Analysis of Influencing Factors of Green View Index Based on Street View Segmentation

Abstract

The calculation of green view index in street view image has become an emerging choice for many researchers. This article proposes a PSAT-Net semantic segmentation model and specifically designs a pyramid pooling attention module. This module cleverly integrates channel and spatial information, implicitly implementing a three-dimensional attention mechanism, and significantly improving performance indicators. By performing detailed semantic segmentation on street view image, this article deeply analyzes the impact of different street view parameter configurations on green view index, and studies in detail how these parameters affect the visual perception of urban greening. The experiment found that: (1) The Green view index(GVI) of most sampling points decreases as the FOV increases. The increase in GVI is mainly affected by the crown width and pitch angle, and the high-value points of GVI are more easily affected. (2) Pitch angle usually causes a slight increase in GVI, which is mainly affected by the distribution of tall trees and low vegetation. Tall trees cause an increase in GVI, while low vegetation causes a decrease in GVI. Similarly, high value points of GVI are more likely to produce larger differences. (3) The Heading parameter has no obvious impact on the overall visual impact of GVI. The difference in independent sampling points comes from the angle between the lens acquisition direction and the road forward direction.These findings not only enrich the theoretical connotation of green view index research, but also provide valuable references for practical applications.