University of Birmingham Assessing the Raspberry Pi as a low-cost alternative for acquisition of near infrared hemispherical digital imagery Kirby, Jennifer; Chapman, Lee; Chapman, Victoria DOI: 10.1016/j.agrformet.2018.05.004 License: Creative Commons: Attribution-NonCommercial-NoDerivs (CC BY-NC-ND) Document Version Peer reviewed version Citation for published version (Harvard): Kirby, J, Chapman, L & Chapman, V 2018, 'Assessing the Raspberry Pi as a low-cost alternative for acquisition of near infrared hemispherical digital imagery', Agricultural and Forest Meteorology, vol. 259, pp. 232-239. https://doi.org/10.1016/j.agrformet.2018.05.004 Link to publication on Research at Birmingham portal Publisher Rights Statement: Published in Agricultural and Forest Meteorology on 15/05/2018 DOI: https://doi.org/10.1016/j.agrformet.2018.05.004 General rights Unless a licence is specified above, all rights (including copyright and moral rights) in this document are retained by the authors and/or the copyright holders. The express permission of the copyright holder must be obtained for any use of this material other than for purposes permitted by law. • Users may freely distribute the URL that is used to identify this publication. • Users may download and/or print one copy of the publication from the University of Birmingham research portal for the purpose of private study or non-commercial research. • User may use extracts from the document in line with the concept of ‘fair dealing’ under the Copyright, Designs and Patents Act 1988 (?) • Users may not further distribute the material nor use it for the purposes of commercial gain. Where a licence is displayed above, please note the terms and conditions of the licence govern your use of this document. When citing, please reference the published version. Take down policy While the University of Birmingham exercises care and attention in making items available there are rare occasions when an item has been uploaded in error or has been deemed to be commercially or otherwise sensitive. If you believe that this is the case for this document, please contact UBIRA@lists.bham.ac.uk providing details and we will remove access to the work immediately and investigate. Download date: 06. Apr. 2021 https://doi.org/10.1016/j.agrformet.2018.05.004 https://research.birmingham.ac.uk/portal/en/persons/jennifer-kirby(fdbba289-be34-41a7-b0c6-b392132fc860).html https://research.birmingham.ac.uk/portal/en/persons/lee-chapman(1b1714b2-a5d1-4dc9-ba54-07b902fa400e).html https://research.birmingham.ac.uk/portal/en/persons/vicky-chapman(26fd4cee-b1f3-4acf-9526-57895a7cadad).html https://research.birmingham.ac.uk/portal/en/publications/assessing-the-raspberry-pi-as-a-lowcost-alternative-for-acquisition-of-near-infrared-hemispherical-digital-imagery(a8ffe21b-d353-4579-a1c1-9f03bd6a55f4).html https://research.birmingham.ac.uk/portal/en/publications/assessing-the-raspberry-pi-as-a-lowcost-alternative-for-acquisition-of-near-infrared-hemispherical-digital-imagery(a8ffe21b-d353-4579-a1c1-9f03bd6a55f4).html https://research.birmingham.ac.uk/portal/en/journals/agricultural-and-forest-meteorology(2fc6d970-b9d6-44cb-879e-0e70cf794cd0)/publications.html https://doi.org/10.1016/j.agrformet.2018.05.004 https://research.birmingham.ac.uk/portal/en/publications/assessing-the-raspberry-pi-as-a-lowcost-alternative-for-acquisition-of-near-infrared-hemispherical-digital-imagery(a8ffe21b-d353-4579-a1c1-9f03bd6a55f4).html Title: Assessing the Raspberry Pi as a low-cost alternative for acquisition of near 1 infrared hemispherical digital imagery 2 3 Authors & Affiliations: 4 Jennifer Kirby 5 University of Birmingham, College of Geography, Earth and Environmental Science, 6 Edgbaston, Birmingham, B15 2TT, JXK067@bham.ac.uk 7 Professor Lee Chapman (Principle Correspondence) 8 University of Birmingham, College of Geography, Earth and Environmental Science, 9 Edgbaston, Birmingham, B15 2TT, l.chapmam@bham.ac.uk 10 Dr Victoria Chapman 11 Met Office Surface Transport Team, Birmingham Centre for Railway Research and 12 Education, Gisbert Kapp Building, University of Birmingham, Edgbaston, 13 Birmingham, B15 2TT, victoria.chapman@metoffice.gov.uk 14 15 'Declarations of interest: none' 16 17 Abstract 18 Hemispherical imagery is used in many different sub-fields of climatology to calculate 19 local radiation budgets via sky-view factor analysis. For example, in forested 20 environments, hemispherical imagery can be used