Home
Back
Soft Computing Research

Online Detection of Contaminants in Packaged Foods with Ultrasound using Signal and Image Processing and Soft Computing

Gauri S. Mittal1 and Otman Basir2
1School of Engineering, University of Guelph, Guelph, Ontario, Canada, and
2Department of Electrical and Computer Engineering, University of Waterloo, Waterloo, Ontario, Canada

Abstract

A signal processing algorithm for time-frequency analysis is used to discriminate between beverage and contaminants using ultrasound echoes. Effective detection ability for materials such as glass, metals, and plastic, with sizes from 10 mm2 to 2.5 mm2 was demonstrated. Unlike the time-gating for flaw detection by non-destructive testing, a foreign body was detected by examining the amplitude ratios between the echoes from the container’s outer and inner surfaces. This technology involved sensor design, signal processing, tomography, real-time pattern recognition and multi-resolution spectrum analysis.

1. Introduction

The presence of foreign bodies (FBs) in processed foods is one of the major concerns of the food industry, and their detection and identification play an important role in quality assurance and safety. When food products are manufactured or packaged, small foreign objects might end up in the product. For example, in the process of packaging food in glass bottles, occasionally fragments of glass result due to shattering and breakage of bottle top, and find their way to the glass containers. Contamination can also be the result of metal scarf that originate from production equipment. Plastic bags and gloves result in dough and ground meat. Bone pieces are found in ground meat and deboned reconstituted chicken products. Therefore, to prevent any potential harm to consumers and maintain their confidence it is desirable for the food industry that all FBs are detected and removed from food products before they reach the consumers. Ultrasound-based measurement is a promising method because it has ability to differentiate discontinuity of acoustic impedance between different regions within a given volume. In addition, it does not spoil foods physically or hygienically when used in non-destructive mode.

Optical method is a possible option for FBs detection but it is limited to clear transparency juices and bottles. Magnetic resonance imaging (MRI) could be another option but it is an expensive and complicated method. Ultrasound based measurement methods have demonstrated potential in various food processing applications, including: concentration gauging, flow measurement, level detection (Ridgway, Henthorn & Hull, 1998), food shelf life monitoring (Kulmyrzaev, Cancelliere & McClements, 2000), and food properties assessment (Mizrach, 2000; Benedito, Carcel, Sanjuan & Mulet, 2000). Povey & Mason (1998) provided an account of ultrasound food processing applications. Nevertheless, in comparison to medical diagnosis and industrial inspections (Birks & Green, 1991), the use of ultrasound in food processing in general, and in FB detection particularly, is still in its infancy. An example of research work in this area is that of Chivers, Russel & Anson (1995) where a clinical B-scanner was used to image stone pits embedded in peach flesh. Haeggstrom & Luukkala (2001) detected and identified FBs in cheeses by checking the amplitude pattern of reflected pressure obtained by water immersion method.

Issues in FB detection in juice bottles are mainly the ultrasonic energy transmission from emitter to bottles being inspected. It includes two aspects: coupling method and incident angle control. The first aspect requires a coupling method that is suitable for on-line inspection. Air-coupled transducer provides more flexibility in implementation during production. However its application is limited to low density container materials due to the reflection of most of the transmitted energy due to acoustic impedance mismatch (Birks & Green 1991). Water is a good couplant for both low and high density materials. However, water immersion coupling method is not practical for implementation in production lines. The second aspect is very important for both air-couple and water-couple method in receiving ultrasound signal. This issue becomes especially critical in evaluating beverages in containers with curved surfaces such as bottles. The ultrasound beam needs to be accurately aligned to the normal of container surface. Birks & Green (1991) reported auto-alignment or surface following devices but no further information was provided. Pahlawan (1996) presented a surface following technique used in pork grading by ultrasound. His method needs also an extra sensor.

