• No se han encontrado resultados

VI. JUSTIFICACION

3. CAPÍTULO III: DISEÑO DEL SISTEMA

3.3. DISEÑO DE BASE DE DATOS

3.3.1. Modelo Lógico de Datos

Some urban features can be extracted using suitable parameters in the commercial software. This is not the case of manhole covers. Therefore, an algorithm was designed to extract these elements.

36

The algorithm was designed to detect the different materials of the manholes based on the returned reflection value (intensity). Since manholes are usually made of steel and iron, this is a characteristic that can help in their detection, and based on as previous studies that have demonstrated the effectiveness of using this parameter from the point cloud information (Tatoglu and Pochiraju 2012; Núñez et al., 2008; Balaguer-Puig et al. 2017).

Color is not usually a representative parameter in the properties of manholes. Therefore, the algorithm designed in this thesis to detect manholes is based on the geometry (circular or rectangular) (Figure 3-6) and intensity, i.e. the algorithm will check the shape for matched intensity objects and skip objects that are not the required shape. It starts by defining the variables that will be used, such as the coordinates (X, Y, Z), time (t) and intensity (i), then uploads the LiDAR data in LAS format.

Figure 3-6: Sample of circular manhole with a diameter of 59.6 cm (left) and a rectangular manhole (right) provided by the manufacturer (http://www.nibf.co.za/product/circular-manhole-cover-and-

frame-type-2a-and-2b/)

Figure 3-7 shows the flowchart of the procedure in every step of the method for manhole detection. The entire algorithm has been implemented in Matlab script, so that it can be used by any operator automatically.

37

Figure 3-7: Detecting manholes workflow

The first step after checking the data is related to the intensity. This variable is used to segment the point cloud regarding the manhole material. The time variable is used to link with other variables, as the LiDAR data is recorded by time. Eventually, coordinates are used to detect the shape of the manhole.

38

Manholes have a different intensity due to their material than the points around them that belong to the pavement. These values were determined based on the manhole material, which was identified after contacting the manufacturer, and the wavelength of the laser beam that was used.

The proposed algorithm first defines the intensity for the input data objects, then if the object intensity 𝐼𝑝𝑜𝑖𝑛𝑡 is different from the required value (𝐼𝑚𝑖𝑛− 𝐼𝑚𝑎𝑥), it skips this feature and

searches for the next one. The intensity value should lie between the minimum and maximum intensity values set previously in the algorithm by the operator who knows the values range. Objects with the required intensity will be run through the next step, which is the shape check. In the second step shown in Figure 3.3. the point cloud is checked to detect the manhole covers based on their shapes. The parameters given by the operator are the sides for rectangular manholes or the radius for circular ones. For rectangular manholes, the shape can be calculated from a plane (Equation 1).

S =|ax + by + cz + d| √a2+ b2+ c2

(1)

Where ‘a’ is dimension in the x direction, ‘b’ is dimension in the y direction, ‘c’ is dimension in the z direction and ‘d’ is the distance from the origin to the required plane. Since all detected manholes are at pavement level, the coefficient “d” will be the same as the pavement height, in this case the “c” value will be zero.

For circular manholes, the shape can be calculated from a plane (2).

𝑥2+𝑧2− 𝑟2 = 𝑜 (2)

Where ‘x’ is dimension in the x direction, ‘y’ is dimension in the y direction, and r is the radius. The manhole shape and dimensions in the selected area are usually known by the municipality from the current basemap. The manhole features that match this known shape and dimensions will then be detected and extracted in the output. These identified cluster objects are irregular, so a last step is required to refine the shape of each identified manhole to a perfect plane with four straight sides and four right angles with equal or unequal adjacent sides, as per the manhole shape. To identify the perfect plane boundaries, the Minimum Bounding Rectangle (MBR) from the convex hull is computed to obtain the envelope of each cluster object (Kwak et al., 2012).

Considering that the target manhole shape in the study area is composed of sets of rectangles, it can be refined as several rectangles that represent the manhole boundaries.

Some steps must be considered to implement the MBR:

 Define the convex hull (Polygon) from all the points. The MBR-derived rectangle is represented by the minimum and maximum coordinate values for the boundary points in 2D. It is determined by choosing the minimum area rectangle among the rectangles

39

with an arbitrary orientation that contains the vertices of boundaries (Figure 3-8: 1,2 and 3).

 For each edge/segment of the convex hull polygon: o Find and calculate the edge orientation

o Find each segment by rotating the convex hull using the previous orientation, in order to easily compute the bounding rectangle area with min/max of x/y of the rotated convex hull (Figure 3-8: 4).

o Return the rectangle corresponding to the minimum area found

Figure 3-8: Manhole detection steps: 1) Result of the filter by intensity, 2) Identification of the points in each cluster, 3) Obtaining the MBR of each cluster, 4) Final shape adjustment based on the known

dimensions.

The process is repeated with all known dimensions until all valid manholes with the required criteria (shape and intensity) are detected. Figure 3-9 shows the minimal bounding rectangle around the LiDAR points for a rectangular manhole.

40

Figure 3-9: Minimal bounding box for a square manhole around 3D MLS point cloud manhole points (units in cm)

In our methodology, unlike other previous studies that use intensity-based images or ground orthophotos (Yu et al. 2014a, 2014b; Yu et al. 2015; Wei et al. 2019), we integrate intensity with collection time to filter the point cloud that is defined as high intensity in the same time period. We add the MBR method that adjusts the final shape of the manhole using the benefit of previously known dimensions in the selected area.