**1. Introduction**

The European Union has set itself the ambitious goal of increasing energy efficiency by 20% by the year 2020. Lighting represents approximately 50% of the electricity consumption of cities. Therefore, European cities can play a very important role in reducing their carbon footprint by implementing innovative solutions that respect the environment in public lighting installations. Examples of such solutions are lighting installations powered by renewable sources [1] or the replacement of lamps by others with better chromatic reproduction [2]. More specifically, it was found that for lighting classes with lower luminance levels, metal halides (MH) lamps were economically comparable or even more favorable than high pressure sodium (HPS) lamps.

Now more than ever, after the recent economic crisis, local governments are obliged to reduce their expenses in order to meet their financial commitments. In this regard, one of the largest expenses in towns and cities is energy consumption, especially by public lighting [3]. Another important consideration is the need to comply with local, national and international regulations [4–6] regarding the optimization of lighting levels in order to reduce or eliminate light pollution [7]. Precisely for this reason, administrations must have access to up-to-date information regarding public lighting and light values in order to remedy any actual or potential non-compliance.

This paper presents an innovative method for compiling information on lighting levels, uniformity and energy consumption in geographic information systems (GIS) [8,9]. Such data are extremely useful when performing energy audits of public lighting. All the information is stored as layers, which form maps of public lighting. This method provides valuable data to town and city administrators, who need to know the current state of their street lighting installations. It also generates street lighting maps of their cities with real luminance or illuminance values as well as electricity consumption levels. This information can thus be used to make decisions and carry out corrective actions.

Currently, public lighting maps are based on countless measurements of luminance or illuminance values at street level (obtained with luminance meters or luxmeters). These values are then transferred to a map. Since each measurement must be associated with a value that identifies its geographic coordinates, measurements must also be georeferenced. The number of measurements required for these maps complicates data collection considerably.

Nowadays, the illumination of roads or streets is evaluated with different methods. For example luminance can be determined with spot luminance meters (or photometers), which measure the luminance of a small area [10]. This method involves making many measurements to evaluate the illumination of a road. Other authors [11] measured illuminance levels using a luxmeter placed on top of a moving vehicle and combined measurements with the location data of a distance measurement instrument or GPS. To quickly measure luminance at various points, another alternative is the use of digital cameras [12,13], where a single image evaluates the illumination of a larger road area.

Our study used digital cameras integrated in low-weight drones to take nocturnal orthoimages [14] of streets or roads and thus evaluate their lighting levels. Unlike other options, this method permitted a more rapid evaluation because a single image provided simultaneous information of several streets. The methodology described in this paper allowed us to obtain a large quantity of luminous data. For this purpose, extensive areas were covered in a short time in order to determine the average real values of luminance or illuminance, electricity consumption, and the energy efficiency of public lighting installations. Our research objectives were the following:


The rest of the paper is organized as follows: Section 2 explains the materials and methods used in the procedure and the background of energy efficiency in street lighting and energy classification. Sections 3 and 4 present the results and discuss them in the context of various examples; and finally, Section 5 lists the most important conclusions that can be derived from this research.

### **2. Materials and Methods**

The general procedure for obtaining lighting levels with airborne digital cameras and estimating energy efficiency levels was composed of the following stages:


Aerial images (orthoimages) were captured with an airborne digital camera with a 4056 × 3040 pixels CMOS (Complementary Metal-Oxide-Semiconductor) sensor, ISO (ISO is a photographic film's sensitivity to light and acronym of International Organization for Standardization) range of 100–1600/3200 (auto/manual mode) and shutter speed ranging from 8 to 1/8000 s. Regarding optics, the camera has a lens with 85<sup>0</sup> Field of View (FOV) and f/2.8.

Even though this camera is not an instrument that measures luminance magnitude, it is possible to transform it into a photometer [12,18]. The information obtained in each pixel of its sensor can thus be related to the real luminance value of luminance captured in an image. To detect the luminance through any image (object image), we used a standard image of a luminous source of known luminance. Both images (object and standard images) had to be taken under the same conditions (exposure time, ISO value, aperture, camera-object distance, etc.). The luminous source consisted of a system of 15 halogen lamps arranged inside a box and covered by a diffusing screen. As shown in Figure 1, the lamps are connected in parallel to a variable power supply, which provided different luminance values. To ascertain the luminance emitted by the luminous source as a function of electrical power, we used a photometer that measures luminance (see Figure 2) and a wattmeter that measures electrical power.

**Figure 1.** Source used as a luminance pattern. The left image shows the lamp distribution inside the box connected to the wattmeter and a variable power supply. The right image shows the screen diffuser over the box.

**Figure 2.** Hagner universal photometer S3 used to measure luminance.

### *2.1. Camera Calibration Procedure*

The average luminance values in street lighting range from 0.2 to 5 cd/m2, depending on the lighting class of the road. This is thus the luminance range of the pattern. To obtain this range of luminance, the lighting source is connected to a variable voltage source. Accordingly, for each voltage or power consumed (measured with a wattmeter), the system emits a certain luminance value. Figure 3 shows the relationship between the luminance emitted and the electrical power consumed. All measurements were carried out in the laboratory, after which a polynomial was formulated to fit them.