In our study an ultrasonic system was developed for the detection of FBs in beverage containers. The prototype is composed of a pulse/echo transducer, a water jet nozzle and a x-y table (Fig. 1). Container bottom was scanned by an ultrasonic transducer mounted on the x-y table. Ultrasonic pulses were transmitted upward to the container bottom through the water jet functioning as a coupling liquid. The ultrasonic pulse was redirected back by the outer surface (front face) of the container bottom as the first reflection and by the inner surface (back face) as the second reflection. By examining the pressure ratio of the two echoes, which is relatively stable, FBs can be detected and identified. By checking the spatial distribution of the ratio, FBs can be localized and their sizes can be estimated. The theory and the algorithm to facilitate the recognition of two echoes in the joint-time frequency domain are described.

2. Theory

The following derivation presents how an ultrasonic pulse/echo method is used to detect and identify FBs in food containers. Symbols z1, z2, z3 and z4 represent (Fig. 2) acoustic impedances of the water jet, container bottom, water in container, and FB, respectively. The z1 and z3 are equal since both signify water impedance. The acoustic impedance is the product of sound speed in the material and material density. The following discussion is limited to 1-D plane wave. Assuming an incident pressure pulse of amplitude P on interface-1 (water jet/front face of container bottom), the reflected pressures from interfaces 1, 2 and 3 are calculated using Eqs. 1 to 3 (Kinsler, Frey, Coppens & Sanders, 1982):

where R1, R2 and R3 are, respectively, the reflections on interfaces 1, 2 and 3; and T1 and T2 are transmission coefficients from z1 to z2 on interface-1, and from z2 to z3 on interface-2 respectively. T1' and T2' are for transmissions in the opposite direction; α1 and α2 are attenuation coefficients of the container, which depends on its material, and of water respectively; and D and d are the thicknesses of the container bottom and that of the water layer between the FB and the container bottom, respectively. Reflection and transmission coefficients are calculated as:

Eq. 2 indicated that the pressure of the second reflection is a function of the impedance difference as calculated in Eq. 9. Therefore, from a theoretical standpoint, the presence of FBs can be detected by measuring the pressure amplitude of the second reflection; given that the incident pressure (P) is known. However, the incident pressure in many cases is unknown as it represents the pressure just before impacting the container. This pressure can be estimated based on the pressure propagation from the transducer through the delay line, to the nozzle, and finally to the outlet of the nozzle. This calculation is inaccurate due to signal attenuation, complexity of the nozzle geometry, instability of transducer driving voltage, and variations in the gap between the nozzle and the container bottom due to container surface irregularity. All these factors may lead to false FB detection.

In contrast, the pressure ratio between P2 and P1 is immune to such anomalies since the above variations vanish as a result of the division operation (P1 and P2 are both proportional to the incident pressure and are subject to the same uncertainties). This is evident from Eq. 10:

The pressure ratio is a function of impedances only. Any variation in the reflection coefficient R3 will change the pressure ratio. But R3 is a function of FB’s characteristics signified by z4 (Eq. 9). Hence, the electric instability is cancelled. Also, Eqs. 4 to 9 indicate that P2 and P1 are in phase, since T1, , T2 and are positive, and normally, both the FB impedances z4 and that of the container z2 are larger than water impedance: (z4 - z3) > 0 and (z2 - z1) >0. When the condition z4 = z3 is true, the value of R3 will be zero and hence P2 will also be zero. This suggests that there is no object in the ultrasound path. In this case, the pressure ratio of the two echoes from the front face and the back face of the container is calculated by:

where P2' is the echo from the back face of the container; the negative sign, coming from the reflection coefficient calculation on interface-2, indicates that the two pressures are in opposite phase.

Since T1, T1' and e-2α1D are known, the ratio P2'/P1 is predictable when no FB is present. In our case, P2'/P1 = 0.92 based on parameters listed in Table 1. If the value increases during transducer scanning, the FB is considered detectable. In the experiments, the container was of polystyrene with an attenuation coefficient of 37 neper per meter at 4 MHz (the peak frequency in the pulse spectrum generated by the transducer). This parameter is computed based on Kaye (1986), He & Zheng (2001), Kumar & Kumar (1996), and Wu & Stepinski (2000) for attenuation coefficient in solid medium. Once an FB is detected, the following Eqs. are used in identifying the FB based on its estimated impedance:

Eqs. 12 and 13 show that the determination of the impedance z4 depends on the measurement of P2/P1 for a given container material and beverage. In practice, the attenuation in water is 0.00012 neper/mm (Birks & Green, 1991) at room temperature, which is much smaller than that of the container material, and therefore can be neglected.

