detection of pine nut rot

2024年10月至11月，研究团队在中国黑龙江省哈尔滨市东北农业大学农业农村部东北智慧农业技术重点实验室试验基地采集数据。数据集构建流程图如图1所示。在数据采集过程中，先准备八个8*15格的育苗盘（280mm×540mm），每格底部垫有湿巾并放入20至30粒大兴安岭偃松，于每日中午12点、晚上21点浇水 ，使湿度保持在80%-95%之间，促使松子发生霉变。同时使用海康威视工业相机与笔记本电脑检测松子霉变状态。摄像头安装至距育苗盘0.55米处，使用延时摄影生成RGB图像序列。基于以上准备，对松子霉变进行为期28天的胁迫实验。同时考虑到松子霉变过程中霉斑形态特征易受光影影响，采集数据集时采用顺光、逆光、变换距离等不同角度拍摄，以增强模型在复杂环境下的检测性能。

From October to November 2024, the research team collected data at the experimental base of the Key Laboratory of Northeast Smart Agricultural Technology under the Ministry of Agriculture and Rural Affairs, Northeast Agricultural University, Harbin, Heilongjiang Province, China. The flow chart of dataset construction is shown in Figure 1. First, eight 8×15-cell seedling trays (280 mm × 540 mm) were prepared, with a wet pad placed at the bottom of each cell and 20 to 30 seeds of *Pinus pumila* from the Greater Khingan Mountains added to each cell. Watering was performed daily at 12:00 noon and 21:00 to maintain the humidity within 80%–95%, thus inducing mold growth on the pine seeds. Meanwhile, a Hikvision industrial camera and a laptop were deployed to monitor the mold status of the pine seeds. The camera was mounted 0.55 meters above the seedling trays, and RGB image sequences were generated using time-lapse photography. Based on the above preparations, a 28-day stress experiment was conducted to induce mold development on the pine seeds. Given that the morphological characteristics of mold spots during the pine seed mold process are readily affected by light and shadow, diverse shooting configurations including front light, backlight, and variable shooting distances were adopted during data collection to enhance the model’s detection performance under complex environmental conditions.

搜集汇总

背景与挑战

背景概述

这是一个用于检测松子腐烂的公开数据集，包含6千张RGB图像，专注于计算机视觉领域的目标检测和图像识别任务。