Multi-label pixel classification method applied to large-scale urban reconstruction

Abstract

The invention provides a multi-label pixel classification method applied to large-scale urban reconstruction. The main content of the method comprises a data set, a network structure and a training method. The process comprises the steps of firstly generating a corresponding feature map by using the data set, then constructing a supervised deep learning frame based on a convolutional neural network, performing label classification and feature extraction on an input image, then inputting extracted features into a linear classifier support vector machine for a multi-classification task, and thusdifferentiating a subordination relation of a pixel to be classified. According to the invention, the classification of each pixel point except for boundary pixels in a high-definition image can be processed, a method for keeping the image size to be unchanged in the training process is provided, and the accuracy of object classification performed in a large-scale urban surface map is improved atthe same time.

Description

technical field [0001] The invention relates to the field of object classification, in particular to a multi-label pixel classification method applied in large-scale urban reconstruction. Background technique [0002] Object classification is an important research topic in computer vision, which has received extensive attention in recent years. Its development can include face recognition, pedestrian detection, intelligent video analysis, pedestrian tracking, etc. in the security field, traffic scene object recognition in the traffic field, vehicle Counting, retrograde detection, license plate detection, etc. are closely related to individual life. At the same time, it also has important applications in the classification tasks of large-scale large-scale images, such as ocean mapping and forest mapping in the field of surveying and mapping. Today, under the effect of economic development, the frequency of urbanization and urban integration construction is accelerating. There...

AI Technical Summary

Problems solved by technology

[0005] To solve the problem of pixel-level classification in large-scale images, the purpose of the present invention is to provide a multi-label pixel classification method applied in large-scale urban reconstruction, first use the data set to generate the corresponding feature map, and then construct a Based on the supervised deep learning framework of convolutional neural network, label classification and feature extraction are performed on the input image, and then the extracted features are input into the linear classifier support vector machine for multi-classification tasks, so as to distinguish the Affiliation of Classified Pixels

Images