Open Access
Add to BibSonomy
Abstract
Polarimetric passive millimeter-wave (PMMW) measurements can provide additional information about scenes. We report on the development of a 94GHz Multi-Polarization Scanning Imaging Radiometer (MPSIR) for outdoor imaging applications, and the polarization characteristic analysis of experimentally acquired images. The imager angular resolution is measured to be 0.37°, and the thermal sensitivity is about 0.46K. Multi-polarization measurements are realized by rotating the detector around the observation axis direction. Several outdoor complex scenes are selected to passive image, which include the outdoor parking lot and the dormitory buildings in the woods. Various polarization parameter images are generated to analyze the polarization characteristics. Based on the multi-polarization imaging, eight more polarization parameters are created and they are all have the special distribution properties. The experimental results indicate that the multi-polarization imaging and the modified polarization parameters have the great potential for three-dimensional reconstruction, edge detection and image segmentation.
© 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Passive millimeter-wave (PMMW) remote sensing has attracted increasing interest for scientific, commercial, and military applications in recent decades [1–4]. Objects emit and reflect the millimeter-wave (MMW) radiation just as they do in the visible and infrared regimes. PMMW sensors operate by detecting natural MMW emissions and reflections of objects. There are the sky windows (low atmospheric losses) of frequencies between the resonances of oxygen and water, which are centered at 35, 94, 140 and 220GHz [5]. For the clear air with the physical temperature of 20° and the H2O density of 7.5g/m3, the attenuation is only 0.6dB/km at 94GHz [4]. Moreover, MMW sensors have the ability to detect through typical obscurants such as smoke, dust, some plastics, and most types of clothing and camouflage. The sky brightness temperature of window frequencies decreases with the decreasing atmospheric loss, which results in the large brightness temperature contrasts within a scene of interest between reflection objects and emission objects. Thus PMMW imagers generate naturally looking high-contrast radiation images and offer the superior poor-weather performance compared to visible and infrared systems. The contrast is larger by up to two orders of magnitude than that of infrared images [6]. The feature of passive operation benefits to the human security screening, where the active illumination may raise health concerns, and to military applications, where the observer may not want to emit electromagnetic signals. The main drawback of PMMW imagers is the poor angular resolution due to the longer wavelength (compared with the visible and infrared regimes) and the limited aperture size. Nevertheless, PMMW sensors have also been utilized for various applications in astronomy, physical parameter retrieval of earth and moon, environmental monitoring, ranging, aircraft landing aids, and concealed object detection [1, 7–10].
The aim of remote sensing is to obtain the physical parameters or information of interest objects. Polarimetric measurements have been experimental demonstrated to be an effective approach to obtain various information [11, 12]. In 2003, the first spaceborne polarization microwave radiometer (Windsat) has been developed and launched for quantitative earth remote sensing to retrieve the wind direction and speed of sea surface, the thickness of the sea ice and so on [13, 14]. In recent years, MMW (particularly around 90GHz) fully polarimetric radiometers have attracted increasing attentions by researchers [15–18]. The fully Stokes parameters are obtained by correlating two linear orthogonal polarization components in a correlator. The PMMW images of a person car are usually generated to discuss the polarization phenomenon. Besides, the dual-polarization radiometers are also developed to analyze the orthogonal polarization properties [19–21]. The above systems are the dual-polarization or Stokes fully polarimetric radiometers.
To investigate the multi-polarization (denotes multiple linear polarizations in this paper) phenomenon, W. -G. Kim et al. has designed a W-band quasi-optical imaging system which can generate all linear polarized radiation images [22]. The indoor objects, e.g., a metal sphere, a metal and a ceramic cup, are selected to study the indoor polarization characteristics, and the linear polarization sum imaging method has been used to improve the capability of target recognition. In our previous works, we have presented several polarization-based method to obtain the object surface orientation information and to classify the metal/dielectric materials [23–26]. Moreover, the PMMW multi-polarization characteristics of simple outdoor scenes have been discussed based on the simulations, and the simulated results have indicated that the outdoor multi-polarization imaging can also benefit to image enhancement [27, 28]. However, in the practical applications, the outdoor observation scenes might be very complex. The multiple reflections, rough surface and coverings may change the polarization properties of complex objects. To our knowledge, these problems have not been addressed experimentally and many polarization phenomenons are waited to be found and discussed.
In this paper, a multi-polarization PMMW imager is designed and developed to investigate the polarization characteristics of outdoor scenes. To ensure multi-polarization measurement and imaging efficiency simultaneously, we have combined the dual-polarization mode and the time-sharing sampling mode to construct multi-polarization imager. The design scheme and system specifications are described in Section 2. In Section 3, several typical measurement experiments have been carried out. According to the common and modified polarization parameters, we have analyzed the polarization characteristics of complex outdoor scenes. The conclusion follows in Section 4.