3. Frequency Domain Analysis

The equations in section 2 seem independent of frequency, as there is no explicit frequency term in these equations. However, this is not the case. The frequency dependence comes from pressure values which are derived from the frequency spectrum of the reflections. To enhance the signal to noise ratio in order to obtain accurate signal peak values from the spectrum, an algorithm for signal selection and separation is needed. In ultrasonic thru-transmission mode, signal selection and separation are not issues since there is only one pulse signal the receiver detects. In the pulse/echo mode used in this study, the transducer, as it operates in dual mode – both transmitter and receiver, will receive at least two reflected signals from the front and back interfaces of the container bottom. If the bottom is thick enough, the two signals will be well separated and spectrum calculation of each echo can be performed by augmenting rest of the signal by zeros (He, 2000). If the bottom is not thick enough, which is our case, one has to determine where to separate the two overlapped signals. To overcome this problem, an algorithm of time-frequency analysis was developed. In this algorithm, a sliding window of width N was used to chop the sampled signal. A Fast Fourier Transform (FFT) was applied to these N points to determine the amplitude of the pulse peak frequency. The window was then moved forward one point along the signal to perform the same operation. Repeating this process throughout the signal, the history of the pulse amplitude at peak frequency was obtained. The two echoes appear as two separated peaks in the time history.

The window width N was important for this technique as the pulse amplitude will be lost if N is chosen to be less than the pulse duration period. If N is chosen to be very large, the correct timing and amplitude information of the signals for two echoes from front face and back face may distort due to the energy merging between the two echoes. In our calculations, we found N equal to 25, which corresponds to pulse period of 4 MHz peak frequency at sampling frequency of 100 MHz, to be effective. To minimize the frequency leakage error on the spectrum amplitude (Cartinhour, 2000), a Hamming window was used prior to FFT operation.

4. Experimental Procedure

The experimental set-up of the prototype of the proposed system is shown in Fig. 1. A polystyrene container with flat bottom of thickness 0.75 mm was supported by a holder. An ultrasonic transducer with delay line was mounted on the bottom of a plexiglass water jet nozzle that was supplied with water by a controllable pump. The delay line was made of a plexiglass cylinder of 7.40 mm in diameter and 10 mm in length that produced a time delay of 7.6 µs. The nozzle has a 3 mm diameter at exit. The ultrasound pulse was transmitted to the bottom of the polystyrene container through the water jet. The ultrasound signal traveling distance was 23.7 mm from the delay line to the exit of the nozzle. A 2 mm gap was kept between the container bottom and the nozzle tip so that the transducer mounted to the x-y table could scan the container bottom smoothly. The water flow was controlled at a rate of 40 L/h. The transducer used was a flat-focused ultrasonic transducer (GRD-1502-HR, Technisonic Research, CT) of nominal frequency 15 MHz. The ultrasonic pulse was coupled to the water by a delay line. A SR-9000 Pulse/Receiver card (Matec, MA) was used to drive the transducer and to receive the echo signal. A pulse of 4 MHz peak frequency was produced by SR-9000 at a repetition rate of 2 kHz. The signal sampling frequency was 100 MHz.

Five specimens: plexiglass, glass, aluminum, stainless steel and copper were tested. All these specimens were cut into pieces of 10 mm in square whose thickness and acoustic parameters are listed in Table 1. The specimens were placed in the container filled with water one hour before the experiment. The water depth in the contained was about 10 mm.

