Image Dataset of Multiple Plant Species Under Different Lighting Modes of Plant Growth Lamps
Description
This dataset is developed for plant factory agriculture scenarios to support research related to the intelligent operation of such factories. As an efficient agricultural model, plant factories enhance crop production efficiency through vertical farming and precise environmental control, and can address global challenges like population growth and arable land scarcity. LEDs have become the mainstream light source in these factories due to their low power consumption and narrow spectral output; however, the red-blue light they consistently use causes significant color distortion in plant images. This reduces the accuracy of machine vision in tasks such as crop physiological monitoring, pest and disease detection, and other AI-driven work, hindering the intelligent upgrading of factories. Currently, most vision-based plant research relies on images captured under white light. Therefore, restoring color-distorted images from specific light sources (e.g., plant growth lamps) to a white-light equivalent state is a key requirement for the intelligent operation of plant factories—and this dataset is constructed to meet this need, serving the development and verification of plant image color correction algorithms.
Files
Steps to reproduce
Build a strictly controlled experimental environment: Use an indoor enclosed photography chamber (with an open top, light-shielded sides, and a dedicated camera aperture) to isolate external stray light and ensure stable lighting. Configure light sources: Use a white LED with fixed parameters as the reference light source for color correction; the experimental light source is a YIRUNFA brand LED plant growth lamp (containing 312 LED beads, specifically 204 red, 72 blue, 24 white, 4 infrared, 4 ultraviolet, and 4 orange beads). This lamp has three preset modes (Growth mode corresponding to blue color cast, Bloom mode corresponding to red color cast, and Growth+Bloom mode corresponding to purple color cast). Select and set up imaging equipment: Use a Sony ZV-E10L camera, operated in remote control mode to avoid shutter vibration. This ensures pixel-level alignment of images of the same sample under different lighting conditions, preparing for subsequent color analysis. Screen plant samples: Select 110 distinct plant materials, covering seedlings, vegetables, flowers, and foliar organs. These samples vary in color (green, red, yellow), morphology (broad leaves, needle-like leaves, succulent leaves), texture (smooth, downy, waxy), and physiological state (healthy, mild stress, different growth stages) to ensure the diversity and representativeness of the dataset. Perform image acquisition: Image each sample under four controlled lighting conditions respectively, namely "only baseline white LED turned on", "plant growth lamp in Growth mode", "plant growth lamp in Bloom mode", and "plant growth lamp in Growth+Bloom mode", and finally form a paired dataset.
Institutions
- Shanghai Jiao Tong University