Infrastructure monitoring is being transformed by the advancements on remote sensing, unmanned vehicles and information technology. The wide interaction among these fields and the availability of reliable commercial technology are helping pioneer intelligent inspection methods based on digital 3D models. Commercially available Unmanned Aerial Vehicles (UAVs) have been used to create 3D photogrammetric models of industrial equipment. However, the level of automation of these missions remains low. Limited flight time, wireless transfer of large files and the lack of algorithms to guide a UAV through unknown environments are some of the factors that constraint fully automated UAV inspections. This work demonstrates the use of unsupervised Machine Learning methods to develop an algorithm capable of constructing a 3D model of an unknown environment in an autonomous iterative way. The capabilities of this novel approach are tested in a field study, where a municipal water tank is mapped to a level of resolution comparable to that of manual missions by experienced engineers but using $63\%$ . The iterative approach also shows improvements in autonomy and model coverage when compared to reproducible automated flights. Additionally, the use of this algorithm for different terrains is explored through simulation software, exposing the effectiveness of the automated iterative approach in other applications.

The smart society, including smart infrastructures is developing quickly. The smart monitoring of the water distribution systems is a significant part of this development, trying to address possible problems in the Water Distribution Networks (WDNs) such as water leakages, pressure instability and water contamination. By sampling and sensing important parameters in the water distribution infrastructure and sending these data to control systems through a low power wide area network (LPWAN), a Cyber Physical System (CPS) of a digitalized WDN can be built. This thesis provides support for the design of such a CPS, by the design of a simulation framework, which includes coordinated WDN and communication network simulators. First, we make a selection on the communication network simulators and protocol stacks. Network simulator 3 (NS-3) is chosen as the simulation tool and Long Range (LoRa)/long range wide-area network (LoRaWAN) is chosen as the network protocol. The EPANET software is used as WDN simulator. Then, the existing implementation of the LoRaWAN protocol stack in NS-3 is modified, to allow the network simulator run parallel to the WDN simulator, and to transmit WDN sampling data to a control center. Finally, the effect of the communication network properties on the performance of the WDN monitoring system is evaluated, and it is shown that the communication delays can affect the monitoring performance even in small systems. The thesis provides the first steps of the development of a simulation environment of the cyber physical system of digitalized water distribution networks, and is expected to support further research in the area. / Det smarta samhället, inklusive smarta infrastrukturer, utvecklas snabbt. Den smarta övervakningen av vattendistributionssystemen är en betydande del av denna utveckling, som försöker lösa eventuella problem i vattendistributionsnäten (WDN) såsom vattenläckage, tryckinstabilitet och vattenförorening. Genom att ta prov och känna av viktiga parametrar i vattendistributionsinfrastrukturen och genom att skicka dessa data till ett kontrollsystem via ett LPWAN, kan ett CPS av ett digitaliserat WDN byggas. Detta examensarbete ger stöd för designen av en sådan CPS, genom utformningen av ett simuleringsramverk, som inkluderar koordinerade simulatorer för WDN och kommunikationsnätverk. Först gör vi ett urval på simulatorer för protokollstackar och kommunikationsnätverk. NS-3 väljs som simuleringsverktyg och LoRa/LoRaWAN väljs som nätverksprotokoll. EPANET-mjukvaran används som WDN-simulator. Sedan modifieras den befintliga implementeringen av LoRaWAN-protokollstacken i NS-3, för att tillåta nätverkssimulatorn att köras parallellt med WDN-simulatorn och för att överföra WDN-samplingsdata till ett kontrollcenter. Slutligen utvärderas effekten av kommunikationsnätverkets egenskaper på prestandan hos WDN-övervakningssystemet, och det visas att förseningarna i kommunikationsnätet kan påverka övervakningsprestandan även i små system. Avhandlingen ger de första stegen i utvecklingen av en simuleringsmiljö av det cyberfysiska systemet för digitaliserade vattendistributionsnät, och förväntas stödja ytterligare forskning inom området.

Structural health monitoring and building assessment are crucial to acquire structures’ states and maintain their conditions. Besides human-labor surveys that are subjective, time-consuming, and expensive, autonomous image and video analysis is a faster, more efficient, and non-destructive way. This thesis focuses on crack detection from videos, crack segmentation from images, and building assessment from street view images. For crack detection from videos, three approaches are proposed based on local binary pattern (LBP) and support vector machine (SVM), deep convolution neural network (DCNN), and fully-connected network (FCN). A parametric Naïve Bayes data fusion scheme is introduced that registers video frames in a spatiotemporal coordinate system and fuses information based on Bayesian probability to increase detection precision. For crack segmentation from images, the rotation-invariant property of crack is utilized to enhance the segmentation accuracy. The architectures of several approximately rotation-invariant DCNNs are discussed and compared using several crack datasets. For building assessment from street view images, a framework of multiple DCNNs is proposed to detect buildings and predict their attributes that are crucial for flood risk estimation, including founding heights, foundation types (pier, slab, mobile home, or others), building types (commercial, residential, or mobile home), and building stories. A feature fusion scheme is proposed that combines image feature with meta information to improve the predictions, and a task relation encoding network (TREncNet) is introduced that encodes task relations as network connections to enhance multi-task learning.