SAR image to be detected by using the trained target detection model.

The invention discloses an SAR target recognition method based on multi-level features, which comprises the following steps of: firstly, establishing pattern expression of SAR target image features, performing feature extraction to obtain feature vectors and a feature saliency map of a target, and fusing the extracted feature vectors to obtain shallow features; connecting an original target image and the extracted image feature saliency map according to a channel to obtain an input image, extracting features of the input image by using a deep convolutional network, extracting an output of a middle convolutional layer as a middle-layer feature, and extracting an output of a final full-connection layer as a deep-layer feature; performing further adaptive weight fusion on the obtained features of the shallow layer, the middle layer and the deep layer; and finally, carrying out classification identification on the fused features by utilizing a trained machine learning classification model to obtain a final identification result. According to the method, the respective advantages of the shallow, middle and deep features are combined, the synergistic effect of the shallow, middle and deep features is brought into full play, and the utilization capability of the multi-level features and the precision of target recognition are improved.

The invention discloses medical mask production equipment. The medical mask production equipment comprises a rack and a conveying device arranged on the rack. A synthesis module, a sewing module, a stacking module, an edge sealing module, a cutting-off module, a hot-pressing fixing module, a detection module and a sterilization module which are used for vertically stacking a plurality of cloth layers are arranged on the rack in the conveying direction of the conveying device. The detection module comprises a physical detection unit, a flexible detection unit and a display unit; the physical detection unit is in communication connection with the display unit; the physical detection unit comprises a camera and a processor; the processor comprises a preprocessing module used for carrying outnoise reduction on the gray image of the medical finished mask, a stroking extraction module used for carrying out edge detection on the image of the medical finished mask, and a feature saliency module used for detecting the image of the edge and improving the edge continuity. The invention further discloses a rapid quality detection method for the medical mask, and the method is rapid in detection, simple to operate and convenient to rapidly popularize and use.

The invention discloses a three-dimensional model fuzzy texture feature salification method based on conditional generative adversarial, and the method comprises the following steps: 1), generating the shape and texture of a three-dimensional geometric model, and obtaining the point cloud features of the three-dimensional geometric model; 2) extracting a fuzzy boundary point cloud; and 3) feature saliency of the fuzzy boundary based on the conditional generative adversarial network. According to the method, based on multi-scale voxelization fuzzy boundary frame selection, textures are generated through the conditional generative adversarial network for mapping, the extracted voxel blocks are embedded into the three-dimensional geometric model, global texture fidelity of the geometric model is achieved through a multi-scale voxel segmentation mode, global texture optimization is changed into local texture optimization, and the calculation amount is reduced.

The invention discloses a saliency target detection method fusing boundary priori and frequency domain information. The method mainly comprises the steps of: grouping pixels through the similarity of features of the pixels, and replacing a large number of pixels with a small number of superpixels to express the feature information of a picture; assuming, by a boundary priori model, that the boundary of the image is a non-salient region and takes the non-salient region as priori knowledge; selecting lBP texture features to measure texture differences between the superpixels so as to represent boundary features of superpixel blocks; converting a non-boundary region of the boundary prior feature map to a frequency domain through Fourier transform, so that detection of a frequency domain saliency region is realized; fusing a boundary prior saliency map and a frequency domain feature saliency map; and carrying out image enhancement on a primary saliency map by adopting homomorphic filtering to obtain a final saliency map with a clear salient region. According to the method, boundary information and frequency domain information are combined, and a salient region with a more complete boundary can be detected for a salient target in a complex scene.