**Figure 3.** Relationship between the luminance emitted by the standard luminous source as a function of the electrical power consumed and the coefficients of a second-order polynomial (*ax*<sup>2</sup> *+ bx + c*) fit to the measured values.

The exposure time of the camera can be adapted to the luminance level of the road, so as not to saturate the detector. For this same reason, it is also necessary to adapt the luminance of the light source (lighting pattern) to the illumination level of the road. To do this, the voltage of the source is modified where the pattern is connected. The luminance is obtained by reading the wattmeter and applying the polynomial adjustment in Figure 3.

The luminance value in cd/m<sup>2</sup> from *RGB* (*RGB* refers to additive primary colors Red, Green and Blue, a color model used in digital cameras) images is calculated with the following expression [12,15]:

$$L = \frac{\mathbf{Y} \cdot f\_s^2}{K \cdot t \cdot S\_{ISO}} \tag{1}$$

where *fs* is the aperture of the camera; *t* is the exposure time in seconds; *SISO* is the ISO value; *K* is the calibration constant; and *Y* is the photopic luminance value [19] calculated from the *RGB* color space [20] as follows:

$$Y = 0.2126 \cdot R + 0.7152 \cdot G + 0.0722 \cdot B \tag{2}$$

Accordingly, the calibration procedure involves obtaining the constant *K*. For this purpose, we employed an image in RAW (RAW format refers to digital natives) format of the luminous pattern. The values *R*, *G* and *B* of each pixel were used to obtain the *Y* value, and the camera parameters and the known luminance of the pattern were used to deduce the calibration constant. The image processing tool for the treatment of the digital images was IRIS, a software for astrophotography [21]. This software is free for non-commercial usage and is able to process images of different formats, including photographic ones.

### *2.2. Method for Measuring Road Lighting*

According to European standard EN 13201-3 [22,23], the average luminance of a surface must be calculated as shown in Figures 4 and 5. A photometer is used to measure luminance at several points located in an area delimited by two consecutive street lights. The photometer must be at a distance of 60 m from the calculation area and at a height above the ground of 1.5 m. In our case, instead of a photometer we used the airborne camera, calibrated as described in the previous section. However, the procedure followed was the one indicated in the European standard.

**Figure 4.** Field of calculation for carriage luminance, where (**a**) is the side-view and (**b**) the top-view.

**Figure 5.** Digital image in which the calculation area of average luminance is indicated by a red frame. According to the standard procedure the camera is located at a distance of 60 m from the calculation area and at a height of 1.5 m.

### *2.3. Energy Efficiency of Street Lighting Installation*

The energy efficiency of the lighting installation can be defined as follows:

$$
\varepsilon\_X = \frac{A\_T X\_{av}}{P\_T} \tag{3}
$$

where *AT* is the total illuminated surface in m<sup>2</sup> of the street; *PT* is the total electrical power in watts installed, including the light sources or lamps and electrical auxiliary devices; and *Xav* is the average value (luminance, in cd/m2, or illuminance, in lx) on the ground. Depending on the road type, the lighting class must be based on the luminance or illuminance magnitude.

A common approach to the energy classification of public lighting installations is the use of the *SLEEC* (street lighting energy efficiency criterion) [5,22] as a whole system indicator, based on the efficiency of the lamp, ballast (when used) and luminaire. The formula for the *SLEEC* indicator (or power density indicator) depends on the photometric measurement (illuminance or luminance) used to calculate street lighting for specific road classes and is the following:

$$SX = \frac{1}{\varepsilon\_X} \tag{4}$$

where ε*X* is the energy efficiency of the street lighting installation; and *X* refers to the type of luminous parameter, based on illuminance (E) or luminance (L).

Although the definition of this indicator is still a topic of debate [24], various compatible approaches are currently available (see Table 1), in which *SLEEC* values are combined with the energy labelling system based on the European standard EN 13201-5. As can be observed in Table 1, the lower the consumption is, the better the energy class of the lighting installation. Therefore, the inverse of energy efficiency is a reasonable indicator that can be used to allocate an energy class to the installation.

Lighting engineering software applications base their calculations on data specifications of the street to be illuminated (i.e., dimensions, desired average illuminance, luminaire height in relation to building height, etc.). Another aspect considered is the configuration or arrangemen<sup>t</sup> of the installation (one-sided, two-side staggered, two-sided coupled, etc.). The main parameters obtained are the spacing between luminaires and overall uniformity.


**Table 1.** Energy efficiency classification of street lighting installations according to the street lighting energy efficiency criterion (*SLEEC)* indicator [25].

Notwithstanding, energy efficiency has become an increasingly important topic. Indeed, in many countries (e.g., the 28 European Union countries), the classification of lighting installations is based on this parameter [5,6]. It is also used in others, such as Australia and New Zealand [26].
