Indeed, in outdoor environments a variety of weather conditions may appear, i.e., highly sunny or cloudy days with different intensities, clear days alternating with different cloud densities, etc.Moreover, it is well known that in outdoor environments, particularly in sunny days, infrared radiation enters the sensor impacting the different spectral channels. The control of illumination factors is addressed in Section 2.2.2.Based on a given camera-based sensor with its corresponding accessories, we study the image accuracy for crop line and weed detection in agronomical images with specific reference to maize crops, where an area 3 m wide must be covered.
This accuracy is studied from two points of view, making the main findings of this paper: (a) geometrical arrangement, based on extrinsic parameters and (b) software corrections for improving the image quality, derived from the uncontrolled illumination in this kind of outdoor environments. This paper is organized as follows: Section 2 describes materials and methods used for accuracy determination considering the above two points of view. In Section 3 accuracy results are provided. Finally, Section 4 presents the relevant conclusions.2.?Materials and Methods2.1. MaterialsThe camera-based sensor consists of three essential physical parts: (a) CCD-based device embedded in a housing with its electronic equipment and interfaces for power supply and to the computer; (b) optical lens and (c) ultraviolet and infrared cut filter. Figure 1 displays these parts assembled as a whole in Figure 1(a) and separated in Figure 1(b).
Figure 1.CCD sensor, lens and UV/IR cut filter: (a) integrated. (b) Batimastat separated.The CCD is a Kodak KAI 04050M/C sensor with a Bayer color filter with GR pattern; resolution of 2,336 �� 1,752 pixels and 5.5 �� 5.5 ��m pixel-size. This device is part of the SVS4050CFLGEA model [20] which is robust enough and very suitable for agricultural applications. This device offers several externally controlled possibilities: (a) exposure time, which determines the time taken to capture the image; (b) Red, Green and Blue gains, where a value can be set for each channel, including gains auto-calculation; (c) definition of specific Regions Of Interest (ROIs); (d) information about the operating temperature. This Gigabit Ethernet device connected to a cRIO-9082 with dual-core controller, 1.33 GHz and LX150 FPGA running under LabView 2011 from National Instruments [21] is robust enough and specifically designed for real-time processing, so both features are very suitable for our agricultural application. Because the application occurs in harsh environments (containing dust, drops of liquid from sprayers, etc.