A low-cost ground-based NDVI imaging solution (ViperPics) for vegetative remote sensing

Human beings have been directing technology markets according to their needs. With the help of computers, algorithms, sensors, and various internet of things (IoT), systems are now being supported by artificial intelligence. New generations have been witnessing the rapid integration of humanoid robots into advanced technology. As a result, studies and applications can be conducted using low-cost, compact single-board computers and sensors. The potential use of such systems in natural resource management and forestry applications is expected to increase steadily. In this study, multispectral image acquisition was carried out using PiCameras in a low-cost imaging system which we refer to as ViperPics. A first-degree polynomial (affine) transformation was applied for image alignment and geometric corrections. A vegetation difference index was calculated based on the ratio of near-infrared (NIR) and red bands, and was tested on an ornamental plant. ViperPics imagery was processed both with and without calibration. Plant index estimations were performed using commercially available low-cost reference cards (black, white, and grey). Calculations were conducted using both the recommended common reflectance values for reference cards, and also measured-reflectance values obtained using a spectrometer. Resulting the Normalized Difference Vegetation Index (NDVI) values ranged between −1 and +1. Radiometric correction resulted in more consistent and reliable NDVI estimates. Relative NDVI showed poor pixel-level agreement in the uncalibrated case (Root Mean Squared Error (RMSE) 0.48–0.51, Pearson R −0.41 – −0.38) despite moderate spectral cross-talk estimation. Calibrated images showed a reduced tendency towards negative NDVI values and a trend towards normalization in areas outside the plant. Our results indicate that ViperPics provides a low-cost and accessible system for imaging with NIR and other colour bandwidths. Further efforts will involve greater integration of the image acquisition software and image processing algorithms into the system.