to assess the leaf canopy, (i.e. 21 leaf area / gap fraction) as well as the radiation below the canopy structure. Nikon 22 Coolpix cameras equipped with an FC-E8 fisheye lens have become a standard 23 device used in hemispherical imagery analysis however as the camera is no longer 24 manufactured, a new approach needs to be investigated, not least to take advantage 25 of the rapid development in digital photography over the last decade. This paper 26 conducts a comparison between a Nikon Coolpix camera and a cheaper alternative, 27 the Raspberry Pi NoIR camera, to assess its suitability as a viable alternative for 28 future research. The results are promising with low levels of distortion, comparable to 29 the Nikon. Resultant sky-view factor analyses also yield promising results, but 30 challenges remain to overcome small differences in the field of view as well as the 31 present availability of bespoke fittings. 32 Key words: Hemispherical fisheye, Near infra-red, Raspberry Pi, Sensors 33 1. Introduction 34 Hemispherical imagery is commonly used to assist in the assessment of radiation 35 budgets. Examples of use include below tree canopies, in urban areas or within 36 riverine environments (Hall et al., 2017; Liu et al., 2015; Chapman, 2007; Chapman 37 et al; 2007; Bréda, 2003; Ringold et al., 2003; Watson and Johnson, 1987). Imagery 38 is usually obtained using a camera equipped with a fisheye lens (Figure 1a) which 39 allows the camera to take an approx. 180˚ hemispherical image (Liu et al., 2015; 40 Chianucci et al., 2015). These images are then processed to analyse the amount of 41 Comment [JK1]: Removed resolution comment visible sky shown in the image (known as the sky-view factor). This can then be used 42 in forestry research to quantify the health of a tree and to compare differences 43 between tree canopies (Schwalbe et al., 2009; Leblanc et al., 2005 Jonckheere et al. 44 2004). 45 (a) (b) (c) Figure 1 (a) FC-E8 Fisheye lens attached to a Coolpix camera Source: Reproduced 46 with permission from Chapman et al. (2007), copyright © 2007 IEEE, (b) First2Savv 47 1850 fisheye camera attached to a Samsung Galaxy S5 Neo; (c) Perspex Dome 48 used to measure distortion. 49 The use of fisheye imagery for this application can be dated back to the early work of 50 Anderson (1964), but it was the advent of digital photography which saw the 51 approach become widely adopted. Following a number of scoping studies, which 52 successfully compared results obtained from film cameras to the new generation of 53 digital cameras (Englund et al. 2000; Frazer et al. 2001; Hale and Edwards, 2002), 54 the new technology quickly became adopted by the scientific community. However, 55 following the successful transition to mass digital photography, studies for the past 56 two decades have become very reliant on the early digital cameras produced by 57 Nikon (Error! Reference source not found.) such as the Coolpix 950 or 4500 58 (Chianucci et al. 2016; Lang et al., 2010; Chapman, 2007; Zhang et al., 2005; Baret 59 and Agroparc, 2004; Ishida, 2004). Indeed, whilst research into hemispherical 60 imagery has also been conducted using alternative cameras and equipment (Table 61 2Error! Reference source not found.), the Nikon Coolpix range equipped with the 62 FC-E8 fisheye lens undoubtedly remains the most popular choice in research to 63 date. 64 Seasonal Changes in Canopy Structure Liu et al., 2015 Used a Nikon Coolpix 4500 camera at sunset / sunrise to capture hemispherical images of tree canopies in order to investigate seasonal changes of tree canopies. Comparing Nikon Coolpix to film cameras and Leaf canopy analysers Homolová et al. 2007 Used a Nikon Coolpix 8700 to compare canopy analysers to hemispherical imagery. Garrigues, et al., 2008 Compares Nikon Coolpix 990 with LAI-2000 and AccuPAR. Frazer et al., 2001 Compared a Nikon 950 to a film camera and highlighted the potential for blurred edges and colour distortion of a Coolpix camera but noted it can be used in calculating canopy gap measurements. Englund et al., 2000 Compared a digital Nikon 950 and a film camera to find that low resolution images from the Nikon 950 were an adequate comparison to film cameras. Grimmond et al., 2001 Compared a Nikon 950 Coolpix to a plant canopy analyser and found that the Nikon was an effective and easy approach to canopy analysis. Gap function Analysis and Estimation of tree canopies Hu et al., 2009 Uses a Nikon 950 Coolpix