7 results about "Mixed resolution" patented technology

The present application relates to a kind of railway tunnel disease intelligent detection method, including mixed resolution image acquisition, mixed resolution normalization preprocessing, double channel parallel disease detection, physical quantity calibration based on dynamic precision coefficient, dynamic splicing algorithm based on ROI feature fusion and mileage error dynamic calibration based on semantic anchor point.The present application is characterized in that, by mixed resolution normalization preprocessing, the compatibility problem of heterogeneous hardware is solved, the disease feature extraction precision of key area is retained using double-scale cascade detection mechanism, with dynamic precision coefficient matrix, the pixel physical mapping deviation caused by heterogeneous resolution is corrected, the accuracy of disease quantitative analysis and physical quantity inverse calculation is ensured;The present application significantly improves the processing efficiency of ultra-high resolution image, eliminates the mileage cumulative positioning error in long-distance inspection, realizes the full-quantitative accurate detection of track traffic tunnel disease in low-cost, high-efficiency environment.

A source video is encoded in a streaming protocol comprising a normal stream and a companion stream. The normal stream contains I-frames and predicted frames at high resolution. The companion stream contains frames at a lower resolution. When needed, an I-frame from the companion stream may be decoded and upscaled from the second resolution to the first resolution and injected into a decoded picture buffer. At an interval corresponding to an amount of time that is shorter than an interval between I-frames in the normal stream, a first frame is downscaled to the second resolution and then upscaled again to the first resolution. The next frame is then encoded with reference to the upscaled frame such that an output stream recovers to the first resolution based on the second frame prior to receipt of the next I-frame in the normal stream.

This invention proposes a conformal array direction finding method and system based on hybrid resolution quantization, relating to the field of direction-of-arrival (DOA) estimation in passive radar. It aims to reduce the high resource consumption of conformal arrays in DOA estimation, thereby solving the deployment challenges of direction finding under space-constrained conditions such as UAV positioning or vehicle-mounted radar. The method involves designing the conformal array structure; constructing models of the received signal from the receiving antenna and the transmitted signal from the transmitting antenna to obtain received data; using the designed structure to receive high-precision and low-precision quantized data, and processing the low-precision quantized data according to an additive Gaussian noise model to obtain received data under hybrid resolution quantization; performing covariance decomposition on the received data under hybrid resolution quantization; and performing spectral peak search on the covariance-decomposed hybrid resolution quantized data to estimate DOA and polarization information. This invention is applicable to direction finding methods for conformal arrays.

The present application provides a video processing method, device and storage medium. The method encodes and decodes the original video to obtain a mixed-resolution video, which includes a first-resolution frame and a second-resolution frame corresponding to the key frame, and also includes a third-resolution frame corresponding to the non-key frame. The resolution of the first-resolution frame is higher than the resolution of the second-resolution frame and the resolution of the third-resolution frame; based on the first-resolution frame and the second-resolution frame, the third-resolution frame corresponding to the non-key frame is amplified and the amplified video is output. The amplified video includes the first-resolution frame corresponding to the key frame and the amplified target frame corresponding to the non-key frame. The above scheme only needs to encode and decode the non-key frame to obtain a low-resolution frame, and then amplify the low-resolution frame according to the information of the key frame to obtain an amplified high-resolution video with less bit rate, thereby reducing the bandwidth pressure of video transmission.

The invention provides a GD-NeRF super-resolution three-dimensional reconstruction method based on a hybrid generative model, and solves the problem of three-dimensional reconstruction under low-resolution input by fusing the advantages of a GAN and a diffusion model. According to the method, various details are generated by using a GAN (such as ESRGAN), and a solution space is optimized by combining a consistency control potential code (CCLC); meanwhile, the iterative refinement capability of a diffusion model (such as Stable Diffusion) is introduced, and the multi-view consistency is ensured through a heavy noise fractional distillation (RSD) module and an iterative 3D synchronization (I3DS) module. A multi-stage training strategy is adopted, details are rapidly generated through GAN in the initial stage, global consistency is optimized through a diffusion model in the later stage, and balance supervision signals are adjusted through dynamic weights. The framework supports mixed resolution input, utilizes an HR view to guide LR generation, and combines physical constraint (CEM) and RSD target constraint to ensure that a generation result not only meets an LR condition, but also retains detail diversity. According to the method, a high-quality and coherent high-resolution result can be generated from the multi-angle low-resolution image.

A source video is encoded in a streaming protocol comprising a normal stream and a companion stream. The normal stream contains I-frames and predicted frames at high resolution. The companion stream contains frames at a lower resolution. When needed, an I-frame from the companion stream may be decoded and upscaled from the second resolution to the first resolution and injected into a decoded picture buffer. At an interval corresponding to an amount of time that is shorter than an interval between I-frames in the normal stream, a first frame is downscaled to the second resolution and then upscaled again to the first resolution. The next frame is then encoded with reference to the upscaled frame such that an output stream recovers to the first resolution based on the second frame prior to receipt of the next I-frame in the normal stream.

The application belongs to the technical field of target detection, and discloses a three-dimensional target detection method for real-time sensing in a mobile environment, which comprises the following steps: extracting image two-dimensional features by inputting a multi-view image of a mobile device into an image backbone network; inputting the image two-dimensional features into a depth prediction network and a semantic prediction network to obtain a depth prediction value and a semantic prediction value, performing multi-level depth supervision on the depth prediction value, and performing semantic supervision on the semantic prediction value; projecting the image two-dimensional features into a three-dimensional space to obtain BEV features; performing bilinear interpolation sampling on the BEV features by using a mixed resolution grid to form new BEV features; inputting the new BEV features into a BEV feature encoding network, and then obtaining a three-dimensional target detection result by using a 3D detection head; and performing detection supervision on the three-dimensional target detection result, and updating network parameters in combination with all the supervision. The application fully utilizes depth information to improve the detection effect, and adapts to actual detection requirements by using a mixed resolution grid.