2. Multi-polarization single-pixel PMMW imager
The configuration and components of 94GHz MPSIR are shown in Fig. 1. The imager consists of a Cassegrain antenna, an ortho-mode transducer (OMT), two direct detection modules, a data acquisition (DAQ) unit, a three-axis scanning turntable, and two personal computers (PC, the one controls the radiometer and the other one controls the scanner). In order to easily reequip, the radiometer and the scan turntable are separately controlled by two PCs. An Ethernet cable is used to realize the physical interconnection, and two PCs intercommunicate based on the user datagram protocol (UDP). The imager can not only generate PMMW images by the scanning turntable, but also continuously collect the target signals in the fixed direction.
Fig. 1 Multi-polarization single-pixel PMMW imager. (a) Design diagram of imager. Multi-polarization measurements are realized by rotating the detector around the observation axis direction. (b) Picture of MPSIR. Two computers control radiometer and scanner. (c) Cassegrain antenna. (d) Receiver box, which consists of OMT, direct detect modules, NI data acquisition system and PC mainboard. (e) OMT and orthogonal polarization components. (f) NI DAQ unit.
2.1. Three-axis scanning turntable
The three-axis scanning turntable is constructed from an 2D imaging scanner and an observation-axis-direction rotator, as shown in Fig. 1(a). The 2D imaging scanner is composed of a two-step motor with the position accuracy of below 0.003°. The antenna and receiver box are mounted on the observation-axis-direction rotator which can be rotated with the position accuracy of below 0.1°. In order to obtain the multi-polarization images, the observation-axis-direction rotator is designed to rotate a given linear polarization angles separately and then fixed during the 2D imaging. Two servomotors are used to change the elevation and azimuth angles of the load, which is a raster scan across a given field of view (FOV). The turntable has an onboard FPGA to feedback the real-time position.
The scanning imaging time is a significant issue of the single-pixel PMMW imager. The motion control of 2D imaging scanner can affect the imaging performance. According to the motion type of horizontal FOV, we have developed two scanning modes to generate an image pixel by pixel. (1) Step Scanning: the horizontal servomotor stops in each given sample position to wait for DAQ unit to collect signals, then moves to the next sample position to repeat the above operation. (2) Continuous Scanning: the horizontal servomotor moves with a constant speed. The speed is calculated based on the FOV and the integration time. It can realize that 100% of scanning time can be used for data collection. Therefore, the former mode needs to spend much more time imaging than the latter one. However, the latter mode can cause the motion blur effect. To relieve this effect, we can reduce the motion speed and collect signals in the given discrete positions instead of capturing data continuously. So the motion speed can be manual modified in the control software.
For example, assuming that the FOV, the integration time and the imaging step (sampling resolution) are 20° × 10°, 5ms and 0.1°, respectively. By converting, the image has 200 × 100 pixels. If we use the 100% of scanning time to sample data, the horizontal motion speed is 0.1°/5ms= 20°/s. Considering the acceleration and deceleration time (both 1.5s), the each horizontal FOV will spend about 2s+1s=3s. So the total imaging time is 100 × 4s=400s. However, this speed is so fast for the motor and the motion blur will be very obvious. In this case, if we reduce the motion speed to 5°/s, the acceleration and deceleration time will both become to about 0.3s, and the total imaging time will be about 460s, which only increases 15%. Because the 5ms integration is completed in the angle range of 0.1°/4 = 0.025°, the motion blur will be relieve apparently. In summary, a proper motion speed should be weighed to obtain the less imaging time and the high quality image.
2.2. Radiometer receiver
The schematic diagram of MPSIR imager is shown in Fig. 2. The multi-polarization imaging is realized by rotating the observation-axis-direction rotator. To ensure the simultaneity of orthogonal polarization data, a dual-polarization receiver is constructed to acquire the dual-polarization radiation signals simultaneously.
Fig. 2 Schematic diagram of the MPSIR imager with two receiver channels. V and H denote two orthogonal-polarization components.
The imager starts with a 94GHz Cassegrain antenna, which is used to collect and focus the naturely emited MMW radiation onto a circular feed-horn antenna and into a circular waveguide. The aperture diameter of main reflector is 0.6096m and the frequency bandwidth is 4GHz. In the test frequency of 92∼96GHz, the antenna gain changes from 50.2dB to 50.95dB, the 3dB or half-power beam width (HPBW) changes from 0.35° to 0.39°, the sidelobes level is from −22.537dB to −18.33dB, and the voltage standing wave ratio (VSWR) is from 1.23:1 to 1.43:1.
