Narrow-Beam Antenna for Short-Distance Non-Destructive Sensor in Fruit-Ripeness Monitoring

Radiation pattern subtraction is a useful technique for narrowing a receiving antenna’s radiation pattern in a short-distance, non-destructive sensor. The sensor’s antenna was developed to be compact and light-weight as well as to operate in real-time. This work applied the principle of radiation pattern subtraction to narrow the beamwidth of an antenna. Our experimental data demonstrated that a simple narrow-beam three-element antenna with a 35° half-power beamwidth and a small size of 2 5.9 6.5 cm  could be realized. In the experiment, the radiation pattern of an array antenna (consisting of two halfwave dipoles with a reflector) was subtracted from the radiation pattern of an element antenna (consisting of a halfwave dipole with a reflector) for the three-element antenna to achieve a narrow beamwidth and compact size. The antenna had a simple feeding network that operated at 10.525 GHz and 35° half-power beamwidth. The angular resolution of the receiving antenna was tested with mango fruits: it was able to resolve an 80 mm separation between the fruit of interest and the nearby fruit, which was sufficient, in terms of practicality, for mitigating the interfering effect of nearby fruits to the fruit of interest. This type of antenna is useful as a short-distance, non-destructive sensor such as a pre-harvest sensor. INDEX TERMS Pre-harvest sensor, short-distance, non-destructive sensor, angular resolution improvement, radiation pattern subtraction, dipole array, far-field phase distribution, top-load dipole.


I. INTRODUCTION
Recently, the United Nations (UN) has announced 17 sustainable development goals (SDG) to make the world free of poverty, hunger, and disease [1]. The 12th sustainable development goal (SDG12) mandates responsible and sustainable food consumption and production. Producers must be able to grow more food while reducing negative environmental impacts. Consumers must be encouraged to shift to a nutritious and safe diet with a low environmental footprint. In the sense of global food security, increasing production through the addition of more production areas is not sustainable. On the other hand, increasing productivity per unit of land is sustainable. For fruit production, the timing of harvest is critical. Over or under-ripe fruits have low or no retail value and are considered a loss or waste of resources. For consumers, fruits harvested too early would have low quality and unacceptable taste. Fruits harvested too late have a short shelf life, poor appearance, and an "off" flavor and odor. Infield assessment of fruit ripeness with an accurate prediction of both the harvest date and yield can significantly reduce waste in the supply chain in the addition to improve food security. To predict an optimal harvest date, a nondestructive technique (to the fruits) for determining the exact ripening date is necessary [2]. The ripening stage can be predicted as well as the rate of ripening that depends on environmental factors.
Numerous modern sensor technologies in fruit harvesting have been developed. Computer vision and proximity sensors (non-vision) can detect spatial features of plants and fruits. They can be used for plant inspection, determination of fruit size and shape, insect detection, and firmness measurement. Ripening stage can also be estimated by computer vision, This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2020.3044908, IEEE Access Author Name: Preparation of Papers for IEEE Access (Month 2020) 2 VOLUME XX, 2020 chemical sensor, and spectroscopy [3]. Computer vision and chemical sensors have severe limitations of interference from sunlight and the concentrated gas in the closed detection system. Microwave technology that is oblivious to these kinds of interference is a better candidate for the task.
Microwave radar at S-band has been developed to estimate grape production yield in the wine industry [4]. It was useful for logistics and supply chain management. The sugar content of apples can be estimated with a millimeter-wave reflectometer [5]. The estimated results from a reflectometer calibrated by an artificial neural network were entirely accurate. The natural frequencies of thick-peel fruits can be determined from a radar cross-section at L-band by the Cauchy method [6]. As the dielectric property of fruit flesh varies with time during the ripening process, its natural frequency can be used to predict a suitable harvest date. Those previously mentioned papers are some fruit classification applications of non-destructive sensors that measure the radar cross-section (RCS) of a target at a short distance. It is a feasible pre-harvest fruit sensor technology since the RCS of fruit changes with passing days [7]. However, if the sensor antenna's beamwidth is too wide, the angular resolution (the minimum angular separation at which two equal-sized targets can be separated when they are placed at the same distance from the sensor [8]) will be too low. In other words, measurement results from a sensor with too broad beamwidth will be too prone to errors due to nearby fruits' presence. Angular resolution characteristics are determined by antenna beamwidth, which is represented by a half-power point (-3 dB) angle. It is necessary to improve the pre-harvest sensor's angular resolution to obtain the most accurate hence the most reasonable estimates. Generally, a narrow beamwidth can be achieved from a large aperture antenna that requires a sizeable far-field range for the plane wave to impinge on the target. It can operate in long-range. One of the attractive approaches that can accomplish a compact size sensor antenna is an approach that uses radiation pattern subtraction, previously applied in radiation pattern measurement field [9]- [10]. Recently, it was applied to improve the spatial resolution of an ultrasonic imaging system [11]. Nevertheless, the phase was not taken into account. Sakogawa et al. [12] presented a beam subtraction scheme to tighten the RFID sensor's beamwidth for a security system. Their work employed hybrid couplers to perform radiation pattern subtraction, and the antenna structure was accomplished for a monopulse procedure. The application of interest was a sensor in a shoplifting warning system.
We presented a technique to narrow down beamwidth by subtracting two different radiation patterns. The developed antenna system was intended to be used as a pre-harvest fruit sensor [13].
Such an antenna needs a memory unit and a signal processing unit. However, the radiation pattern subtraction antenna would be done with a suitable feeding network to remove the need for memory and signal processing unit. The antenna would be operating at 10.525 GHz. It was designed to be compact and light-weight so that it could be installed on a drone. To our best knowledge, there has not been any recent work focusing on the investigation of radiation pattern subtraction related to far-field phase distribution of antenna. This paper presents the details of the principle of radiation pattern subtraction that had led to the simple design and implementation. This work's contributions are a wholly derived set of far-field equations from the radiation pattern subtraction principle, applied to such kind of antenna mentioned above, and a practical and straightforward antenna as a pre-harvest sensor. That set of equations allowed us to determine the feasibility of achieving narrow beamwidth by subtracting array antennas' radiation patterns. Instead of switching between two array antennas' radiation patterns and memorizing each pattern for subtraction, a feeding network was introduced so that the antenna would be compact and light-weight.
The rest of this paper is organized as follows. Section II presents a problem statement of the effect of a nearby fruit on the measured RCS of the fruit of interest and the sensor antenna's desirable characteristics. Section III describes the principle of radiation pattern subtraction related to far-field phase distribution. The design and experiments on the antenna characteristics, as well as its performance in eliminating the effect of nearby fruit, are described in section IV. Finally, the conclusion is drawn in section V.

