Intelligent analysis and evaluation method for city hue

Abstract

The invention belongs to the technical field of urban color planning, and particularly relates to an intelligent analysis and evaluation method for city hue. The method comprises the following steps:performing semantic segmentation on streetscape image data by adopting an artificial intelligence algorithm, and extracting buildings in the streetscape image data; adopting a character intelligent detection algorithm and an image processing technology to extract shop signs in the streetscape; based on an HSV color system, extracting and counting building and shop call colors; analyzing and judging whether the shop call and the building tone are coordinated with the dominant tone of the appointed area, and marking out a non-coordinated area; and evaluating and giving a rectification scheme ofthe inharmonious area by utilizing an image aesthetics evaluation algorithm. According to the method, image segmentation, shop call extraction and aesthetic evaluation are carried out by using a deeplearning algorithm, and the precision of the algorithm can be gradually improved along with the increase of the data volume. The method is suitable for urban image big data, high in robustness and good in real-time performance, and has popularization and application prospects.

Description

[0001] field of invention [0002] The invention belongs to the technical field of urban color planning, and in particular relates to an intelligent analysis and evaluation method of urban color. Background technique [0003] Intelligent analysis and evaluation of urban color is an important part of urban design, which plays a prominent role in continuing urban context, improving urban quality, and shaping urban spirit. However, the development of urban color planning is facing many difficulties, the biggest problem is the investigation and evaluation of the current status of color. This type of work requires systematic research on the status quo of urban color, but there are problems such as ignorance of the color system, immature technical methods, and a large area of ​​current research. Using manual research methods is time-consuming, laborious and incomplete, and it is difficult to meet the requirements of color planning. . [0004] Artificial Intelligence (AI), abbrevia...

AI Technical Summary

Problems solved by technology

[0006] At the same time, traditional machine learning algorithms are also constantly developing, and gradually surpassed neural networks in the late 1990s, becoming a hot spot in academic research, focusing on classifiers such as SVM, AdaBoosting, and random forests. There are two main reasons. First of all, the relevant hardware level cannot meet the huge amount of calculation of the deep neural network, which resulted in the insufficient depth of the neural network at that time, and could not give full play to its advantages.

Secondly, the amount of data at that time was relatively small, which could not meet the needs of deep networks.

Images