An OMT is utilized to separate two orthogonal polarization states of the incident MMW radiation signal into two front-end modules (direct detection modules, DDMs), as shown Fig. 1(e), which are called the V (vertical) and H (horizontal) polarization channels based on the orientation of the waveguide relative to the flange. The OMT is the Millitech product (OMT-10-094RR) with the insertion loss of 0.55∼0.59dB and the VSWR of 1.2:1. The MMW signal from the OMT is coupled to the 94GHz low-noise amplifier (LNA), through the square-law detector (SLD), and then the low-frequency amplifier (LFA). Two DDMs are supplied by DC of +5V@42mA, and the video outputs are 1.17V (V channel) and 1.28 (H channel) for the +25°C blackbody material. The noise figures are 3.65dB (V channel) and 3.48dB (H channel), which mean that the noise temperatures are 382.0K and 356.2K.
The output voltages are acquired by the DAQ unit, as shown Fig. 1(f), which is the National Instruments product (NI 9223 BNC) with the maximum sampling rate of 100MHz. The graphic user interface (GUI) is in the radiometer PC and is programmed with C++ codes. The GUI software has two main functions: (1) Communicating with the turntable to scan and capture data; (2) and displaying the dual-polarization images at the same time. To simplify the wire arrangement, we have manufactured a receiver box to pack an OMT, two DDMs, DAQ unit and PC mainboard together.
2.3. System performance
Thermal temperature sensitivity and angle resolution are two important performance parameters of radiometer systems. For the passive radiometric measurement or imaging, illumination reflection and atmospheric attenuation are the nature phenomenons. In the performance tests, the corresponding analysis and design should be considered.
The noise performance (i.e., thermal temperature sensitivity) of MPSIR is described by the minimum detectable temperature or noise-equivalent temperature difference (NETD), which can be measured as by using the following equation [29]:
where Nrms is the RMS noise of measured signal, Rslope is the system responsivity (i.e., the slope of the calibration curve in the range of system response). Nrms is measured by capturing the signals of an absorbing material for an extended period time. Considering the integration time of 5ms and the blackbody temperature of 303K, the measured Nrms are 1.327mV (V channel) and 1.326mV (V channel). Rslope is measured by observing a hot blackbody (the absorbing material at ambient temperature) and a cold blackbody (the absorbing material at liquid nitrogen temperature), then calculating the slope of linear calibration curve.Therefore, the linearity of MPSIR is measured firstly and then the calibration curve can be used to calculate the NETD. A calibration material was placed in front of the Cassegrain antenna and controlled at a variety of physical temperatures (300K, 308K, 318K, 328K, 338K and 348K). To obtain the cold brightness temperature, an calibration material was dipped in the liquid nitrogen (∼77K). The calibration material is a blackbody absorbing material with the emissivity of 0.999 and the size of 0.8m×0.8m. Because the reflectivity and transmissivity are both lower than 0.001, the contributions of reflection and transmission have been ignored in the test. The calibration material is closely place in the front of antenna, so the atmospheric attenuation has been also neglected. The measurement results of the linearity tests are shown in Fig. 3(a). The V channel has the responsivity Rslope of 2.837mV/K and the offset voltage of 3.9329V. The H channel has the responsivity Rslope of 2.921mV/K and the offset voltage of 3.5803V. According to Eq. 1, NETDs of V and H channels are 0.468K and 0.454K, respectively.
Fig. 3 Measurement results of MPSIR performance. (a) Radiometer response curve to the absorbing material at a variety of physical temperatures. (b) Calculated and measured PSF of the radiometer from the point source measurments.
As is known, the theoretical NETD can be calculated by [30]
where TA and TR are the antenna brightness temperature and the receiver noise temperature, respectively, B is the radiometer bandwidth, τ is the integration time, and G and ∆G are the system gain and its variation. According to the parameters of above measurements, TA=303K, TR=382.0K (V channel), TR=356.2K (H channel), B=4GHz, τ=5ms, and if we assume there is no gain variation, we have NETD=0.153K (V channel) and NETD=0.147K (H channel). This difference mainly results from the gain variation effect. Consequently, the gain variation can be calculated as 6.4558×10−4 (V channel) and 6.5140 × 10−4 (H channel). It can indicates that the MPSIR has a good performance of gain variation [30].The angle resolution is another important parameter to describe the imaging performance of PMMW imager, which is measured by the method described in [29] using a point target with a high MMW radiation. A noise source with a high MMW radiation (about 6000K) is placed in the position with the distance of 50m. In order to reduce the background radiation, a metal plate is placed at 45° to reflect the cold radiometric sky. For the noise source, the cold sky background radiation is low enough. The measured PSF is normalized, so the atmospheric attenuation between the antenna and the point target do not influence the PSF curve within 2° nearby the main lobe. For a linear shift-invariant system, the image of the point target is the point spread function PSF of the system. The diffraction limited PSF can be theoretically calculated based on the Cassegrain dish diameter of 0.6096m [31]. The point target has been placed in the far-field and the measured results are shown in Fig. 3(b). It can indicate that MPSIR is approximately diffraction limited with the angle resolution of ~ 0.37°.
