Acquiring Instantaneous Multispectral Imagery Using a Single Image Sensor with Multiple Filter Mosaic Fredrik Bergholm 1 , Hamed Hamid Muhammed 2 and Anders Larsolle 3 1 Dep. of Numerical Analysis and Computer Science, NADA; Royal Institute of Technology, KTH, SE-100 44 Stockholm, Sweden. e-mail (corresponding author): fredrikb@nada.kth.se 2 Uppdaga HB, c/o Hamed Hamid Muhammed, Stenrösvägen 42, SE-75266 Uppsala, Sweden. e-mail: hamed@uppdaga.com 3 Dep. of Biometry and Engineering, SLU. P.O.Box 7032, SE-750 07 Uppsala, Sweden. e-mail: Anders.Larsolle@bt.slu.se Abstract This paper presents a technique called CAMSPECT where an external filter mosaic was placed in front of an RGB-camera. The external filter mosaic overlapped the camera's own RGB Bayer mosaic and up to 10 spectral bands could be produced. Field measurements of potato plant canopies in 2005 and 2006 with CAMSPECT and spectrometer resulted in comparable spectra. The small filter areas of the mosaic enabled the identification of details with differing spectral characteristics, and areas with for example deep shadows could be excluded from the analysis. In CAMSPECT images, plant could be separated from soil, which eliminated those errors caused by mixed spectra as compared to spot spectrometer measurements. Keywords: multi-spectral image, filter mosaic, potato, reflectance spectra, spectrometer Introduction Estimating status of vegetation through instantaneous measurements of spectral reflectance often implies the choice of either using spot high-resolution spectral measurements or using monochrome or colour image data. The optimal monitoring sensor should produce an image from where the object of interest could be identified and localised, and also having multispectral data in each pixel to be able to perform more efficient analysis. Such scenarios could be the estimation of stress through measuring reflectance from leaves in an image of a crop, excluding the disturbance from soil reflectance. Another example could be spectral identification and discrimination of plant species in an image, e.g. weed detection in a cereal or oil seed crop for variable rate spraying purposes. The objective of this paper is to present a new technique for imaging spectrometry, using an RGB-camera equipped with an external optical filter mosaic, and comparing the results with high-resolution spectrometer measurements. Materials and methods Recently, a novel cost-effective technique, called the CAMSPECT, has been developed where an external filter mosaic is to be placed in an intermediate image plane in a lens