**2. Image acquisition system**

Image acquisition is the most important step for robust image processing. Indeed, the use of natural images involves some difficulties tied to outdoor conditions (lighting variations) and object complexity (the leaf area presents several planes, high contrast and lighting variations, scale variations, wheat growth stages ….). Photographic slides were taken at different growing stages of the wheatears.

Different solutions exist for image acquisition according to the vertical distance between the camera and the scene (1m, 2m or more), to the illumination used (natural or controlled light), to the kind of images (color, grey level, hyperspectral ...)

Figure 1 shows one solution based on specific boxes with opaque protection and controlled illumination by power-leds, for wheat ear counting, allowing to obtain color or grey level images.

Fig. 1. Example of image acquisition system (left) and its specific illumination by Led (right)

Figure 2 shows some examples of color images of wheat crop at three different growth stages with controlled illumination.

We propose in this chapter to explore the proxy-detection domain by focusing first on the development of robust image acquisition systems, and secondly on the use of image processing for different applications tied on one hand to wheat crop characterization, such as *the detection and counting of wheat ears per m² (in a context of yield prediction) and the weed detection*, and on the other hand to the evolution of seed development/germination performance of chicory achenes. Results of the different processing are presented in the last

Image acquisition is the most important step for robust image processing. Indeed, the use of natural images involves some difficulties tied to outdoor conditions (lighting variations) and object complexity (the leaf area presents several planes, high contrast and lighting variations, scale variations, wheat growth stages ….). Photographic slides were taken at

Different solutions exist for image acquisition according to the vertical distance between the camera and the scene (1m, 2m or more), to the illumination used (natural or controlled

Figure 1 shows one solution based on specific boxes with opaque protection and controlled illumination by power-leds, for wheat ear counting, allowing to obtain color or grey level

Fig. 1. Example of image acquisition system (left) and its specific illumination by Led (right)

Figure 2 shows some examples of color images of wheat crop at three different growth

part just before a conclusion.

images.

**2. Image acquisition system** 

different growing stages of the wheatears.

stages with controlled illumination.

light), to the kind of images (color, grey level, hyperspectral ...)

Fig. 2. Example of image acquisition system (left) and its specific illumination by Led (right)

Proxy-Hyperspectral images can also be used and are obtained with specific sensors. As an illustration, images of wheat crop have been acquired in march 2011 on the domain of Melgueil (INRA, 34-France) by means of dedicated apparatus developed by the Cemagref. This device is constituted by a motorized rod installed on a tractor which allows a longitudinal displacement of a push-broom hyperspectral camera located 1 meter above the vegetation (figure 3).

Fig. 3. Hyperspectral acquisition device

Other imaging system is based on chlorophyll fluorescence. A xenon light source (Hamamatsu Lightingcure L8222, model LC5, 150 W) is used to produce a white light which passes through an interference filter (03FIB002, Melles Griot, Carlsbad, USA) with a central wavelength 410 nm and width at half-maximum 80 nm to excite the chlorophyll-a. The light is conducted to the seeds by an optical fiber. The blue light issued from the filter is absorbed by the seeds chlorophylls, which resulted in fluorescence emission. A high-pass filter (665 nm, 03FCG107, Melles Griot 03FCG107) ensured the selection of the fluorescent signal from the blue light reflected by the object.

The images are acquired by a CCD monochrome camera (Hamamatsu C5405-70), with a resolution of 640 x 480 pixels and 256 gray levels. The system is enclosed within a black box to avoid interference from ambient light and is air-conditioned (Tectro TS27, PVG Int. B.V., Oss, The Netherlands) to maintain a temperature of 20°C. The lamp is optically isolated and white light did not escape inside the box. The blue filter transmitted an unexpected, small amount of infrared light between 770 and 900 nm (detected with spectrometer AVS-SD2000, Avantes, Eerbeek, The Netherlands).

Texture, Color and Frequential Proxy-Detection

(corresponding to the four angles) and we note:

Fig. 5. Examples of the three textural patterns

energy (angular second moment) :

inverse different moment :

the cooccurrence matrix elements for the 3 texture classes:

'

 

*P <sup>p</sup> <sup>P</sup>* 

> 

 ' , '

'

(1973).

progressive.

 

Image Processing for Crop Characterization in a Context of Precision Agriculture 53

( , ) and ( ', ') '

For the other angles, the corresponding frequencies are defined similarly. In this document, we consider the distance **d** = 1 (this means we consider neighboring pixels). Moreover, texture of the wheatears, and, more generally, of the pictures, are not oriented along a common direction. Thus, for each pair of grey levels we sum over the four frequencies

' *P P*( , ', ,1)

Generally cooccurrence matrices are not used directly because of their size. The principle is to compute a set of measures from these matrices, such as those proposed by Haralick

Figure 5 shows three samples of ear ,soil and leaf textures: wheatearstexture reveals many transitions between very different grey tones. For leaves and soil the transitions are more

ears leaves soil

2 '

 

(3)

(4)

Thus, we compute four Haralick's features which characterize the particular disposition of

2 '

, ' *p* 

1 1 ( ') *<sup>p</sup>*

 

, '

 

 

(1)

(2)

' = , = ' *fij fi j*

*i i d jj*

 

is the element of the normalized cooccurrence matrix.