The invention discloses a satellite in-orbit application-oriented remote sensing image visual salient region intelligent search method. The method comprises the following steps: S1, collecting a remote sensing image of a target through a satellite; s2, inputting the acquired remote sensing image into a visual saliency calculation model to obtain a frequency domain feature saliency map; s3, filtering the frequency domain feature saliency map based on the information entropy, dividing a valuable target area image, and removing pseudo target information of a remote sensing background; s4, takingthe valuable target area image as the input of an improved U-Net network model; and S5, searching a target area of the remote sensing image through training and processing of the improved U-Net network model. According to the invention, the visual saliency based on the remote sensing image is combined with the improved U-Net network model for carrying out the quick search of a valuable target region of the remote sensing image, and a saliency map extracted by the visual saliency calculation model is taken as the input of the adjusted U-Net network model, thus achieving the light-weight and quick region search of a valuable remote sensing target.

The invention provides a weak ship target fusion detection method and device based on multi-vision significant features, and the method comprises the following steps: carrying out the sea-land segmentation, cloud layer elimination and morphological filtering of a sea area sequence image obtained by an optical remote sensing staring satellite, and obtaining a sea area near-infrared image; based on the sea area near-infrared image, obtaining a target gray feature saliency map through a mean shift algorithm; obtaining a target point location distribution saliency map through a human vision system model based on the sea area near-infrared image; and taking the target point location distribution saliency map as guidance, performing image segmentation based on region growth on the gray feature saliency map, and extracting a ship target existing in the sea area near-infrared image in combination with ship target size features. The weak and small target detection method solves the problems that a weak and small target detection method in the prior art is high in missed alarm rate and false alarm rate, low in robustness and accuracy and not suitable for weak and small multi-target detection under the conditions of low signal-to-noise ratio and ocean clutter.

The invention discloses an image feature extraction method and device, and the method comprises the steps: S1, generating an initial feature map based on a convolutional neural network and a to-be-detected target image; S2, inputting the initial feature map into an initial module to further learn high-order abstract feature information to obtain an enhanced feature map; and S3, based on the enhanced feature map, generating a feature saliency map with the same length and width as the enhanced feature map by using supervised learning; and S4, performing normalization processing on the feature saliency map; and S5, multiplying the initial feature map by the normalized feature saliency map to obtain a purified feature map. The noise in the feature map can be inhibited, the target information is highlighted, and the small target detection capability of the model is improved.

The invention provides an SAR image salient target detection method based on multi-scale superpixels, and belongs to the field of signal processing, in particular to the field of synthetic aperture radar image feature extraction. According to the method, super-pixel-level frequency characteristics, super-pixel-level local contrast and super-pixel-level global contrast are included. Linear integration is carried out on the same features under different scales, saliency maps of the features are obtained, and it is guaranteed that the superpixel features obtained under the multiple scales can participate in fusion of the feature maps. And finally, multiplying the feature saliency maps to obtain a final saliency target detection result map. Finally, simulation experiments are carried out on SAR images in different scenes, and compared with various classic saliency detection algorithms and a two-parameter CFAR detection method, it is proved that the saliency target detection effect of the algorithm under most conditions can be better than that of a comparison algorithm, and meanwhile the background clutter suppression effect of the algorithm is higher than that of the comparison algorithm; under comprehensive evaluation, the SAR image salient target detection method has a good SAR image salient target detection effect.

The invention discloses a disinfection robot design method based on multi-feature fusion and attention weight obstacle avoidance. The method comprises the following steps: 1, designing hardware facilities of a disinfection robot; 2, designing a steering mode of the robot; 3, constructing a map model; 4, robot autonomous obstacle avoidance control; 5, the robot monitoring the body indexes of the patient in the working operation interval and sterilizing and killing the specific place of the hospital, and an ultraviolet disinfection lamp being installed above the robot. Therefore, the robot can have a certain sterilization and killing function. The newest Mecanum wheels are adopted on the basis of the disinfection robot, the problems that dead angles are prone to occurring in a narrow space, steering is inconvenient and the like are solved, medical workers can be replaced for basic work of some hospitals, meanwhile, contour features and color features are fused through feature saliency monitoring and attention weighting, the accuracy of the pedestrian detection model is improved.