II. DESIRABLE CHARACTERISTICS OF AN ANTENNA FOR SHORT-DISTANCE, NON-DESTRUCTIVE, FRUIT-RIPENESS MONITORING SENSOR
When a nearby fruit is sufficiently far away from the fruit of interest, the problem of scattered waves from such fruit is negligibly small. However, when a nearby fruit is located in close contact with the fruit of interest, the back-scattered wave from the nearby fruit will alter the fruit of interest's measurement result. Therefore, the beamwidth of the sensor's antenna must be sufficiently narrow. Fig. 1(a) shows a scenario of mango fruits (Mangifera indica) on a tree. It can be seen in the diagram in Fig. 1 (b) that the angle s  is measured from the direction pointing to the fruit of interest to that pointing to the nearby fruit, which is s d apart. Mango fruit is assumed to have an ellipsoidal shape. If the nearby fruit is in contact with the fruit of interest 0 () dd = , the angle between the centers of the two fruits is 0  . The distance between the sensor and the fruit of interest () r is in the far-field range of the sensor's antenna 2 ( 2 ) rD   , where D is the largest dimension of the sensor's antenna and  is the wavelength at the operating frequency.
The intended sensor operates as a monostatic radar. A detected mango's radar cross-section ( ) RCS : is determined in terms of gains of transmitting and receiving antennas, power of the transmitter and the receiver, and distance between the sensor and the mango () r [8].
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2020.3044908, IEEE Access Author Name: Preparation of Papers for IEEE Access (February 2017) VOLUME XX, 2020 3