Labview

A Labview 5.0 program (National Instruments, Cambridge, Ontario, Canada) was developed to control the SR-9000 card and the x-y table, and to process the echo signals simultaneously. The program consists of a front panel, a block diagram and an icon/connector. The panel includes switches, knobs, and graphical outputs that provide the appearance of an instrument. Fig. 3 is a front control panel of the main program for the FB detection system. It is composed of three sections: The first section appearing in the upper left corner of the control panel is to control the movement of the x-y table. The detection procedure is designed to scan the bottom of a beverage container line by line. By inputting the starting position (X0, Y0), ending position (Xn, Yn) and step length (Xstep,Ystep), the transducer scans the container bottom using the x-y table. The real-time position of the transducer is monitored and displayed on the monitor. The second section, “P – E control” (Fig. 3), is to control the signal delay time and signal size, and also controls sample frequency. The third section is to process and display the sampled signal (left lower part window control and right lower part). This includes the signal slicing, offset level removing, windowing, FFT algorithm, and pressure ratio calculation for peak frequency.

The main program runs in the following order (i) positioning the transducer using the x-y table, (ii) sending and receiving ultrasound signal, and (iii) signal processing. This timing is realized using LabView’s sequence structure. The first sequence that controls the x-y table’s movement using input parameters given on control panel. The second sequence provides a waiting time. This waiting time ensures that ultrasound signal is emitted and received after the x-y table is mechanically stopped to avoid any measurement when the x-y table is moving. This avoids errors for FB localization and reduces signal to noise ratio. The third sequence of the main program, which performs the signal sampling and displaying. These three sequences are planted and executed in “While loop”, symbolized by an unclosed wall with an arrow at its break. “While loop” is a post-iterative-test structure that repeats a section of the code until a condition is met. The “While-loop” is controlled by the Boolean continue terminal situated near the lower right corner, which receives the output of the sequence two. The loop will stop if the output is logic 1, otherwise the loop will continue.

5. Results and Discussion

Fig. 4 depicts the signal reflected from a FB free container. The echo timing is shown for the piezo/plexiglass interface delay line to the water jet/container-bottom interface, where 7.6 µs was the round trip time of ultrasound in the delay line, and 32 µs was that of the water in the nozzle. The 32 µs delayed tip pulse was due to the discontinuity of tip boundary. After the reflection on the interface of delay line/water, a pulse was observed representing the second reflection on the same interface. Some noise before the container outer surface reflection time could be attributed to shear wave of the delay line. The reflections from front and back faces were after the nozzle tip pulse and occurred approximately after 42 µs.

Fig. 5 shows two reflections having a time difference of 0.69 µs, which corresponds to the round trip time between the front and back surfaces of the container bottom. The two echoes are in opposite phase because of the impedance discontinuity, and agree very well with the analysis in the preceding sections. Fig. 6 shows two echoes for a specimen of glass placed in a water container. The two echoes are in phase since the conditions (z4 - z3) > 0 and (z2 - z1) > 0 are satisfied. In fact, the phase information could also be used as criterion for FB detection. In Figs. 5 and 6, two echoes seem to be overlapped and are difficult to determine the separation point.

Fig. 7 is a time history of the amplitude at peak frequency of 4 MHz for the data of Fig. 5 using the Hamming window technique. The first peak is the amplitude of reflection of the container’s front face and the second peak is that of the back face. The time difference between the two reflections remains unchanged similar to that for the real time signal (Fig. 5), but the peaks were easy to sort out. Using the two peak values, the pressure ratio was computed. If the pressure ratio is larger than a threshold value (see section 2), an FB is considered to be present.

Using the slide window method and the pressure ratio criterion, specimen of glass in the container was inspected. It was found that the specimen can be detected as small as 2.5 x 2.5 mm square that is smaller than the nozzle cross section area. Fig. 8 is a schematic of FBs detection by using an ultrasonic beam scanning with a x-y table. By this way the FBs’ size can be estimated by directly reading the distribution of reflections pressure ratio. Fig. 9 is a pressure ratio distribution of a glass specimen of 2.5 x 2.5 mm. The resolution of x-y table was 0.025 mm. The scale in the Fig. is 0.25 mm per division. A matrix of 17 x 18 for pressure ratio was obtained for 17 line scannings with 18 measurements on each. Roughly the amplitude ratio is larger when the transducer is right underneath the specimen rather than when far away from the specimen. The space between two peaks is 5 divisions corresponding to 2.5 mm of the specimen size. However, the maximum value did not occur in the center position, since the ultrasound beam was not perfectly perpendicular to the front face due to nozzle tilting.

