Multimodal image dataset of tomato fruits with different maturity

Accurately identifying tomatoes of different maturity and determining the appropriate harvest time are important prerequisites for achieving efficient tomato harvesting and ensuring post-harvest quality. However, in actual harvesting scenarios, complex lighting conditions can degrade the quality of RGB images, making it difficult for models to acquire and utilize key visual features, leading to application bottlenecks. Multimodal data has complementary and consistent properties, which can provide additional feature information for models. Depth and near-infrared images are mainstream multimodal research data due to their ease of acquisition and low cost. Additionally, tomato fruits have dense distribution and asynchronous maturity characteristics during the growth process. Existing tomato cluster detection and segmentation datasets are difficult to meet the needs of maturity-oriented harvesting. Building a multimodal image dataset of tomato fruits of different maturity based on visible light, depth, and near-infrared can effectively fill the gaps in the current research field. This dataset contains 4,000 sets of multimodal image data samples, covering tomato sample images under four lighting conditions: natural light, artificial light, weak light, and sodium yellow light. It includes target detection and semantic segmentation annotations for three maturity stages: unripe, half-ripe, and ripe. The total size is 10.4 GB. This dataset can provide basic data support for the development of visual intelligence systems for tomato intelligent management and harvesting equipment.

精准识别不同成熟度的番茄并确定适宜的采收时机，是实现番茄高效采收、保障采后品质的重要前提。然而在实际采收场景中，复杂的光照条件会降低RGB（Red-Green-Blue）图像的质量，使得模型难以获取并利用关键视觉特征，进而引发应用瓶颈。多模态数据具备互补与一致的特性，可为模型提供额外的特征信息。深度图像与近红外（near-infrared）图像因易于获取、成本低廉，成为主流的多模态研究数据。此外，番茄果实生长过程中存在分布密集、成熟度异步的特性，现有的番茄簇检测与分割数据集难以满足面向成熟度的采收需求。构建基于可见光、深度与近红外模态的不同成熟度番茄果实多模态图像数据集，可有效填补当前该研究领域的空白。本数据集共包含4000组多模态图像数据样本，涵盖自然光、人工光、弱光及钠黄光四种光照条件下的番茄样本图像，包含针对未成熟、半成熟、成熟三个成熟度阶段的目标检测与语义分割标注信息。数据集总容量为10.4 GB，可为番茄智能管理与采收设备的视觉智能系统研发提供基础数据支撑。

搜集汇总

背景与挑战

背景概述

该数据集是一个针对番茄果实不同成熟度的多模态图像数据集，包含4,000组基于可见光、深度和近红外图像的数据样本，覆盖四种光照条件并提供三个成熟阶段的目标检测和语义分割标注，总大小为10.4 GB。其目的是通过多模态数据弥补复杂光照下RGB图像质量下降的不足，为番茄智能收获和管理设备的视觉系统开发提供基础数据支持。