FIGURE 1. A scenario of mango fruits on a tree (a) Photograph (b) Diagram (c) Required normalized gain in the direction
Naturally, fruits of the same age have the same size and dielectric properties; hence it is reasonable to assume that they have the same RCS. Fig. 1  The nearby fruit's scattered wave power is required to be less than 10% of that of the fruit of interest to limit the interfering RCS of the nearby fruit. For a short-distance microwave sensor application, the separation () s d between the fruit of interest and the nearby fruit can be comparable to the distance () r between the sensor system and the fruit of interest (I). Therefore, the received power from the nearby fruit (N) not only depends upon the beamwidths of the transmitting and the receiving antennas, but also upon the angle () where o G  is the normalized gain of the sensor's antenna in the direction 0  .
The standard size of mango is the parameter that controls the value of the distance 0 () d . The far-field range of the sensor's antenna controls the minimum distance () r that the sensor can be placed far away from the mango of interest. The angle Dr   . In practice, a standard mango size is determined by the average weight, length, and width of a batch of mangoes. The standard length and width of exported mangoes are 18 cm and 8 cm, respectively [14]. Therefore, the distance This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2020.3044908, IEEE Access Author Name: Preparation of Papers for IEEE Access (Month 2020) 4 VOLUME XX, 2020 considering the case of two linear uniform arrays that have the same kind of element and same phase center, shown in Fig. 2 (a). The desired radiation patterns of the array antenna #1 and the array antenna #2 can be generated by setting appropriate values for the parameters ( ) 11 , Fig. 2 (a), the received electric field intensities at array antenna 1 (output of power combiner #1) and array antenna 2 (output of power combiner #2) were combined at the power combiner #3 where the inputs were designed to be out-of-phase. The array antennas #1 and #2 utilized in this work are shown in Fig. 2 (b). They were two-element array antennas of which each element was a halfwave dipole with a reflector. The electric field intensity of the array antenna can be calculated by the pattern multiplication principle shown in [15]. To achieve the end-fire radiation pattern of a halfwave dipole with a reflector, the distance () h was set to 4  .
where h is the distance between the halfwave dipole to the reflector, and , d  are the distance and phase shift between the two elements, respectively. The first square bracket in (3) represents the radiation pattern of a halfwave dipole. Those in the second and third square brackets are the contributions from the reflector and the two-element array antenna, respectively.
The far-field phase of the halfwave dipole with a reflector can be calculated from the far-field phase of the halfwave dipole element ( 2) where  is the phase shift between the array and its element, calculated by, 0 for cos , for cos , Eq. (9) indicates that the far-field phase of the two-element array varies with the distance ( ) This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2020.3044908, IEEE Access Author Name: Preparation of Papers for IEEE Access (February 2017) VOLUME XX, 2020 5  −  , the subtracted radiation pattern had a main beam in the direction of 54 and a sidelobe in the direction of 117 as depicted in Fig. 3 (a). The half-power beam width (HPBW) of the subtracted radiation pattern was 38.3 . The sidelobe level was −13.9 dB.
Case 2: For a shorter distance between elements, the sizes of array antenna #1 with ( ) 11   The HPBW of the subtracted radiation pattern was 39.5 , and its main lobe and sidelobe directions were 53 and120 , respectively. The sidelobe level was −16.5 dB. Case 3: In this case, the distance between the elements in array #1 shrinks to zero; the design parameters for array #1 is ( ) This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2020.3044908, IEEE Access =  . Fig.3 (d) shows that the far-field phase of array #2 is in-phase with that of the element in every direction. The subtracted radiation pattern was achieved by subtracting the magnitude of array #2 from that of array #1. The HPBW of the subtracted radiation pattern was 38.2 , and its main lobe and sidelobe directions were 51 and 143  , respectively. The sidelobe level was −15.9 dB. It is relevant that the subtracted radiation pattern of Case 3 achieved the narrowest beamwidth of 36.9 compared to other cases.
From the above discussion, the mechanism of radiation pattern subtraction antenna that provides the narrowest beamwidth is illustrated. The proposed antenna can be considered as a phased array antenna with three elements. The optimal parameters of this antenna are amplitude and phase excitation of each element and distance between elements can be obtained from the optimization algorithm.  For the SUM BEAM, its HPBWs were narrower as the distance between the elements was increased. The nonuniform excitation (1:2:1) possessed a wider beamwidth than the uniform excitation (1:1:1), as is generally known. The rate of change of HPBW of the subtracted radiation pattern was slower than that of the SUM BEAM. All HPBWs of the subtracted radiation pattern were narrower than that of the SUM BEAM for 0.9 d   , where the 1:2:1 excitation of the subtracted radiation pattern was the widest one. The HPBWs of the 1:1:1 excitation radiation pattern subtraction and of the 1:1.5:1 excitation radiation pattern subtraction were narrower than that of the SUM BEAM.
Consider the amplitude excitation of the subtracted radiation patterns, the HPBW of the subtracted radiation pattern was almost unchanged with increasing distance () d , while the one of the SUM BEAM descreased significantly with increasing distance () d . This occurred because both the array's and the element's patterns for nonuniform excitation subtraction were much different. . In this regard, a proper separation between elements and a proper amplitude excitation had to be manually determined.
It can be seen in Fig.5 that the narrowest HPBW is 33 . This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2020.3044908, IEEE Access