An identification experiment was conducted with FB size 10 x 10 mm2 to eliminate the edge effects. Table 2 provides a summary of the results for the measurement of P2/P1 and a comparison with the theoretical predictions. These experiments were performed by applying the ultrasound beam right underneath the FB. The experimental results are in agreement with the theoretical calculations with a maximum relative error of 6.3%.

The pressure ratios with FB are approximately 3 times larger than those without FB. This confirms that using pressure ratio comparison is an effective criterion for FB detection. Three orange juices, with and without pulp, have similar pressure ratios in the absence of FB (Table 3). This may be due to the absence of pulp at the container-juice interface in the path of the ultrasound beam. In the presence of FB, the pressure ratios for pulp-free and pulpy orange juices (fairly large pulps) are identical. The pressure ratio for orange juice with medium size pulp is about 10% less than that of the former two orange juice types. This could be attributed to pulp insertion between the container bottom and the FB. The pressure ratios in the cases of juices with FB are smaller than those of water.

6. Further Discussion

The modeling of the method presented in the theoretical part is based on the assumption of plane wave 1-D propagation. The transversal dimension of the container bottom and FB should be much larger than the ultrasound beam width and the interface should be orthogonal to the beam when the inner surface reflection P2 and P2’ derivation is performed. In practice, these two conditions can be fairly met for the container bottom: the diameter of a container is normally in the order of 80 mm compared with ultrasound beam width of 6 mm, and the ultrasonic transducer can be aligned to the surface of the container. In this case, the pressure ratio (P2’/P1) in Eq. (11) in the absence of FB can be a priori calculated or measured for a given product container. This was confirmed by the experimental results of polystyrene container (Table 2).

In a real beverage container, dimensions of a real FB are no longer larger than the ultrasound beam width. The beverage space between the bottom and the sedimented FB is no longer a uniform layer, and FB geometric shape and position are variables. The P2/P1 (in Eq. 10) does not hold anymore. Material identification by Eq. 10 can not be performed. However, the presence of a FB in the path of ultrasound beam does change the reflection signals by superposing the reflections from FB to that from the inner wall surface, as long as FBs’ impedance is different from that of beverage. Values of the superposing depends on the impedance difference between FB and beverage, FB’s cross-section to ultrasound beam, and roughness of FBs. Beverage layer between the FB and the container bottom affects time shifting between two reflections of the superposition. For a FB sedimented at the bottom, the time shift is not perceptible. For a FB suspended in beverage, the time shift demonstrates itself by an echo after the reflection of inner face, if its intensity is strong enough. Quantitative evaluation for FB reflection using analytical solutions is given by Haeggstrom & Luukkala (2001) for large size FB (as compared to wavelength), and by Pierce (1981) for small size scatter. For industrial inspections, ability to measure the pressure ratio deviation from the criterion is important.

To test the sensitivity of the proposed method, the following experiment was conducted. Three steel balls of diameter 2.38, 3.17 and 4.73 mm were respectively placed on a flat glass plate of thickness 3.03 mm. Balls, glass plate and a transducer of center frequency 7 MHz were immersed in a water tank. Balls were selected due to the following reasons. Acoustic cross-section of a ball is smaller than its geometric cross-section since much of the sound incident on the ball is not reflected back to the transducer. This represents, to some extent, the reflection characteristics of randomly formed FBs. Liquid layer between the ball and the glass plate is no longer of equal thickness. This represents also, to some extent, a real FB sedimented on the bottle bottom. Balls are similar in the path of ultrasonic beam, so that the size effect of a FB on the reflection signals can be compared without the worry of its positioning to the beam.