In summary, the instrument main specifications of MPSIR are listed in Table 1.
Table 1. Instrument main specifications of MPSIR
2.4. System outputs and polarization parameters
All the linear polarization brightness temperatures can be obtained by rotating the antenna and receiver box around the observation-axis-direction. In this paper, the reference plane of linear polarization angle is the ground. For an azimuthal symmetrical and flat “opaque” surface, the arbitrary linear polarization brightness temperature can be described by [23]
where α represents the linear polarization angle with the range from −180° to 180°. α = 0° denotes the horizontal polarization (i.e., TB0◦ = TBh) and α = 90° denotes the vertical polarization (i.e., TB90◦ = TBv). θ is the incident angle with respect to the flat surface, eα is the MMW emissivity of object surface, Tobj is the physical temperature of object, and TBiα is the ambient brightness temperature incident on the object at the reflection angle. The first term in Eq. (3) represents the surface self-emission and the second term is the surface reflection portion.There are four information parameters to describe the polarization state of natural emission, which are expressed by the Stokes vector of brightness temperature and conventionally defined as [15]
where TBv, TBh, TB45◦, TB−45◦ (i.e. TB135◦), TBcl and TBcr are the brightness temperatures of vertical, horizontal, linear 45°, linear −45°, circular left cl and circular right cr polarization, respectively. TI represents the radiance intensity received by radiometers. According to the theoretical calculations and experimental measurements from many references [15–18, 32–35], TV signals of most natural and man-made objects have been demonstrated low level. The level of TV is mainly related to the object internal particle distribution and the layered surface structure [34, 35]. TV polarization signal increases at the large incident angle and many object edges generally have the large TV value (several even dozens of Kelvin) [15, 16]. The reported MPSIR can only measure the linear polarization signals, so TV will not be discussed hereafter in this paper.In addition, we define other orthogonal polarization difference parameters as written by
where TBα, (α = 15°, 30°, 60°, 75°, 105°, 120°, 150°, 165°), has the same definition with the Eq. (3). Moreover, several polarization feature parameters are usually defined to describe the polarization states, which are written as where PP, DOLP and AOP are the polarization percentage, degree of linear polarization and angle of polarization. LPR is the linear polarization ratio, which is defined for the metal/dielectric material classification in our previous works [23, 25]. Additionally, we define a modified DOLP (mDOLP), which combines several other polarization differences, and three modified LPR (mLPR), which have other forms. where the denominator of mDOLP becomes (6·TI) due to six polarization differences, E(TI) is the mean value of all pixels of TI image, M(TI) is the median value of all pixels of TI image.3. Experiment results, analysis and discussion
At the latter half of 2017, MPSIR was assembled successfully, and first measurements were conducted for the various natural and man-made objects in the outdoor complex scenes. Several selected experiment topics are presented in the following subsections. The polarization state in this paper is referenced to the spherical elevation over azimuth coordinate system, which means that the horizontal ground is the reference plane.
3.1. Outdoor parking lot
The overlook imaging is the common outdoor application, and the cars are the typical objects of outdoor scenes. Figure 4 shows the visible image of overlook imaging scenes. The center angle of elevation is −28°, which means that imaging is taken at the centet angle of 28° below the horizon. The imaging distance is 33m. Figure 5 shows the imaging experimental results of twelve linear polarizations. All the gray-scale images have been inverted, i.e., the black denotes the high brightness temperature and the white denotes the low brightness temperature. The cars have the low brightness temperature due to reflecting the cold sky. It is seen that there seems no contrast among the twelve images. Because many pixel positions have the multiple reflection phenomenons and then the polarization contrast decreases even disappears. In fact, there are small contrasts in the window areas but they are not obvious in the original brightness temperature images. Therefore, polarization difference images are further calculated.
To analyze the polarization characteristics, the fifteen polarization feature parameters have been calculated, as shown in Fig. 6. The experimental results show that there are many interesting facts in the polarization feature images, which imply the physical parameters or information of observed objects.
