In order to develop a low-cost and easy to implement technical solution to map inside-field spatial variability, and to explore its relationship with crop conditions, several experiments were conducted using ultra-light aircraft and Unmanned Aerial Vehicle (UAV) equipped with visible and infrared cameras. The sensors consisted of a ramp of 3 small format digital cameras (EOS 350D, Canon®): one for the visible part of the spectrum, and two modified cameras in order to acquire red edge and near infrared radiations. The images acquisition on the 3 cameras is simultaneous using external triggers and can be activated through the operator remote control on the ground or programmed to be automatically done using an on-board GPS navigation system. On ultra-light aircraft we also add a micro-bolometer thermal camera to the system. This paper describes the components of this acquisition system and focuses on the geometric and radiometric processing steps necessary for quantitative use of the data. At an altitude of 500 m this system acquires images with a ground resolution of 8 cm for the visible and near infrared bands and 55 cm for the thermal band. Unmanned Aerial Vehicle common altitude stretches over several tenth of meters up to 500 m and is adapted to the survey of fields of several hectares with very high spatial resolution. Ultra-light aircraft offers a range of altitude up to 1 to 2 km and a larger survey capacity with smaller spatial resolution. The spectral sensitivity of the cameras was measured using monospectral emittance sources. We worked both on the raw multispectral images and on the computed jpeg standard output. This allowed us to select the best band (or band combination) to produce red edge and near infrared images. We also developed an algorithm to compensate some radiometric distortion in the acquired images, particularly on vignetting effect. Classical photogrammetric calibration was used in order to measure lens geometry of each camera and evaluate as precisely as possible the coefficients of the lens polynom needed by commercial photogrammetric software. Several sets of images were acquired over experimental fields in temperate zone (on wheat) and tropical zone (on sugarcane). These images were radiometrically and geometrically corrected used the above elements and are stored as georeferenced stackable images in a Geographic Information System. The next step for a quantitative use of the data is to compensate changes due to atmospheric and illumination conditions in the image time series.