Signals without and with FBs (steel balls) in the path of an ultrasound beam were sampled and are shown in Fig. 10. It is shown that signals obtained at different reflection conditions are almost superimposed each other, and it is difficult to find the difference among the signals. Using the sliding window algorithm presented in the third section, the pressure ratios between front and back face reflections were measured at frequencies from 5 to 10 MHz about the center frequency of 7 MHz. Fig. 11 shows that the pressure ratio “with FB” is higher than those “without FB” by about 1.3% for frequencies up to 8 MHz. Pressure ratios for smaller diameter FB deviate further from that of “without FB”. Error bars in Fig. 11 also show the standard deviation for each pressure ratio calculated using 100 samples. The pressure ratios “without FB” are not merged with those “with FB” for frequencies from 6 to 8 MHz even after considering measurement errors. This indicates that the method proposed in this study is sensitive for FB detection and robust to noise in the neighbourhood of center frequency.

Another challenge to this method for FB detection comes from the non-parallelness of the outer and inner surfaces of the bottle bottom. This non-parallelness leads to either the decrease of reflected signal intensity or the loss of the reflection signal to the transducer, even the incidence of ultrasound is at right angle to the outer surface of the bottle bottom. This is because the ultrasound beam follows Snell’s law in propagation. To overcome this difficulty, use of a transducer array would be effective, where transducers in the array are fired one by one and all of the transducers are at the same time set on “receiving” after one pulse is fired. By this way, part or the whole lost echo due to non-parallelness can be recorded by the neighbour transducer of the emitter.

7. Conclusions

An ultrasound system for real-time detection of foreign bodies in food containers was described and worked satisfactorily. The system consisted of an ultrasound echo/pulse transducer mounted on an x-y table and coupled with a food container using a water jet. Time-frequency domain analysis worked reliably to locate different interfaces in the detection. The proposed detection system is capable of detecting fine FB’s for various materials including glass contaminants. The object-oriented graphical programming language (LabView) is suitable to control an ultrasonic scanning system composed of an ultrasound transducer, an x–y table and a computer for foreign body (FB) detection inside beverages packaged in containers.

This also describes the principle and design of an auto-aligned ultrasonic transducer system using a water jet for foreign body detection inside beverages in glass bottles. Variation of reflection amplitude as a function of the ultrasound beam incident angle to beverage container surface is proven sensitive theoretically and experimentally and suitable for being as feedback for auto-alignment control. Experiments conducted using the prototype showed its ability to detect foreign bodies inside juices in glass bottles. Interesting echo signals due to curved wall were found and discussed for ultrasound propagation in bottles. Threshold in confined time region is therefore set to avoid false warning for detection in low signal to noise ratio. This design is also applicable to non-destructive inspection for metal canned food.

8. References

Benedito, J., Carcel, J. A., Sanjuan, N., & Mulet, A. (2000). Use of ultrasound to assess cheddar cheese characteristics. Ultrasonics, 38, 727–730.

Birks, A. S., & Green, R. E. (1991). Ultrasonic testing: Nondestructive testing handbook, vol. 7, (p. 226 and pp. 837-839). Columbus, Ohio: American Society for Nondestructive Testing, Inc.

Cartinhour, J. (2000). Digital signal processing (pp. 253-258). New Jersey: Prentice Hall.

Chivers, R. C., Russel, H., & Anson, L. W. (1995). Ultrasonic studies of preserved peaches. Ultrasonics, 33, 75-77.

Haeggstrom, E., & Luukkala, M. (2001). Ultrasound detection and identification of foreign bodies in food products. Food Control, 12, 37 – 45.

He, P. (2000). Measurement of acoustic dispersion using both transmitted and reflected pulses. Journal of Acoustic Society of America, 107, 801-807.

He, P., & Zheng, J. (2001). Acoustic dispersion and attenuation measurement Using Both Transmitted and Reflected Pulses. Ultrasonics, 39, 27-32.

Kaye, G. W. C. (1986). Tables of Physical and Chemical Constants and Some Mathematical Functions (p. 76). New York: Longman Inc.

Kinsler, L. E., Frey. A. R., Coppens, A. B., & Sanders, J. V. (1982) Fundamentals of acoustics, (pp. 125-126). Toronto, Canada: John Wiley & Sons.

Krautkramer, J., & Krautkramer, H. (1969). Ultrasonic Testing of Materials, (p. 235 and pp. 476-477). New York: Springer-Verlag.