camera to take hemispherical images to calculate gap size and shape within a tree canopy. Gap function Analysis and Estimation of tree canopies Zhang et al., 2005 Researched the effect of exposure on calculating the leaf area index and gap function analysis using a Nikon Coolpix 4500. Lang et al., 2010 Calculated gap function of canopies using a Nikon Coolpix 4500 and compared it to the Canon EOS 5D cameras. Chianucci et al. 2016 Used a Nikon 4500 to compare gap functions in forested canopies. Danson et al., 2007 A Nikon 4500 was used as a comparison to terrestrial laser scanning. Adaption or calibration of Nikon cameras Chapman, 2007 Adapted a Nikon 4500 camera to make in near infra-red in order to better estimate sky-view factors and the woody bark index of tree canopies. Baret, & Agroparc, 2004 Used a Nikon 4500 in order to determine the optical centre of an image using a fisheye lens. Ishida, 2004 Created threshold software for colour images from a Nikon 950 camera. Table 1 List of sample studies that use Nikon Coolpix cameras.65 66 Studies Camera used Approach Kelley and Krueger, 2005 HemiView 2.1 digital image system Used a 20-megapixel SLR CMOS camera as part of the HemiView software (Delta Devices 2017) to record canopy structure in riparian environments Duveiller and Defourny, 2010 Canon PowerShot A590 camera Used a Canon PowerShot A590 camera to assess batch processing of hemispherical images Rich, 1990 Canon T90 Minolta X700 Nikon FM2 Olympus OM4T Comprehensive instructions on how to take hemispherical photography with a list of cameras suitable for research Urquhart, et al., 2014 Allied Vision GE-2040C camera Uses sky-view factors from a high dynamic range camera to calculate short term solar power forecasting Wagner and Hagemeier, 2006 Canon AE-1 camera Used a Canon camera to estimate leaf inclination angles on tree canopies Table 2 Studies using alternative cameras for hemispherical photography. 67 68 The Nikon Coolpix range of cameras remains a key tool in forest climatology (Error! 69 Reference source not found. and Table 2Error! Reference source not found.). 70 Unfortunately, the Coolpix range is no longer readily available (Nikon, 2016) with 71 digital camera technology advancing considerably in the interim making models such 72 as the Coolpix 4500 camera appear large and bulky with a relatively poor battery life 73 and low image resolution (3.14 megapixels). However, even today, the FC-E8 74 fisheye lens remains one of the least distorted on the market (Holmer et al., 2001) 75 and as such, the camera series remains very popular with researchers as a tried and 76 tested means to collect hemispherical imagery (Chapman, 2007). A significant 77 further advantage of the Coolpix range of cameras was the ability to easily convert 78 the camera to take near infra-red (NIR) imagery. By adapting a camera in this way, it 79 significantly enhances its functionality in the forest environment as due to the highly 80 reflective nature of vegetation it becomes easier to distinguish this from woody 81 elements and other features in imagery when taken in NIR; which can then be used 82 to assess the health and density of tree canopies (Chen et al., 1996; Turner et al., 83 1999). 84 Overall, the Nikon Coolpix camera has reached the point where it is informally 85 viewed as a standard device for this purpose, but with dwindling numbers now 86 available for purchase on internet auction sites, there is a need to investigate new 87 and more sustainable means to collect data in the long term. Whilst new digital 88 cameras are available on the market, the approach explored in this paper is to 89 investigate whether a low-cost alternative can be developed using readily available 90 off-the-shelf components. 91 2. Methods 92 2.1 Adapting a Raspberry Pi 93 The Raspberry Pi is a range of small computers designed to minimise the cost of 94 computing and thus make it, and computer programming more generally, accessible 95 to a wide audience. After a prolific launch, it now has a worldwide following of 96 developers focussed on producing generic code and peripherals for use in a range of 97 applications. As an example, the computer can now be readily fitted with a 98 Raspberry Pi camera and subsequently programmed to take images at set time 99 intervals. 