- TQ, TU, TW, TX, TY, and TZ images show that the car windows have the apparent polarization contrasts. The window orientation direction determines the contrast level. Selecting the window area #A and #B as the examples. The normal vector of area #A directs to the upper right side, while that of area #B directs the upper left side. TB45◦ is nearly the horizontal polarized brightness temperature for area #A but is nearly the vertical polarized brightness temperature for area #B. Therefore, there is the biggest difference in TU for area #A and #B (but the one is negative, the other one is positive).
- AOP is the angle of the major axis of polarization ellipse with respect to x axis. In other words, it is the orientation angle of normal vector of the corresponding pixel position (only for the polarized radiation surfaces with one reflection of non-polarized radiation object). For the non-polarized radiation surfaces (e.g., sky, rough shrub, multi-reflection cavity and so on), the computed AOP can not describe the surface orientation property. Consequently, there are many noise areas which have non-stable AOP values, e.g., area #C and all car metal roof areas reflecting the sky. Therefore, AOP distribution can also distinguish between the polarized and non-polarized surfaces/objects.
- PP, DOLP and mDOLP represent the degree of polarization difference. PP only considers the difference between horizontal and vertical polarization, while mDOLP considers several other orthogonal polarization differences. The corresponding images and statistical curves in Fig. 6 (the third row) show that mDOLP has more obvious polarization feature than PP and DOLP. The noise level of mDOLP image is low due to the average computation of multiple polarization differences.
- In LPR image, the range of color bar has been limited to [0.5, 2] to display clearly. The ground has higher LPR values than the car, especially the ground area #D with the elevation angle of about −25°. According to the Ref. [25], the LPR of dielectric surface at the near Brewster angle has high value, because the vertical polarization brightness temperature at this elevation angle is very close to the physical temperature T0. Theoretically, the LPR of all material surfaces is equal to or higher than 1. But the measured LPR might lower than 1, because the measured brightness temperatures generally have noises and so TBh may be higher than TBv for the non-polarized objects (e.g., sky, rough shrubs). From the LPR image, we might also classify the metal/dielectric materials using the method in the Ref. [25]. However, the selection criterion of classification threshold should be modified, because the brick ground is rough surface and its LPRs decrease [26]. mLPR1, mLPR2 and mLPR3 are the modified LPR, which have the similar form of LPR and only replace Tobj by other parameters. The new parameters are generally lower than Tobj. Based on the modified operations, the outlines of cars are automatically displayed. These results indicate that the modified LPR parameters are more sensitive to the surfaces with the large incident angle (i.e. the surface normal become nearly orthogonal to viewing direction as in the sides of a cylinder). Therefore, the modified LPR parameters might be used for occluding edge detection and image segmentation. Subjectively, mLPR2 has better segmentation performance than mLPR1 and mLPR3. The modified LPR used for image segmentation will be deeply investigated in future work.
3.2. Dormitory buildings with woods background
The dormitory buildings in the woods are selected to be a typical complex scene because this scene has metal objects, objects with coverings, multiple reflection objects, sky and so on. Figure 7 shows the visible image of complex scenes. The elevation angle is from −17° to 7°, which means that imaging is taken around the horizontal direction. The imaging distance of rear dormitory is about 120m and that of car shed is about 55m. Figure 8 shows the imaging experimental results of twelve linear polarizations. All the gray-scale images have been inverted, i.e., the black denotes the high brightness temperature and the white denotes the low brightness temperature. To display clearly, the range of color bar has been limited to [240K, 300K]. (1) Area #G, #J and #K have the low brightness temperature due to the reflection of cold sky. Note that area #J actually has different brightness temperatures for twelve images. This phenomenon will be displayed clearly in the polarization parameter images, which will be explained in the following paragraphs. (2) The brightness temperatures of building roof (area #H) have noticeable differences for twelve images. (3) However, many other areas seem have no difference among the twelve images, which is caused by the polarization decreasing of multiple reflections. (4) The imaging experiments were carried out at dusk and the sky physical temperature was obviously reducing during the measurements. Six pairs of orthogonal polarization images were obtained orderly from 0°/90°, 45°/135°, 30°/120°, 60°/150°, 15°/105°, 75°/165°. Thus, the sky brightness temperature of six pairss of images has orderly declined over time. The trees in front of building also have this phenomenon.
Fig. 7 Visible images of horizontal imaging scene. (a) Imaging scene. The red block is the imaging area. (b) Imaging area.
Fig. 8 Multi-polarization PMMW images of dormitory buildings in horizontal direction. To display clearly, the color bar has been manually scaled to 240K ∼ 300K. The real brightness temperature range is about 192K ∼ 306K
As shown in Fig. 9, the fifteen polarization feature parameters also have been calculated to analyze the polarization characteristics.