Kulmyrzaev, A., Cancelliere, C., & McClements, D. J. (2000). Characterization of aerated foods using ultrasonic reflectance spectroscopy. Journal of Food Engineering, 46, 235-241.

Kumar, B., & Kumar, A. (1996). Evaluation of Ultrasonic Attenuation without Invoking the Diffraction Correction Separately. Ultrasonics, 34, 847-853.

Mizrach, A. (2000). Determination of avocado and mango fruit properties by ultrasonic technique. Ultrasonics, 38, 717 – 722.

Pahlawan, R. (1996). Surface following and detection of the normal to a surface – application to ultrasonic – based pork carcass grading. MS thesis, University of Toronto.

Pierce A D. (1981). Acoustics: An introduction to its physical principles and applications, (pp. 424-431).New York, McGraw-Hall.

Povey, M. J. W., & Mason, T. (1998). Ultrasound in Food Processing. London: Blackie Academic & Professional.

Ridgway, J., Henthorn, K. S., & Hull, J. B. (1998). Controlling overfilling in food processing, In Ultrasound in Food Processing, pp. 1-16, edited by Povey, M. J. W., & Mason, T. London: Blackie Academic & Professional.

Wu, P., & Stepinski, T. (2000). Quantitative estimation of ultrasonic attenuation in a solid in the immersion case with correction of diffraction effects. Ultrasonics, 38, 481-485.


Fig. 1. A prototype of a foreign body detector using an ultrasonic technique.
(All dimensions are in mm)


Fig. 2. A schematic of transmission and reflection calculations.


Fig. 3. Front control panel of an ultrasound foreign bodies detector using LabView.


Fig. 4. Timing of ultrasound signal without foreign body in the path of ultrasonic pulses.


Fig. 5. A zoomed view of Fig. 3 around 42 µs.


Fig. 6. Pulses in phase between the reflections from front and back faces.


Fig. 7. Amplitude-time at peak frequency of 4 MHz for FB-free container calculated by sliding window algorithm.


Fig. 8. An ultrasonic water jet nozzle scanning path and various dimensions.


Fig. 9. Pressure ratio distribution around a glass specimen of 2.5 x 2.5 x 1.2 mm.


Fig. 10. Real time signals sampled at conditions of without and with FB at 7 MHz center frequency.


Fig. 11. Results and their errors calculated by sliding window algorithm for pressure ratios between the front and back faces at different frequencies and conditions without and with FB.


Table 1. Thicknesses and acoustic impedances for materials used


Table 2. Experimental results and comparison with theoretical predictions for P2/P1


Table 3. Experimental results of pressure ratios for FB detection in various juices


About the Authors

GAURI S. MITTAL, P.Eng., Professor of System and Food Engineering at the School of Engineering, University of Guelph, Guelph, Ontario, Canada. An author of more than 270 refereed journal research papers and 210 other publications, as well as three books and various book chapters. He is the recipient of many awards. A registered professional engineer, professor Mittal received the B.Sc. from India (1969); the M.Sc. (1976) from the University of Manitoba, Canada, and Ph.D. (1979) from the Ohio State University, USA. Mittal presently conducting research in the areas of modelling, simulation and optimization; non-thermal pasteurization of liquid foods using pulsed electrical field; sensor and detection systems development for food quality and safety; image processing for food quality and safety, robotics in food processing, and novel techniques for food processing.

OTMAN BASIR, holds a Ph.D. degree in Systems Design Engineering from the University of Waterloo, Canada, a M.Sc. degree in Electrical Engineering from Queens University, Kingston, Canada, and a B.Sc. degree in Computer Engineering from al-Fateh University, Libya. He is currently an Associate Professor with the Department of Electrical and Computer Engineering, University of Waterloo, Dr. Basir is the Associate Director of the Pattern Analysis and Machine Intelligence Laboratory. His research interests include Intelligent Transportation Systems, Embedded Real-time Systems, Sensor Networks, Sensor and Decision Fusion, and Biologically Inspired Intelligence.