IV. ANTENNA DESIGN AND EXPERIMENTAL RESULTS
This section describes the design and fabrication of the radiation pattern subtraction antenna, based on the subtraction principle presented in the previous section. CST Microwave Studio [16] was used to analyze and design the antenna. The design and analysis took into account the dielectric substrate, mutual coupling between antennas, and the type of feeding network. The halfwave-dipole element antenna structure was adopted from [17] but added a topload structure to shorten the length of the dipole. This also reduced the mutual coupling between these elements. Therefore, the proposed antenna consisted of three top-load dipoles, a microstrip feeding circuit, and two-phase shifters. All were placed on an RG4003C substrate ( ) 3.35, tan 0.0027 r  == , as can be seen in Fig. 6. The substrate was 0.508 mm thick. The copper layer was 0.035 mm thick. The metallic segments were made of copper.   6 shows three Wilkinson power combiners that provide the same input phase and a 1:1.41:1 voltage excitation (1:2:1 power excitation) to the three top-load dipoles. The HPBW of the 1:1.41:1 voltage excitation was close to that of the 1:1.5:1 voltage excitation calculated in the previous section. A phase shift of 102.8 from the feeding line II 5 ( 4.7 mm) L = was set for dipole #3, while a phase shift of 83.7 from the feeding line I 4 ( 4.2 mm) L = was set for dipole #1. The balanced feed of dipole #1 was configured to be in the opposite direction to those of the other dipoles, so it already had a 180° phase difference to the other dipoles.
The dimensions of the proposed antenna that had the HPBW of 35 was 2 5.9 6.5 cm  , which were 2 4.5 cm smaller than the dimensions of the 1:1:1 voltage excitation SUM BEAM ( 2 6.6 6.5 cm  ). To represent a conventional aperture antenna, an E-plane sectoral horn with HPBW in the E-plane of 35 [15] was used. Its dimensions (length and aperture height) were 14.2 cm, and 4 cm , respectively. Hence, the proposed antenna, which had smaller dimensions, could be used at a shorter distance from the target. Furthermore, the proposed antenna was much lighter than a horn antenna. For the proposed antenna and a switched beam antenna, their dimensions were the same. However, one of the power combiners in the proposed antenna would be substituted by a phase shifter in a switched beam antenna. The benefits from the proposed antenna were that memory circuit could be eliminated, and the sensor could be operated in real-time. Our investigation of the far-field phase distribution of the top-load dipole and the array of two top-load dipoles was conducted with CST Microwave Studio. The simulated result of the far-field phase distribution of the element (topload dipole #2) was obtained by replacing the top-load dipole #1 and #3 with two 50  load resistors. To determine the far-field phase distribution of the array antenna (top-load dipole #1 and #3), top-load dipole #2 was substituted with a 50  load resistor. In E-plane, the farfield phase distribution of the element and the array antenna  Fig. 8 shows that, at 10.36 GHz, the proposed array antenna was well-matched with the 50  input, with a reflection coefficient ( ) 11 S of −30.9 dB. The bandwidth of the antenna for 11 S less than −10 dB was around 2 GHz, from 9.5 GHz to 11.5 GHz. The slight difference in simulated and measured resonance frequencies could be from fabrication error.    9 shows the radiation patterns in E-and H-planes of the fabricated antenna. The HPBW in the E-plane was 35 , with a main beam direction of 60 . The peak of the sidelobe was −13.5 dB below the peak of the main beam in the direction of 146 . For the H-plane at 60  = , the main beam was in the direction of 90 with a HPBW of 100 .

A. RADIATION-PATTERN-SUBTRACTION ANTENNA
All of the measured results agreed very well with the simulated results of the proposed design. Any discrepancies can be attributed to the effect of the feeding structure because of slight phase shift error in fabrication.

B. ANGULAR RESOLUTION IMPROVEMENT
An improved angular resolution could be obtained more readily with a narrower beam of the subtracted radiation patterns ( 35 HPBW in E-plane) than with the broader beam of the halfwave dipole with a reflector ( 80 HPBW in E-plane). In our experiment, the scattered electric field intensity from the mango of interest (I) was measured with and without the influence of scattered wave from a nearby mango (N). The relative scattered electric field intensity, which is the ratio of the electric field measured from I and N and the measured electric field from only I, was calculated for several center-to-center separation distances () s d between I and N. In Fig. 10 (a) and (b), the main beam direction of the sensor's antenna system was configured to point to the mango of interest (I). In the experiment, the proposed radiation-pattern-subtraction antenna was rotated 30 counter-clockwise, as could be seen in Fig. 10 (a) r between the mango of interest and the sensor's antenna system was fixed at 135 mm (from the antenna