- TQ, TU, TW, TX, TY, and TZ images show that the building roofs (area #H) with the oblique incidence have apparent polarization contrasts. However, for the roof covered by trees (area #I), the polarization difference reduces even disappears, because the trees are almost the non-polarized MMW radiation objects. The roof normal vector determines the difference level. The orientation direction of normal vector influences the polarization change law, and the elevation direction of normal vector denotes the incident angel. For area #H, the orientation direction of normal vector is nearly vertical. So 45° and 135° polarizations are nearly symmetrical and TU values are close to zero. For area #G and #K, two objects are both metal material and reflect the non-polarized cold sky. So six orthogonal polarization differences are all near-zero. As noted in above paragraph, area #J is a metal car shed with the coverings of sunscreen fabrics, which has an arc-shaped roof. The thermal radiation of shed roof can be described by a three-layer model (including metal, fabrics and sky) and it is a typical polarized radiation surface [30]. The highlight area positions of six images are distinct, which indicates that linear polarization difference is sensitive to object shape. If we fuse all highlight areas of six images, the overall shape will be reconstructed. For the building sides and the trees, the MMW radiations are nearly non-polarized. The former one is caused by the multiple reflections among the object surfaces. The latter one is due to an amount of the random scattering and transmission between leaves on trees.
- As is similar to the first experiments in section 3.1, AOP image also shows the discrimination ability for the polarized and non-polarized areas. Area #H and J are polarized surfaces. These results are agree with the above six polarization difference images. The other areas with noise fluctuation denote approximately non-polarized surfaces.
- PP, DOLP and mDOLP represent the degree of polarization difference. The polarized surfaces have the non-zero polarization degree. As is similar to the first experiments in section 3.1, these three images also indicate that mDOLP has more obvious polarization feature than PP and DOLP. mDOLP image has low noises due to the average computation of multiple polarization differences. The outlines of area #H and #J is very clear in mDOLP image. The peak areas in the sub-figure of statistic curves represent area #H and #J, which are more obvious than the first experiments in section 3.1.
- According to the LPR application limits, LPR in this imaging case is unsuited to classify the metal/dielectric materials. Because the incident angles and roughness of object surfaces are not in the typical application scope, or the multiple reflections of many dielectrics can diminish polarization (e.g., the concrete ground in front of buildings, and the side surfaces of buildings). Area #L has larger LPR than other areas, because the incident angle is more close to the application scope (60° ∼ 70°) and the concrete surface is relatively smooth. Furthermore, the three modified LPR parameters also highlight the object occluding edges. As the scattering and transmission coverings, the trees interference the polarization property. The local detail features are not good but the overall outline basically agrees with the real one. Therefore, it is necessary to further investigate the application scenes and conditions if we use the modified LPR for occluding edge detection and image segmentation.
4. Conclusion
In this paper, we report a multi-polarization passive millimeter-wave imager MPSIR for remote sensing, and analyze the multi-polarization characteristics of outdoor complex scenes. The three-axis scanning turntable is constructed to generate 2D scan imaging of single-pixel radiometer, and to obtain multiple linear polarized images separately. The orthogonal dual-polarization receivers are used to acquire the dual-polarization radiation signals simultaneously. It not only ensures the simultaneity of dual-polarization data, bu also reduces the multi-polarization imaging times. The MPSIR system has the aperture diameter of 0.6096m, the angular resolution of 0.37°, and the thermal brightness temperature sensitivity of about 0.46K (5ms integration time). A typical imaging with 200×100 pixels spends about 7min.
Two typical outddor scenes were selected to conduct the multi-polarization imaging experiments. The experimental results indicate that: (1) The multiple reflections, rough surface and volume scattering coverings can decrease the polarization degree or feature; (2) Several additional orthogonal polarization differences can help for the reconstruction of overall shape; (3) AOP image displays a random noise distribution feature for the non-polarized surfaces/objects, and has the deterministic angles for the polarized surfaces/objects. AOP can be not only utilized to retrieve the orientation angle, but also used to distinguish the polarized/non-polarized surfaces/objects; (4) The modified DOLP can obtain more stable distribution of polarization degree, which is help for the detection of man-made smooth dielectrics; (5) The modified LPR is sensitive to the surfaces with big incident angles, which could be applied to the occluding edge detection and image segmentation. The corresponding theories, methods and experiments will be further investigated in future work.
In summary, the multi-polarization PMMW imaging has an enormous potential for three-dimensional reconstruction, material classification, edge detection and image segmentation. In fact, several researches have already been reported in recent years. There are also many detail problem to solve, such as the quantitative research of the depolarization effect of multiple reflection and covering.
Funding
Fundamental Research Funds for the Central Universities (HUST: 2017JYCXJJ036); China International Joint Research Fund (2015B01008).