100 At the time of writing, the most popular Pi compatible camera available on the market 101 is the Pi camera which comprises of a Sony IMX219 9-megapixel sensor. This is 102 available either as a standard device or as a Pi NoIR camera where the infra-red 103 blocking filter (needed by modern digital cameras due to the inherent capability to 104 see beyond the visible spectrum: Chapman, 2007) has been removed (Raspberry Pi, 105 2016). As outlined in the previous section, NIR capability improves the utility of the 106 approach for use in forested environments. 107 2.2 Comparison of Fisheye lenses 108 Unfortunately, a fisheye lens is presently not available that has been specifically 109 designed for the Pi NoIR camera. However, due to the recent proliferation of 110 smartphone photography, there is a wide range of fisheye lenses that are now 111 available for smartphones which have the potential to be used. The key 112 consideration here, as per Holmer et al, (2001), is to select a lens with minimal 113 distortion to reduce error in later image analyses. This can be achieved by testing 114 the equiangularity of the lens by calculating any distortions in the radial distance. As 115 shown in Figure 2, the aim is to acquire an image where the radial distance is 116 directly proportional to the zenith angle (Chapman, 2008). 117 118 (a) (b) (c) (d) (e) Figure 2 (a) Visual comparison of Nikon Coolpix camera, (b) smart phone camera 119 with attached 185˚ fisheye lens, (c) smart phone camera with attached fisheye lens 120 198˚, (d) smart phone camera with attached fisheye lens 180˚ and (e) smart phone 121 camera with attached fisheye lens 235 ˚ 122 123 A range of available fisheye lenses were tested for distortions (Table 3). In this initial 124 test, the fisheye lenses were clipped onto a Samsung Galaxy S5 Neo (Figure 1 b) 125 and placed under a large Perspex calibration dome marked at equal points along the 126 sides using a compass (Figure 1 c). A plumb bob was then used to position the 127 device directly below the centre of the dome before a series of images collected 128 (Figure 2). Measurement distortions were then calculated using Image-J software 129 (Figure 3). 130 131 Product Field of view Cost (At time of writing) Yarrashop fisheye lens 180 £7.99 First2Savv JTSJ-185-A01 fisheye lens 185 £8.99 AUKEY fisheye lens 198 £11.99 MEMTEQ universal fisheye lens 235 £10.99 Table 3 Mobile fisheye lenses specification. 132 133 134 Figure 3 Comparison of radial distortion between different mobile fisheye lenses and 135 Nikon Coolpix 4500 camera FC-E8 lens. 136 137 0.0 10.0 20.0 30.0 40.0 50.0 60.0 0.0 10.0 20.0 30.0 40.0 50.0 60.0 D IS T O R T E D D IS T A N C E ( C M ) MEAN RADIAL DISTANCE FROM DOME CENTRE (CM) Equiangular Dome 180˚ fisheye lens 185˚ fisheye lens 198˚ fisheye lens Nikon Coolpix FC-E8 lens 235˚ fisheye lens The results show that the 185˚ fisheye lens (Figure 2b) is most comparable with the 138 Nikon Coolpix FC-E8 lens (Figure 2a). It has a similar field of view (FOV) and 139 despite a slight reduction in image clarity at high radial distances, the 185˚ lens has 140 the lowest level of distortion (Figure 3). However, comparisons between the Nikon 141 Camera FC-E8 lens and other mobile fisheye lenses are not as favourable and all 142 display clear distortions and/or significant reductions in FOV. For example, the 180 ˚ 143 (Figure 2d) camera captures the lowest FOV of the compared fisheye lenses (Figure 144 3). The 198˚ fisheye lens (Figure 2c) has excellent clarity at high radial distances 145 however has a lower FOV then reported and high levels of distortion (Figure 3). 146 Conversely, the 234 ˚ fisheye lens (Figure 2e) has a high FOV however has high 147 levels of distortion, especially at high radial distances (Figure 3). Based on these 148 analyses, the 185˚ fisheye lens was chosen for further investigation. 149 2.3 Adapting a Pi Noir camera to take hemispherical images 150 In order to use the 185 ˚ fisheye lens with the Pi NoIR camera, a series of small 151 adaptations are required. Whilst these adaptations could be achieved using 3D 152 printing technology, this was achieved in this study using parts scavenged from the 153 First2Savv 185º fisheye lens (Figure 4Error! Reference source not found.a) and 154 tubing from a Waveshare Raspberry Pi Camera Module Kit (Figure 4b). The camera 155 component of the Waveshare kit was removed, using a saw and drill, to leave a 156 hollow tube. The tubing (Figure 4b) was then tied and secured to the base of the 157 Raspberry PI NoIR camera using thin wire (Figure 4c). The camera was then 158 attached to the Raspberry Pi board using the connector port (Figure 4d). 