Quality Control of Hazel Nuts Using Thermographic Image Processing

This paper describes some new approaches to the quality inspection of hazels, analyzing thermographic images. On the one hand we present a quality inspection system checking large amounts of hazels carried on a conveyer belt and on the other hand a system, which permits a more detailed quality classification controlling single objects. We emphasize the characteristic properties of thermographic images, which lead to the kind of image processing algorithms we use. Thresholding and texture analysis algorithms, as well as fuzzy logic for the classification of single objects are applied. 1 General instructions Each year thousands of tons of hazels are processed or prepared for processing by the sweets industry. The procedure includes cleaning and classifying with respect to the quality of the hazels as well as the removing of foreign bodies. The detection of foreign bodies rests on mechanical, optical and ultrasonic methods. These techniques enable the detection of foreign bodies, which differ from hazels with respect to the mass, the colour or the surface density. By using these methods a large portion of foreign substances is detected, but unfortunately there is still a quality problem for the remaining ones. Moreover, there are also approaches, using the X-ray technology, but due to the radiation exposure they are often undesired for safety reasons. Within the framework of a research project, possibilities were investigated using infrared thermography for the online detection of foreign bodies in food and particularly in hazels. The different thermal behaviour of different materials is used to detect foreign bodies. A thermographic camera produces images that represent the thermal radiation of material in gray values. The food under investigation is slightly heated and then after a fixed period of cooling time the thermographic camera takes an image. The differences in the cooling behaviour appear in different gray values in the thermographic images. This property yields the choice of the image processing algorithms. In the first attempt algorithms resting exactly on the above mentioned characteristic of the gray level difference between the hazels and the foreign bodies were analyzed. Furthermore texture analysis algorithms were applied in order to find defects in the texture consisting of a high density packing of hazels. As an alternative approach, individual nuts were inspected in order to account object oriented features instead of measuring global properties of large quantities of hazels. 2 Thermography Studies concerning the possibility of detecting foreign bodies in food by using IR-thermography were reported by Meinlschmidt, Maergner et al. [4, 5, 6]. The results show that one obtains images with highest contrast, using a flash light to heat the hazels. Furthermore they discovered considering the cooling behaviour of both hazels and typical foreign bodies that the most advantageous time to inspect the hazels is about one or two seconds after the heating pulse. The thermographic camera used for the experiments is a Thermosensorik-System CMT 384 M [3]. The camera uses a matrix of 384 x 288 HgCdTe (CMT) stirling cooled infrared sensors, capable of detecting middle infrared radiation in the range of 3,4 – 5, m. The temperature resolution is < 20mK (NETD) and each pixel has a 14 bit resolution. The pixel pitch is 24 x 24 m and the maximum full frame rate can be up to 130 Hz. 3 Image processing algorithms In the following the image processing algorithms are described in detail, which were used to detect foreign bodies or nuts of bad quality in a thermographic image made of slightly heated nuts on a conveyer belt. Figure 1 shows an example of a photo and the corresponding thermographic image. Figure 1: a)Gray level photo of hazels with foreign bodies; b)Thermographic image captured 0.25s after heating the hazels with a flash (stones and nutshells are marked with a circle)