Acknowledgments
The authors would like to thank Miss Jingjing Liu for the spiritual encouragement and support, Mr. Pengtao Tian for the system technical discussion, and Mr. Yuanbo Sun for the graphical user interface discussion.
References and links
1. L. Yujiri, M. Shoucri, and P. Moffa, “Passive millimeter-wave imaging,” IEEE Microw. Mag. 4(3), 39–50 (2003). [CrossRef]
2. N. A. Salmon, “Outdoor passive millimeter-wave imaging: Phenomenology and scene simulation,” IEEE Trans. Antennas Propag. 66(2) 897–908 (2018). [CrossRef]
3. G.-P. Hu, K. L. Chan, Y.-C. Zheng, K. T. Tsang, and A.-A. Xu, “Comparison and evaluation of the change microwave radiometer data based on theoretical computation of brightness temperatures at the apollo 15 and 17 sites,” Icarus 294, 72–80 (2017). [CrossRef]
4. R. Appleby, “Passive millimetre-wave imaging and how it differs from terahertz imaging,” Philos. Trans. R. Soc. A-Math. Phys. Eng. Sci. 362, 379–393 (2004). [CrossRef]
5. J. R. Pardo, J. Cernicharo, and E. Serabyn, “Atmospheric transmission at microwaves (ATM): an improved model for millimeter/submillimeter applications,” IEEE Trans. Antennas Propag. 49(12), 1683–1694 (2001). [CrossRef]
6. D. G. Gleed, R. Appleby, N. A. Salmon, S. Price, G. N. Sinclair, R. N. Anderton, J. R. Borill, and M. R. M. Wasley, “Operational issues of passive millimeter wave imaging systems,” Proc. SPIE3064, 23–33 (1997). [CrossRef]
7. S. E. Clark, J. A. Lovberg, C. A. Martin, and V. Kolinko, “Passive millimeter-wave imaging for airborne and security applications,” Proc. SPIE5077, 16–21 (2003). [CrossRef]
8. T. Luthi and C. Matzler, “Stereoscopic passive millimeter-wave imaging and ranging,” IEEE Trans. Microw. Theory Tech. 53(8), 2594–2599 (2005). [CrossRef]
9. M. R. Fetterman, J. Grata, G. Jubic, W. L. Kiser, and A. Visnansky, “Simulation, acquisition and analysis of passive millimeter-wave images in remote sensing applications,” Opt. Express 16(25), 20503–20515 (2008). [CrossRef] [PubMed]
10. C. Zhang, H. Liu, J. Wu, S. Zhang, J. Yan, L. Niu, W. Sun, and H. Li, “Imaging Analysis and First Results of the Geostationary Interferometric Microwave Sounder Demonstrator,” IEEE Trans. Geosci. Remote Sensing 53(1), 207–218 (2015). [CrossRef]
11. J. S. Tyo, D. L. Goldstein, D. B. Chenault, and J. A. Shaw, “Review of passive imaging polarimetry for remote sensing applications,” Appl. Optics 45(22), 5453–5469 (2006). [CrossRef]
12. F. A. Sadjadi, “Passive three-dimensional imaging using polarimetric diversity,” Opt. Lett. 32(3), 229–231 (2007). [CrossRef] [PubMed]
13. P. W. Gaiser, K. M. St Germain, E. M. Twarog, G. A. Poe, W. Purdy, D. Richardson, W. Grossman, W. L. Jones, D. Spencer, G. Golba, J. Cleveland, L. Choy, R. M. Bevilacqua, and P. S. Chang, “The windsat spaceborne polarimetric microwave radiometer: Sensor description and early orbit performance,” IEEE Trans. Geosci. Remote Sensing 42(11), 2347–2361 (2004). [CrossRef]
14. P. S. Narvekar, G. Heygster, T. J. Jackson, R. Bindlish, G. Macelloni, and J. Notholt, “Passive polarimetric microwave signatures observed over antarctica,” IEEE Trans. Geosci. Remote Sensing 48(3), 1059–1075 (2010). [CrossRef]
15. H. Suess and M. Soellner, “Fully polarimetric measurements of brightness temperature distributions with a quasi-optical radiometer system at 90 GHz,” IEEE Trans. Geosci. Remote Sensing 43(5), 1170–1179 (2005). [CrossRef]
16. A. Duric, A. Magun, A. Murk, C. Matzler, and N. Kampfer, “The fully polarimetric imaging radiometer SPIRA at 91 GHz,” IEEE Trans. Geosci. Remote Sensing 46(8), 2323–2336 (2008). [CrossRef]