159 (a) (b) (c) (d) Figure 4 (a) 185˚ fisheye lens attached to base (b) base component of Raspberry Pi 160 fisheye module, (c) fisheye module attached to Raspberry Pi NoIR camera (d) 161 Camera module attached to a Raspberry Pi computer. 162 163 3. Comparison of Nikon camera and Pi NoIR Raspberry Pi camera. 164 3.1 General Specifications 165 Table 4 shows the specification comparison of both the Pi NoIR camera version 1 166 and 2, the Nikon Coolpix 4500 and the Nikon Coolpix 9000 camera. As has been 167 demonstrated in the previous section, the reported FOV can vary with individual 168 cameras (Grimmond et al., 2001) and therefore this has been estimated in this study 169 using a mechanical clinometer. The adapted Pi camera FOV (164 ˚ ) is less than the 170 Comment [JK2]: Removed resolution argument Nikon Coolpix FOV (176 ˚ ) which is hypothesised to be a consequence of the added 171 tubing (Figure 4b) causing some distortion and loss of image at ground level. 172 Nikon 900 Nikon 4500 Pi NoIR V1 Pi NoIR V2 Pixel range 1.2 megapixels 3.14megapixels 5 megapixels 8 megapixels Optical Zoom 3 x optical zoom lens 4 x optical zoom lens N/A N/A Field of View 183 0 FC-E8 lens (176 ˚ using a mechanical clinometer) 183 0 FC-E8 lens (176 ˚ using a mechanical clinometer) 185 o mobile fisheye lens (164 ˚ using a mechanical clinometer) 185 o mobile fisheye lens (164 ˚ using a mechanical clinometer) Dimensions 143 x 76.5 x 36.5mm (5.6 x 3.0 x 1.4 in.) 130 x 73 x 50mm (5.1 x 2.9 x 2.0 in.) 25 x 24 x 1mm 25 x 24 x 1mm Cost £100* £200* £25 £25 * Approximate Second-hand price 173 Table 4 Comparison of Coolpix cameras to Raspberry Pi cameras 174 175 3.2 Distortion Analysis 176 As hemispherical imagery is mostly used in the analysis of tree canopies, the loss of 177 information at ground level (i.e. high radial distances) is less of a concern. It is at 178 these extremities of the image where distortions are also more common and indeed 179 one of the main attractions of the Nikon Coolpix range of cameras (Holmer et al., 180 2001). Whilst an equiangular lens is not an essential requirement of a camera 181 system for this application, it does ensure fewer corrections are required and 182 minimises error in subsequent analysis. The distortions of the adapted fisheye lens 183 are again tested by using the Perspex calibration dome (Figure 5). 184 (a) (b) Figure 5 (a) Nikon Coolpix camera in a Perspex dome and (b) Raspberry Pi NoIR 185 camera with fisheye attached under Perspex dome. 186 187 The FOV of the adapted Pi camera is demonstrated to be less than the Nikon 188 camera however there is a greater level of distortion when using a Nikon Coolpix 189 camera (190 ). This difference is likely due to the size of the equipment with the Nikon Coolpix 191 camera being larger in size than the Pi camera lens (145 mm compared to 25mm). 192 With respect to equiangularity, there is a strong correlation between radial distance 193 distortions of the Nikon Coolpix FC-E8 lens camera and Raspberry Pi NoIR adapted 194 fisheye camera at 99.9% confidence level (195 ). 196 197 198 Figure 6 Radial Distortion of a Nikon Coolpix FC-E8 lens camera and a Raspberry Pi 199 camera with attachable fisheye lens. 200 201 0.0 10.0 20.0 30.0 40.0 50.0 60.0 0.0 10.0 20.0 30.0 40.0 50.0 60.0 D IS T O R T E D D IS T A N C E ( C M ) MEAN RADIAL DISTANCE FROM DOME CENTRE (CM) Nikon Coolpix FC-E8 lens 185˚ fisheye lens attached to Raspberry Pi NoIR camera Equiangular Dome 3.3 Sky-view factor Analysis 202 To further demonstrate the inter-device comparability, images were captured 203 devices for sky-view factor analysis (204 205 Nikon Coolpix FC-E8 Lens Nikon Coolpix FC-E8 Lens SVF Raspberry Pi Camera Raspberry Pi Camera SVF Raspberry Pi Camera – After Threshold Analysis Raspberry Pi Camera – Leaves only SVF (a) (b) (c) (d) (e) (f) (g) (h) Figure 7). The Images were then analysed using ‘Sky-View Calculator’ software 206 (Göteborg Urban Climate Group, 2018) developed by Lindberg and Holmer 207 (2010) using a process where the image was converted to binary (208 209 Figure 7), divided into concentric annuli before calculating the number of white Pixels 210 (sky) in each annulus and summed (Holmer et al. 2001; Johnson and Watson, 1984; 211 Nikon Coolpix FC-E8 Lens Nikon Coolpix FC-E8 Lens SVF Raspberry Pi Camera Raspberry Pi Camera SVF Raspberry Pi Camera – After Threshold Analysis Raspberry Pi Camera – Leaves only SVF (a) (b) (c) (d) (e) (f) (g) (h) Steyn 1980). Analyses were performed on the original imagery as well as images 212 cropped to have the same FOV. Table 5 shows that when the FOV is uncorrected, 213 the Pi overestimates the sky-view factor, but when this is corrected, the output is 214 very similar and is significant at the 99.9% level. 