Application Research of Adaboost Regression Prediction Based on Machine Vision for the Brightness of a Specified Distance Environment
Year: 2025
Type: article
Abstract: This paper proposes a decision guidance method based on an environmental brightness prediction model to address the problem of path planning failure and energy waste caused by environmental perception errors in robot recognition of dark areas in the environment. Firstly, by combining machine vision with light intensity sensors and depth cameras, a lightweight dataset containing light intensity, darkness parameters, dynamic parameters, and depth parameters is collected. Secondly, considering the poor accuracy of directly inferring dark areas based on traditional methods of obtaining environmental images and calculating image brightness information, this paper innovatively introduces dynamic parameters and depth parameters, which to some extent consider the impact of short-term environmental changes and spatial distribution on environmental brightness. Thirdly, the Adaboost regression model is used to train the self built lightweight data set. The analysis of feature importance shows that the dynamic and depth parameters account for 30% in total, which confirms the rationality and progressiveness of the introduction of dynamic and depth parameters. Finally, in order to more accurately evaluate the performance of different machine learning methods under the specific objectives of this study, Spearman correlation coefficient and Kendall rank correlation coefficient were introduced to evaluate the performance of the model. The experiment confirmed that the Adaboost model outperformed the decision tree, gradient boosting tree and other comparison models in Spearman (0.826) and Kendall (0.691) correlation coefficients. This method provides a high-precision and high security solution for predicting the brightness of the specified distance environment and identifying the lowest brightness point, with both theoretical value and engineering application potential.
DOI: 10.61784/jcsee3057
Authors: JinTao Hu, ZhuoFan Yang, ZeHuai Yuan, YingZi Chen, YiTong Hu
Language: English
Subfield: Aerospace Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4409682898