7 results about "Optical branch" patented technology

An optical input / output chip and distributed computing system, including: output sub-chip includes first light processing module to form N first light branches output from laser source, divide and modulate each first light branch into M first light sub-branches according to M wavelengths, form second optical branches containing N different modes from N modulated first optical branches before being output in first polarization state; input sub-chip includes second optical processing module, to form N third optical branches containing same mode from N second optical branches containing different modes, divide out part of light in each third optical branch having been converted from first polarization state to second, and form first polarization state again, combine and demodulate light in the first polarization state without conversion in original third optical branch after having been divided into M second optical sub-branches according to M wavelengths respectively. Thus increasing both communication rate and bandwidth effectively.

The application discloses a kind of SAR data active integration's remote sensing data timing seamless reconstruction method and device, belong to remote sensing image processing and space-time information analysis technical field.Sentinel-2 waveband and Sentinel-1 dual polarization data are screened and time-space alignment, radiation normalization and cloud mask construction are completed;Parallel operation dual-branch deep network optical branch extracts multi-scale space-time characteristics, SAR branch uses deformable convolution+texture self-attention to mine multi-scale directional texture;Then optical characteristics are used as Query, in cloud mask marked area active search SAR Key-Value complementary information, realize deep fusion by cross-modal multi-head attention;Fusion result is jointly corrected by relative time distance coding, time attenuation factor and cloud perception mask, and long-time sequence attention drift is inhibited;Final level fusion multi-level feature and adaptive weighting are outputted, and full-time seamless image is outputted.

This invention relates to a packaging box for a dual-channel PLC optical splitter for a space station and its manufacturing method. The packaging box for the dual-channel PLC optical splitter for a space station includes a box body, two input optical cable protective sleeves installed inside the box body, two sets of PLC cores that serve as backups for each other, a multi-output optical cable protective sleeve, and an array slot output positioning block. The two input optical cable protective sleeves and the multi-output optical cable protective sleeve are respectively installed at both ends of the box body. One axial end of each set of PLC cores has a single fiber of multi-output protection tube. An array slot output positioning block with an array of positioning slots is installed inside one end of the box body. One axial end of the multi-output optical cable protective sleeve passes through and is fixed in multiple positioning slots. The single fiber of the multi-output protection tube passes through the multi-output optical cable protective sleeve to form a multi-output optical cable. The other axial end of each set of PLC cores has a single fiber of two input protection tubes. The single fiber of the two input protection tubes passes through the two input optical cable protective sleeves to form two input optical cables.

This invention discloses a method and apparatus for seamless temporal reconstruction of remote sensing data by actively integrating SAR data, belonging to the field of remote sensing image processing and spatiotemporal information analysis technology. The method involves screening Sentinel-2 and Sentinel-1 dual-polarization data and performing spatiotemporal alignment, radiometric normalization, and cloud mask construction. A dual-branch deep network is run in parallel; the optical branch extracts multi-scale spatiotemporal features, while the SAR branch mines multi-scale directional textures using deformable convolution and texture self-attention. Subsequently, the optical features are used as queries to actively retrieve complementary key-value information from the SAR data within the cloud mask-marked region, achieving deep fusion through cross-modal multi-head attention. The fusion result is then jointly corrected by relative temporal distance encoding, temporal decay factor, and cloud-sensing mask to suppress long-term attention drift. Finally, multi-level features are hierarchically fused and adaptively weighted to output a seamless, full-temporal image.

The application provides a remote sensing image cloud removing method and system based on optical remote sensing images and synthetic aperture radar images, and belongs to the technical field of remote sensing image processing. The method first acquires paired optical remote sensing images and synthetic aperture radar images; then a double-branch feature extraction network of an optical branch and a SAR branch is constructed to extract spectral and context features of the optical images and structural and texture features of the SAR images respectively; on this basis, the optical features and the SAR features are dynamically weighted and fused through a hierarchical cross-modal adaptive fusion module to obtain a fused feature map; and finally, the fused feature map is reconstructed to output a cloud-free remote sensing image. The application can realize high-quality remote sensing image cloud removing reconstruction under complex cloud conditions.

The application discloses a landslide detection method based on adaptive fusion of optical satellite remote sensing and DEM data, and takes optical satellite remote sensing data and DEM data as core inputs to construct a heterogeneous double-branch high-resolution network: the optical branch adopts HRNet-W48 to extract spectral texture features, and the DEM branch adopts a lightweight HRNet-W18 to extract terrain geometric features. A terrain perception gating fusion module is designed, a pixel-level weight mask is dynamically generated based on the terrain features reflected by the DEM data through the gating network, and the adaptive weighted fusion of optical and terrain features is realized in the form of residual error; a CE+BCL combined loss function is proposed, and the boundary contrast loss is used to strengthen the supervision of the landslide edge area. The application effectively solves the problems of rigid fusion strategy, loss of spatial details and fuzzy boundary in the traditional method, and significantly improves the accuracy, boundary clarity and complex scene adaptability of landslide recognition.