215 216 Nikon Coolpix FC-E8 Lens Nikon Coolpix FC-E8 Lens SVF Raspberry Pi Camera Raspberry Pi Camera SVF Raspberry Pi Camera – After Threshold Analysis Raspberry Pi Camera – Leaves only SVF (a) (b) (c) (d) (e) (f) (g) (h) Figure 7 Visual variations in sky-view factors when comparing a Nikon Coolpix FC-217 E8 lens with a 185˚ Raspberry Pi NoIR camera. 218 219 Image Sky-view factor Leaf-view factor Nikon Coolpix Camera (Non- adjusted FOV) Nikon Coolpix Camera (adjusted FOV) Raspberry Pi Camera. Raspberry Pi camera – contribution of leaves (a) 0.25 0.17 0.17 0.55 (b) 0.24 0.26 0.29 0.68 (c) 0.40 0.42 0.44 0.45 (d) 0.4 0.45 0.45 0.45 (e) 0.3 0.34 0.35 0.26 (f) 0.4 0.45 0.47 0.33 (g) 0.37 0.40 0.44 0.42 (h) 0.48 0.33 0.34 0.53 Table 5 Sky view factors of Nikon Coolpix camera adjusted FOV, Raspberry Pi NoIR 220 camera, Nikon Coolpix unadjusted FOV and Raspberry Pi leaves only images. 221 222 3.4 Near Infrared Capabilities 223 In addition to hardware availability, the advantages of using a Raspberry Pi NoIR 224 camera over a Nikon 4500 camera is the in-built near infra-red (NIR) technology. 225 Comment [JK3]: Changed an to a Comment [JK4]: Removed resolution argument Although it is also possible to convert the Nikon Coolpix camera to take NIR images 226 (Chapman, 2007), this involves substantial effort which risks damaging the camera. 227 The capability of the Pi NoIR was confirmed in this study. A simple threshold 228 analysis proved sufficient to remove all other aspects of the image except for 229 vegetation (230 231 Nikon Coolpix FC-E8 Lens Nikon Coolpix FC-E8 Lens SVF Raspberry Pi Camera Raspberry Pi Camera SVF Raspberry Pi Camera – After Threshold Analysis Raspberry Pi Camera – Leaves only SVF (a) (b) (c) (d) (e) (f) (g) (h) Figure 7Table 5). The differences in sky-view factor can then be calculated; from 232 this a leaf-view calculation were made and presented in Table 5, indicating an 233 approximation of leaf cover in the image and further highlights the utility of the 234 camera in forestry applications. 235 4. Conclusions 236 The Nikon Coolpix camera range has provided a reliable ‘standard’ solution for 237 obtaining hemispherical fisheye imagery for many years. However, whilst still fit for 238 purpose, an alternative is needed to ensure a sustainable means of data collection 239 moving forward. This paper has shown that comparable results can be provided with 240 a low-cost image collection system using readily available components. 241 The Pi NoIR camera provides an off-the-shelf NIR solution, making it perfect for use 242 in forested environments and thus removing the need for further adaptation (i.e. 243 removal of blocking filters and addition of cold mirrors: Chapman, 2007). However, 244 fisheye lenses are not yet readily available and hence there is presently a need to 245 carry out alternative adaptations such as those outlined in this paper, or the use of 246 simple 3D printing technology. However, the most positive result from this study is 247 the direct comparability of the imagery (and subsequent results from sky-view factor 248 analyses) obtained from the two techniques. Both systems have similarly low levels 249 of distortion, but there are minor differences in relation to the FOV. Further research 250 is needed to adapt the Raspberry Pi to make the sensor usable in the field; this 251 includes waterproofing the technology and testing the equipment at various 252 temperature ranges. A limitation of this study is that the technology was not tested 253 for interference from electronic or radio waves. 254 255 Comment [JK5]: Removed resolution comment Further advantages of the Raspberry Pi approach are the computing capability of the 256 device, which means it has internal logging capabilities and (once waterproofed) 257 could be left in the field in time lapse mode for long periods at a time, even relaying 258 imagery over the internet in real-time if communications are available. Overall, 259 moving forward there are many advantages to using the Raspberry Pi, however the 260 key conclusion is that a fit for purpose and dynamic solution for the collection of 261 hemispherical imagery can be readily produced at a low cost. 262 Acknowledgements 263 Funding for this research was provided by the Rail Safety and Standards Board 264 (RSSB) and the Engineering and Physical Sciences Research Council (EPSRC). 