17. M. Peichl, S. Dill, M. Jirousek, J. W. Anthony, and H. Süß, “Fully-polarimetric passive MMW imaging systems for security applications,” Proc. SPIE 7837, 78370C (2010). [CrossRef]
18. M. Peichl, S. Dill, and D. Rudolf, “Investigation of fully-polarimetric signatures from targets with some relevance to security applications,” Proc. SPIE 8022, 80220E (2011). [CrossRef]
19. J. P. Wilson, C. A. Schuetz, T. E. Dillon, P. Yao, C. E. Harrity, and D. W. Prather, “Passive 77 GHz millimeter-wave sensor based on optical upconversion,” Appl. Optics 51(18), 4157–4167 (2012). [CrossRef]
20. J. P. Wilson, C. A. Schuetz, C. E. Harrity, S. Kozacik, D. L. Eng, and D. W. Prather, “Measured comparison of contrast and crossover periods for passive millimeter-wave polarimetric imagery,” Opt. Express 21(10), 12899–12907 (2013). [CrossRef] [PubMed]
21. S. Liao, N. Gopalsami, T. W. Elmer, E. R. Koehl, A. Heifetz, K. Avers, and A. C. Raptis, “Passive millimeter-wave dual-polarization imagers,” IEEE Trans. Instrum. Meas. 61(7), 2042–2050 (2012). [CrossRef]
22. W.-G. Kim, N.-W. Moon, H.-K. Kim, and Y.-H. Kim, “Linear polarization sum imaging in passive millimeter-wave imaging system for target recognition,” Prog. Electromagn. Res. 136, 175–193 (2013). [CrossRef]
23. Y. Cheng, F. Hu, L. Gui, L. Wu, and L. Lang, “Polarization-based method for object surface orientation information in passive millimeter-wave imaging,” IEEE Photonics J . 8(1), 5500112 (2016). [CrossRef]
24. Y. Cheng, B. Qi, S. Liu, F. Hu, L. Gui, and X. Peng, “Polarization ratio property and material classification method in passive millimeter wave polarimetric imaging,” Proc. SPIE 9993, 9993K (2016).
25. F. Hu, Y. Cheng, L. Gui, L. Wu, X. Zhang, X. Peng, and J. Su, “Polarization-based material classification technique using passive millimeter-wave polarimetric imagery,” Appl. Optics 55(31), 8690–8697 (2016). [CrossRef]
26. F. Hu, X. Zhang, Y. Cheng, Y. Xiao, and M. Song, “Influence of surface roughness on polarization property in passive millimeter-wave imaging,” IEICE Electron. Express 14(21), 20171005 (2017). [CrossRef]
27. Y. Cheng, F. Hu, L. Gui, J. Su, B. Qi, S. Liu, and M. Huang, “Linear polarisation property and fusion method for target recognition in passive millimetre-wave polarimetric imaging,” Electron. Lett. 52(14), 1221–1223 (2016). [CrossRef]
28. Y. Cheng, F. Hu, L. Gui, Y. Hu, and Z. Han, “Passive millimeter-wave polarization characteristics of several common structures,” Proc. SPIE 10407, 1040711 (2017).
29. R. G. Driggers, C. M. Webb, S. J. Pruchnic, C. E. Halford, and E. Burroughs, “Laboratory measurement of sampled infrared imaging system performance,” Opt. Eng. 38(5), 852–862 (1999). [CrossRef]
30. F. T. Ulaby, R. K. Moore, and A. K. Fung, Microwave Remote Sensing: Microwave remote sensing fundamentals and radiometry, (Addison-Wesley Publishing Company, 1981), Vol. 1.
31. B. Qi, L. Lang, Y. Cheng, S. Liu, F. Hu, X. He, P. Deng, and L. Gui, “Passive millimeter-wave scene imaging simulation based on fast ray-tracing,” IEEE International Geoscience and Remote Sensing Symposium (IGARSS) (2016), 2642–2645.
32. N. A. Salmon, R. Appleby, and P. Coward, “Polarimetric passive millimetre wave imaging,” Proc. SPIE 4373, 82–85 (2001). [CrossRef]
33. N. A. Salmon, R. Appleby, and P. Coward, “Polarimetric passive millimetre wave imaging,” Proc. SPIE4373, 82–85 (2001).
34. L. Tsang, “Polarimetric passive microwave remote sensing of random discrete scatterers and rough surfaces,” J. Electromagn. Waves Appl. 5(1), 41–57 (1991).
35. L. Tsang, P. Xu, and K. S. Chen, “Third and fourth stokes parameters in polarimetric passive microwave remote sensing of rough surfaces over layered media,” Microw. Opt. Technol. Lett. 50(12), 3063–3069 (2008). [CrossRef]
