High-Definition Garden Plant Images Threshold Segmentation Mechanism Based on PSO and DRL

Introduction

Garden plants, as a renewable resource, affect all aspects of human development and constrain the progress of society. With the development of society, people are gradually highlighting garden plants’ value, and they have unlimited development prospects and are thus valued by humans (Yuan et al., 2023). The diversity of garden plants plays a huge role in the ecological environment. In recent years, due to the increase in industrial production, the competition between humans and nature has become increasingly fierce. It has affected the balance of the ecological environment and led to some precious species being on the brink of extinction. The disappearance of these garden plants will in turn disrupt the balance of the ecosystem, affect human life and scientific progress, and exacerbate a series of environmental problems such as the greenhouse effect, land desertification, and soil erosion (Valiente et al., 2015). Therefore, the protection of garden plants is an important task today.

Garden plants play a crucial role in ecological environment protection. In the growth process of garden plants, in order to avoid some adverse factors affecting them, we need to monitor various phenotypic parameters of garden plants in real time, such as leaf color, leaf area, and leaf droop angle (H. Zhang, Wang, et al., 2023). Due to the fact that in the traditional field of garden plants, the extraction of phenotypic parameters of garden plants is usually done through manual measurement. This causes certain harm to garden plants during the use of measuring devices. Therefore, in recent years, methods that use image processing technology to safely extract phenotypic parameters of garden plants have been widely applied, avoiding some drawbacks of traditional measurement methods (Miao et al., 2023). In the field of modern garden plant science, high-definition garden plant image segmentation technology not only is applied to monitor the growth of garden plants and identify weeds around garden plants, but is also often used to distinguish the types of garden plants and count the number of garden plant leaves.

Garden plant image segmentation is a prerequisite for further operation of garden plant images and has always been a highly anticipated and challenging research direction in the area of high-definition image processing. With the popularization of the Internet and the rapid development of image and video equipment, data for existing high-definition garden plant images has increased (Besson et al., 2022). The original garden plant image processing technology has been stretched to the limit, and the upgrading of garden plant image processing technology is imminent. How to quickly process a large number of high-definition garden plant images has become a problem worth studying (Y. Yu, Wang, et al., 2023).

The threshold segmentation method is one of the traditional segmentation methods used for garden plant images, and it is the most basic and widely used segmentation technique in high-definition garden plant image segmentation (Dronova et al., 2012). For example, in the application of infrared technology, it is applied in the segmentation of infrared thermal non-destructive image testing (Bu et al., 2019). In biomedical engineering, it is applied in the segmentation of magnetic resonance images (Hao et al., 2020). In agricultural engineering applications, it is applied in the segmentation of non-destructive image testing for fruit quality (Van de Looverbosch et al., 2021). As a highly practical segmentation method, the real-time performance of threshold segmentation directly affects the operational efficiency of the entire system. Particularly when there are many thresholds, the real-time and security requirements for selecting the optimal threshold become important indicators to measure the superiority of threshold segmentation methods (Z. Wang et al., 2024).