31 results about "Scale-space segmentation" patented technology

An apparatus includes an object detector configured to receive image data of a scene viewed from the apparatus and including an object. The image data is associated with multiple scale space representations of the scene. The object detector is configured to detect the object responsive to location data and a first scale space representation of the multiple scale space representations.

A scale-Space feature extraction technique is based on recursive decomposition of polyhedral surfaces into surface patches. The experimental results show that this technique can be used to perform matching based on local model structure. Scale-space techniques can be parameterized to generate decompositions that correspond to manufacturing, assembly or surface features relevant to mechanical design. One application of these techniques is to support matching and content-based retrieval of solid models. Scale-space technique can extract features that are invariant with respect to the global structure of the model as well as small perturbations that 3D laser scanning may introduce. A new distance function defined on triangles instead of points is introduced. This technique offers a new way to control the feature decomposition process, which results in extraction of features that are more meaningful from an engineering viewpoint. The technique is computationally practical for use in indexing large models.

The invention brings forward a Chinese word segmentation method by using character embedding based on word context and a neural network. The character embedding is learnt on large-scale automatic segmentation of data. The learnt character embedding is used as inputting of a segmentation model of the neural network so as to effectively help model learning. The method comprises the concrete steps of learning character embedding on large-scale automatic segmentation of data according to the word context and lexeme marking; and utilizing character-embedding as inputting of the segmentation model of the neural network so that segmentation performance is improved. Compared with other Chinese word segmentation method technologies based on the neural network, the method adopts the character-embedding technology based on the word context so that word information is effectively integrated into the segmentation model. Therefore, the accuracy of a word segmentation task is improved.

Advanced driver assistance systems need to be able to recognize and to classify traffic signs under real time constraints, and under a wide variety of visual conditions. The invention shown employs binary masks extracted by color space segmentation, with a different binary mask generated for each sign shape. Temporal tracking is employed to add robustness to the detection system. The system is generic, and is trainable to the signs used in various countries.

The invention relates to the technical field of remote sensing image processing, in particular to a remote sensing image color enhancing method based on multi-scale image segmentation and color transferring. The method comprises steps: S1, conducting wavelet denoising preprocessing for a target gray level image; S2, according to a multi-scale normalized segmentation method, conducting segmentation of the target gray level image after preprocessed; S3, combining a reference image, and conducting brightness matching for all sub-domains of the target gray level image; and S4, selecting a sub-domain with the similar property to conduct mapping and color transferring. Aesthetic measure and target recognized degree of a result image are obviously enhanced, under the condition that image registering is not needed, the method can achieve full-automatic color enhancing for a remote sensing image which is relatively simple in texture structure and big in brightness contrast just through a color reference image which is similar in character and style, and therefore, the remote sensing image color enhancing method provides strong technical support for remote sensing image processing.

The invention discloses an ionospheric tomography technology based on multi-scale subdivision and an ionospheric delay correction method. The three-dimensional space of the regional ionosphere is subdivided according to different "pixel" scales, thereby obtaining multiple different The single-scale ionospheric tomography model, the unknown variables of these models are uniformly solved, and according to different weight factors, finally weighted to obtain the solution of the multi-scale tomography model, the ionospheric electron density distribution in the region is reconstructed, and the regional ionospheric electron density distribution is obtained. Ionospheric delay. The ionospheric space activity law reconstructed by the invention has a high degree of fitting, strong timeliness, and is convenient to use; the regional ionospheric delay amount calculation result obtained according to the invention has high accuracy, which expands the application range of CORS measurement results. After analyzing the application results of a large number of engineering examples, the ionospheric electron density distribution reconstructed by the invention is smoother and more reasonable than the traditional single-scale ionospheric tomography model, and the accuracy of ionospheric delay correction is increased by 30% on average.

The invention discloses a registration method for visible light and infrared images based on multi-scale segmentation and SIFT, which comprises the following steps: step 1, preprocessing is performed to visible light images and infrared images, wherein Gaussian filtering processing is performed to the visible light images, and linear enhancement processing is performed to the infrared images; step 2, an initial parameter value which includes a scale factor, a compact factor and a shape factor is set, and the multi-scale image segmentation is performed to the visual light and infrared images subjected to preprocessing; step 3, the SIFT registration is performed to the visible light and infrared images subjected to the multi-scale image segmentation, calculation parameters of homonymy point selection transformation models are found out and the registration is performed to source images; step 4 and step 5, judgment is performed, if the visible light images and the infrared images are not subjected to accurate registration or the registration accuracy does not meet the requirement of the set threshold, the operation is returned to step 2 to adjust the parameter value of the multi-scale image segmentation method till the registration accuracy of the visual light images and infrared images meets the requirement of the set threshold, iteration is stopped, and the registration result is output.

The invention provides a high-performance implementation method for multi-scale segmentation of remote sensing images, and in particular a method for implementing fast and multi-scale image segmentation of large amount of remote sensing images and establishing the structural relationship of segmentation results in the process of high-resolution remote sensing image information extraction. The method comprises the following steps: on the basis of analyzing the implementation procedure of the algorithm and finding out the computing-intensive segment of the algorithm, realizing the parallel segmentation of the algorithm-intensive segment based on MPI and OMP models, and carrying out date join on the parallel segmentation results; implementing the multi-scale image segmentation by storing theinitial segmentation results of the algorithm and merging the subsequent multi-scale remote sensing images; and establishing a multi-scale object topological relation model. The generated multi-scalesegmentation region and corresponding hierarchical relationship can be applied to various services; and the corresponding implementation method is applicable to various segmentation algorithms such as mean shift and the like, and can greatly improve the data amount processed by algorithm and the processing efficiency.