Using computer vision for analysis of plant growth condition: what to consider? Hans Jørgen Andersen Computer Vision and Media Technology laboratory Aalborg University
Background ► Lower prices of cameras opens new possibilities for sensor development ► Cameras – Computer Vision – used in industry normally takes place in a controlled environment ► Within agriculture this is often not feasible
Problem ► If the surrounding environment for image acquisition cannot be controlled Then the computer vision system has to adjust to the environment
Outdoor Images of Wheat Plants Sunshine, unclouded. Sunshine, clouded. Skylight, clouded.Skylight.
If Spectra of Light Source Changes Spectra of Reflected Light Changes Spectral Variation of the Illumination
Light Sources ► Outdoor Condition poses the problem of Two Illumination sources The Sun and The Sky Sunlit Sky light How can you analysis the green color of the vegetation ?
Characteristic of Reflections ► Sunlit condition may pose two reflection components: Highlight, i.e. the color of the sun / light source Body, i.e. the color of the plant / object Plastic cup With HighlightPure Body reflection
Outdoor Image Formation Light Sources Object Plant Transmittance Absorption Reflectance Observer Camera Ambient Point Uniform
Modeling of Daylight Black Body T, Kelvin Black Body spectra Daylight may be modelled as a Black Body Correlated Colour Temperature, CCT Daylight model spectra
Segmentation
Classifying Reflections ► Reflections from Coffee classified into Body and Highlight Components Probability of Body Reflection
Use within Gap Fraction Estimation Original Image Classifying each pixel as Soil (”gap”) = 0, Plant = 1
Multi - Spectral Images ► nm, 26 bands
Modeling Spectra ► Endmember Spectra Known (measured)
Classifying Reflections ► First-order body scattering Vegetation Soil
Classifying Reflections (2) Specular reflectionVegetation - Vegetation Vegetation - SoilSoil - Soil
Conclusion ► Modeling the Image Formation Process: Is valuable for Robust segmentation Analysis of vegetation growth status Assessment of various reflection components
Perspectives ► Modeling of vegetation