265 References 266 1. Anderson, M.C. 1964: Studies of Woodland light climate. Journal of Ecology 267 52, 27-41 268 2. Baret, F. and Agroparc, S. 2004: A simple method to calibrate hemispherical 269 photographs. INRA-CSE, France 270 (http://147.100.66.194/can_eye/hemis_calib3. pdf). 271 3. Bréda, N. J. 2003: Ground‐based measurements of leaf area index: a review 272 of methods, instruments and current controversies. Journal of experimental 273 botany, 54(392), 2403-2417 274 4. Chapman, L. 2008: An introduction to upside-down remote sensing, Progress 275 in Physical Geography 32 (5), 529-542 276 5. Chapman, L. 2007: Potential applications of near infra-red hemispherical 277 imagery in forest environments. Journal of Agricultural and Forest 278 Meteorology, 143(1), 151-156. 279 6. Chapman, L., Thornes, J. E., Muller, J. P., and McMuldroch, S. 2007: 280 Potential applications of thermal fisheye imagery in urban environments. IEEE 281 Geoscience and Remote Sensing Letters, 4(1), 56-59. 282 7. Chen, J. M. 1996: Optically-based methods for measuring seasonal variation 283 of leaf area index in boreal conifer stands. Journal of Agricultural and Forest 284 Meteorology, 80(2-4), 135-163. 285 8. Duveiller, G., and Defourny, P. 2010: Batch processing of hemispherical 286 photography using object-based image analysis to derive canopy biophysical 287 variables. Proceedings of GEOBIA, 1682-1777. 288 9. Chianucci, F, Leonardo D, Donatella G, Daniele B, Vanni N, Cinzia L, 289 Andrea R and Piermaria C 2016: "Estimation of canopy attributes in beech 290 forests using true colour digital images from a small fixed-wing 291 UAV." International journal of applied earth observation and 292 geoinformation 47. 60-68. 293 10. Chianucci, F., Macfarlane, C., Pisek, J., Cutini, A., and Casa, R. 2015: 294 Estimation of foliage clumPing from the LAI-2000 Plant Canopy Analyzer: 295 effect of view caps. Trees, 29(2), 355-366. 296 11. Danson, F. M., Hetherington, D., Morsdorf, F., Koetz, B., and Allgower, B. 297 2007: Forest canopy gap fraction from terrestrial laser scanning. IEEE 298 Geoscience and remote sensing letters, 4(1), 157-160. 299 12. Englund, S. R., O'brien, J. J., and Clark, D. B. 2000: Evaluation of digital 300 and film hemispherical photography and spherical densiometry for measuring 301 forest light environments. Canadian Journal of Forest Research, 30(12), 302 1999-2005. 303 13. Delta 2017: HemiView Forest Canopy Image Analysis System, 304 https://www.delta-t.co.uk/product/hemiview/ accessed 30/01/2018 305 14. Frazer, G. W., Fournier, R. A., Trofymow, J. A., and Hall, R. J. 2001: A 306 comparison of digital and film fisheye photography for analysis of forest 307 canopy structure and gap light transmission. Journal of Agricultural and forest 308 meteorology, 109(4), 249-263. 309 15. Garrigues, S., Shabanov, N. V., Swanson, K., Morisette, J. T., Baret, F., 310 and Myneni, R. B. 2008: Intercomparison and sensitivity analysis of Leaf 311 Area Index retrievals from LAI-2000, AccuPAR, and digital hemispherical 312 photography over croplands. Journal of Agricultural and Forest meteorology, 313 148(8), 1193-1209. 314 16. Grimmond, C. S. B., Potter, S. K., Zutter, H. N., and Souch, C. 2001: RaPid 315 methods to estimate sky-view factors applied to urban areas. International 316 Journal of Climatology, 21(7), 903-913 317 17. Göteborg Urban Climate Group, 2018: Sky-view Calculator, University of 318 Göthenburg, https://gvc.gu.se/english/research/climate/urban-319 climate/software/download. Accessed 30/01/2018 320 18. Hale, S. E., and Edwards, C. 2002: Comparison of film and digital 321 hemispherical photography across a wide range of canopy densities. Journal 322 of Agricultural and Forest Meteorology, 112(1), 51-56. 323 19. Hall, R. J., Fournier, R. A., and Rich, P. 2017: Introduction. In Hemispherical 324 Photography in Forest Science: Theory, Methods, Applications (pp. 1-13). 325 Springer Netherlands. 326 https://www.delta-t.co.uk/product/hemiview/ https://gvc.gu.se/english/research/climate/urban-climate/software/download https://gvc.gu.se/english/research/climate/urban-climate/software/download 20. Holmer, B., Postgård, U. and Eriksson, M., 2001: Sky view factors in forest 327 canopies calculated with IDRISI. Theoretical and Applied Climatology, 68(1), 328 pp.33-40. 