9 results about "Shape context" patented technology

The invention discloses an accurate attachment process optimization system based on computer vision, and the system comprises an image collection module which is used for obtaining a product and label image; the contour construction module is used for extracting edges and generating a contour point set; the pose recognition module is used for determining the spatial positions and angles of the product and the label based on the shape context and the Hungary algorithm; the coordinate mapping and compensation module is used for mapping the spatial pose to a tool coordinate system and generating a compensation amount; the track generation and motion control module is used for generating a smooth attaching track and driving an attaching mechanism to execute operation; the quality detection and distribution module is used for collecting the post-pasting image, judging the qualification and controlling the distribution of defective products; and the beat optimization module is used for dynamically adjusting the conveying speed and the camera triggering opportunity. According to the method, stable product and label pose recognition, rapid point pair matching and dynamic beat optimization based on the queuing theory can be realized under complex working conditions, so that the attachment precision and the production efficiency are both considered.

This invention discloses a method and system for point cloud completion of complex plant structures based on an end-to-end network. The method constructs a hierarchical geometric encoder to extract and fuse multi-scale geometric features and global shape context features from an incomplete input tree point cloud. A seed generator then generates a coarse point cloud containing the main skeleton of the tree. A cascaded tree growth module progressively upsamples the coarse point cloud, and dynamic deformation constraints ensure that the upsampling process conforms to the natural growth pattern of trees, ultimately outputting a structurally complete and realistically shaped tree point cloud. Furthermore, this invention addresses the scarcity of real, complete tree point cloud data by fusing UAV and ground-based LiDAR scanning data, providing high-quality supervisory data for model training. This effectively solves the problems of insufficient semantic understanding, loss of detail, and morphological distortion in existing point cloud completion methods when dealing with complex tree branch structures, significantly improving completion accuracy and morphological fidelity.

The invention belongs to the technical field of computer vision, and particularly relates to an end-to-end shape recognition method and system based on a complex network. The method comprises the following steps: carrying out dense discretization on the edge of a to-be-identified image to obtain an edge description point set; selecting a series of key points on the obtained edge description point set at equal intervals; extracting local features of a key point surrounding edge description point set by using a shape context; taking each key point as a node, defining edges in the network according to spatial correlation among the key points, and constructing a multi-layer complex network by adopting a dynamic evolution strategy; local features and topological features of the complex network are processed through a graph convolutional network, and robust expression of the shape classification task is obtained; and performing classification through a full connection layer network to obtain a classification result, and realizing end-to-end shape recognition. According to the method, the adaptability and robustness of the method to different shape classification tasks are enhanced.

The invention provides a scene coding and code matching method and device for an underground parking lot, and electronic equipment, and the method comprises the steps: carrying out the rough matching through a random walk descriptor, determining a candidate mapping topological interval matched with a positioning topological interval, carrying out the fine matching through a shape context descriptor, and obtaining a candidate mapping topological interval matched with a positioning topological interval; and finally, obtaining a target matching result of matching the first target shape context descriptor of the positioning topological interval with the second target shape context descriptor of the candidate mapping topological interval, so that the matching workload of subsequent fine matching is greatly reduced in the rough matching process, and the matching efficiency of overall matching is improved. And the accuracy of the fine matching process is good, the matching between the points is centimeter-level, and the cross validation effect is also formed by fusing the two matching modes, so that the scene recognition result is more accurate and robust.

The present invention relates to the field of scene recognition technology, and is a real-time semantic segmentation method and system for multi-shape pyramids in traffic scenarios. The method comprises the following steps: initializing an original image to be segmented to obtain an initial segmentation feature map; performing channel splicing on the initial segmentation feature map and an output feature map obtained by first downsampling the original image to be segmented to obtain a fused feature map, then downsampling and extracting detail information to obtain a detail information feature map; downsampling the detail information feature map and extracting semantic information; extracting horizontal and vertical receptive field information from the semantic information feature map, performing dilated convolution to obtain feature maps with different receptive fields, obtaining multi-scale and multi-shape contextual information feature maps, then upsampling them, performing point-by-point convolution on the detail information feature map to reduce dimensionality, performing channel splicing on the two to obtain a fused feature map containing semantic and detail information, adjusting the number of channels to the number of predicted categories, and normalizing to obtain a predicted segmentation result. The method has good segmentation accuracy and inference speed.

The invention discloses an RGB-T dual-light camera system parameter self-calibration method based on shape context, and relates to the technical field of camera parameter self-calibration, and the method comprises the steps: synchronously collecting a plurality of frames of RGB images and thermal imaging images of an object at different visual angles, respectively extracting the outlines of the images through employing a contour extraction algorithm, carrying out the sampling of the outlines, and carrying out the self-calibration of the parameters of the RGB-T dual-light camera system. Obtaining a contour point set and calculating a shape context histogram; comparing the shape context histograms to obtain a matching point set, and when the number of matching points in the set exceeds a set threshold value, obtaining a matching point essential matrix based on an RANSAC algorithm; and decomposing the essential matrix through SVD to obtain decomposed external parameters, optimizing the decomposed external parameters through a nonlinear optimization technology, and completing RGB-T dual-light camera system parameter self-calibration by using the optimized external parameters. According to the method, the calculation of the relative pose external parameters between the cameras can be realized without special processing, and the calibration of the RGB-T dual-light camera system is facilitated.

This invention provides a method for road marking extraction and classification based on a local shape Transformer network. The method includes: acquiring an original LiDAR point cloud and performing non-ground point filtering preprocessing to obtain a road surface point cloud; inputting the roadside point cloud into a Transformer-based encoder network, and obtaining semantic features containing local geometric details and global shape context through multi-stage downsampling and feature extraction; using a decoder network to upsample and recover the semantic features, fusing multi-scale information with skip connections, and outputting the road marking category prediction result for each point through a classification head. This invention improves the extraction accuracy and boundary integrity of fine markings such as lane lines and arrows.

The present invention relates to a comprehensive evaluation method for the similarity of a bionic vehicle from multiple perspectives. First, images are collected from different perspectives of the bionic vehicle and real organisms, their contours are extracted, and as uniform contour sampling points as possible are obtained through the contour point sampling method. Secondly, the shape context algorithm is used to calculate the similarity evaluation matrix. Finally, the similarity between each view is calculated through the shape context distance. The similarities of each view are weighted and summed to obtain the overall shape similarity of the bionic vehicle. This method calculates the similarity value of the bionic vehicle based on the shape context algorithm, indicating the direction for the optimization of its shape and structure.

The present invention relates to the technical field of image processing, and specifically discloses an e-commerce image dynamic optimization and adjustment method based on market feedback. Color noise is removed and interference in edge detection is suppressed through gray conversion and smoothing filtering techniques. Canny edge detection and wavelet transform are used to extract and analyze edge curvature features, accurately identify serrated regions. Shape context descriptors and the Hungarian algorithm are used to align the boundaries of pictures of the same commodity from different perspectives, and the shape difference coefficient is calculated to evaluate its consistency. For further optimization, the shape difference coefficient and edge curvature features are transformed into a comprehensive feature vector, which is used as the input of the graph convolutional network model. Global features are extracted through multi-layer graph convolutional operations to identify regions with severe serration, and hard thresholding is used for denoising, and cubic spline interpolation is used to accurately repair the serrated regions.