329 21. Homolová, L., Malenovský, Z., Hanuš, J., Tomášková, I., Dvořáková, M., 330 and Pokorný, R. 2007: Comparaison of different ground techniques to map 331 leaf area index of Norway spruce forest canopy. In Proceedings of the 332 International Society for Photogrammetry and Remote Sensing (ISPRS) 333 22. Hu, L., Gong, Z., Li, J., and Zhu, J. 2009: Estimation of canopy gap size and 334 gap shape using a hemispherical photograph. Trees, 23(5), 1101-1108. 335 23. Ishida, M. 2004: Automatic thresholding for digital hemispherical 336 photography. Canadian Journal of Forest Research, 34(11), 2208-2216. 337 24. Johnson, G. T., and Watson, I. D. 1984: The determination of view-factors in 338 urban canyons. Journal of Climate and Applied Meteorology, 23(2), 329-335. 339 25. Jonckheere, I., Fleck, S., Nackaerts, K., Muys, B., CopPin, P., Weiss, M., 340 and Baret, F. 2004: Review of methods for in situ leaf area index 341 determination: Part I. Theories, sensors and hemispherical photography. 342 Journal of Agricultural and Forest meteorology, 121(1), 19-35. 343 26. Kelley, C.E. and Krueger, W.C., 2005: Canopy cover and shade 344 determinations in riparian zones. Journal of the American Water Resources 345 Association, 41(1), pp.37-046. 346 27. Lang, M., Kuusk, A., Mõttus, M., Rautiainen, M., and Nilson, T. 2010: 347 Canopy gap fraction estimation from digital hemispherical images using sky 348 radiance models and a linear conversion method. Journal of Agricultural and 349 Forest Meteorology, 150(1), 20-29. 350 28. Leblanc, S. G., Chen, J. M., Fernandes, R., Deering, D. W., and Conley, A. 351 2005: Methodology comparison for canopy structure parameters extraction 352 from digital hemispherical photography in boreal forests. Journal of 353 Agricultural and Forest Meteorology, 129(3),187-207. 354 29. Lindberg, F., and Holmer, B. 2010: Sky View Factor Calculator, Göteborg 355 Urban Climate Group, Department of Earth Sciences, University of 356 Gothenburg 357 30. Liu, Z., Wang, C., Chen, J. M., Wang, X., and Jin, G. 2015: Empirical 358 models for tracing seasonal changes in leaf area index in deciduous broadleaf 359 forests by digital hemispherical photography. Journal of Forest Ecology and 360 Management, 351, 67-77. 361 31. Nikon, CoolPix 4500 product archive, 2017: 362 http://imaging.nikon.com/lineup/CoolPix/others/4500/ accessed 30/01/2018 363 32. Raspberry Pi 2016: https://www.RaspberryPi.org/products/Pi-noir-camera/, 364 Accessed 2017 365 33. Rich, P. M. 1990: Characterizing plant canopies with hemispherical 366 photographs. Remote sensing reviews, 5(1), 13-29. 367 34. Ringold, P. L., Sickle, J., Rasar, K., and Schacher, J. 2003: Use of 368 hemispheric imagery for estimating stream solar exposure. Journal of the 369 American Water Resources Association, 39(6), 1373-1384. 370 35. Schwalbe, E., Maas, H. G., Kenter, M., and Wagner, S. 2009: Hemispheric 371 image modeling and analysis techniques for solar radiation determination in 372 forest ecosystems. Journal of Photogrammetric Engineering and Remote 373 Sensing, 75(4), 375-384. 374 http://imaging.nikon.com/lineup/CoolPix/others/4500/ https://www.raspberrypi.org/products/pi-noir-camera/ 36. Steyn, D.G., 1980: The calculation of view factors from fisheye‐lens 375 photographs: Research note. 376 37. Turner, D. P., Cohen, W. B., Kennedy, R. E., Fassnacht, K. S., and Briggs, 377 J. M. 1999: Relationships between leaf area index and Landsat TM spectral 378 vegetation indices across three temperate zone sites. Remote sensing of 379 environment, 70(1), 52-68. 380 38. Urquhart, B., Kurtz, B., Dahlin, E., Ghonima, M., Shields, J. E., and 381 Kleissl, J. 2014: Development of a sky imaging system for short-term solar 382 power forecasting. Atmospheric Measurement Techniques Discussions, 7, 383 4859-4907. 384 39. Wagner, S., and Hagemeier, M. 2006: Method of segmentation affects leaf 385 inclination angle estimation in hemispherical photography. Journal of 386 Agricultural and Forest Meteorology, 139(1), 12-24. 387 40. Watson, I. D., and Johnson, G. T. 1987: Graphical estimation of sky view‐388 factors in urban environments. Journal of Climatology, 7(2), 193-197. 389 41. Zhang, Y., Chen, J.M. and Miller, J.R., 2005: Determining digital 390 hemispherical photograph exposure for leaf area index estimation. Agricultural 391 and Forest Meteorology, 133(1-4), pp.166-181. 392 393