education

1. Duration: 60 min High School Grade: 11 CCSS, NGSS
Deep Learning based Airborne LIDAR Image
Detection and Classification
towards
Intelligent Surveys
of
EHV Transmission Line Corridor
Priti Nahar (32216001)

2. Introduction
Power
System
Reliability &
Resilience
Equipment
Installation
Construction
Operation &
Maintenance
Manpower
Planning
Challenges
Workforce safety hazards
Low working efficiency
Limitation at hard to reach location
Misjudgement

3. Introduction
Routine Line and Tower Inspection
Line Corridor Inspection
Emergency Management

4. Introduction
Inspection Object Fault / Observations Specification
Line & Tower Tower Tilted, deformed, and corroded tower; the missing bolt and pin;
damaged and displaced foundation; foundation defects.
Conductor Loose, rupture, damaged lines; wire break; overheating; corona
detection, corroded, ice-coating lines; line sagging; attachment,
twining; distance to the ground object.
Insulator Overheating; contamination; Missing, cracking piece; corona
detection; parts corrosion.
Hardware Fittings Fitting fracture, abrasion, rust; missing pins; corrosion; loose part;
burning trace; adherent object; corona detection.
Line Corridor Buildings Unauthorized building or structure; unexpected antenna; newly
insufficient safe distance.
Vegetation Ultra-high or high vegetation encroachment; fast growing plant.
Emergency
Management
Natural Disaster Impact of floods, earthquakes, landslide, lightning, etc.
Survey &
Mapping
Terrain & Vegetation Optimise Routing : avoidance of Residential, Wild Life, Reserve
Forest, Crop and Tree damages

5. Comparison
Data Source Main Application in Transmission Line Specific Advantages Challenges
ALS data
(Highest point density
from fixed-wing aircraft
dozens of points/m2 and
from helicopters hundreds
of points/m2)
• Mapping of conductors and pylons
• Mapping of individual trees near power lines
• Detailed 3D data directly available, which allows
accurate modelling of power line structures and nearby
trees
• Not dependent on external lightning conditions
• Possibility for high-quality measurement of tree height
growth and for estimating future threats
• Flexibility in data acquisition, especially from
helicopters
• Recently possible also from UAVs
• High point density required for reliable mapping of small and
narrow objects
• Depending on the scanning geometry, not necessarily many
returns from vertical poles
Land-based mobile
mapping data
(Highest point density
thousands of points/m2)
• Mapping of conductors and pylons
• Inspection of power line components
• Very detailed 3D data directly available by laser
scanning
• Detailed images can also be acquired
• Panoramic imaging geometry enables accurate
reconstruction of towers and poles
• Possibility for high-quality estimates of vegetation
• Allows assessment of tree diameters and modelling of
the tree trunks
• Power lines not necessarily easy to reach with ground vehicles
• Land-based mapping can be slow, especially in forests
• The measurement geometry does not allow detection of
components from the top
UAV data
(Highest resolution of
images < 1 cm and highest
point density of laser
scanning hundreds of
points/
m2)
• Mapping of power lines and towers
• Inspection of power line components
• Detailed mapping of vegetation
• High flexibility in data acquisition
• Lower costs compared to other aerial platforms
• Potential for diverse applications
• Rapid development of UAV components and battery
technology is improving the feasibility for operational
applications
• Developing technology (e.g., operational and regulatory issues)
• Maximum flying time with current battery technology about 2 h
• (dependent of the platform)
• Combustion engines often unsuitable because of high-
amplitude vibrations

6. Introduction
Motivation Patrolling through UAV/ Drone: Decision conventional
mode
Drone Navigation: Manual
Target Detection : Manual
Vegetation Management
Drone patrolling in day time only

7. Introduction
Motivation Conventional Surveying not providing micro level
information
Not up to date :mismatching of information at field level.
Time Consuming: Manual Assessment of land use
including vegetation/ cropping pattern along the corridor
Time lag between survey and actual start of construction
Time Overrun & Cost Overrun

8. Power Transmission Project : Existing Practice of assessment
of damages and compensation
• Surveying is an important aspect of transmission line for the purpose
of assessing RoW.
• Preliminary route alignment/walk over survey is carried out through
desk top study with the help of Topo map, GIS/GPS, Google Map etc
to select most optimum route out of at least 3 alternative alignments
studied.
• At this stage, preferred route with schedule of angle point is identified
for taking of the detailed survey.

9. Power Transmission Project : Existing Practice of assessment
of damages and compensation
• During detailed survey a final alignment and pegging (location marking of
towers) is carried out to draw complete profiling apart from recording all
features in the corridor.
• Details of existing crops, general crop patterns, seasonal particulars, nature
and extent of yield are compiled and analysed to study the extent and nature
of impact.
• Additionally, details of land owners (name, area impacted etc.) coming under
RoW identified to facilitate assessment/ valuation of compensation during 3
different stages of project activities viz. foundation, erection and stringing.
• Since damages to tree/crop can only be ascertained after actual damage
happened, valuation of compensation and disbursement of payment to land
owners follow thereafter involving several procedural activities.

10. Power Transmission Project : Gaps in Present System
• Survey using topo map, google maps and GIS/GPS etc. provide overall
topographical features but not giving micro level information about type
vegetation/ cropping pattern of the project area.
• Often these secondary tools are not up to date which may result in
mismatching of information at field level.
• Assessment of land use including vegetation/ cropping pattern along the
corridor is done manually through survey which takes substantial time
considering linear nature of the projects.
• Time lag between survey and actual start of construction is quite substantial
resulting in inconsistency in actual ground level condition vis-à-vis
assessment done during survey.
• Often there is cost and time overrun due to inconsistency in assessment
process and subsequent resistance for public demanding enhance of
compensation

11. use of modern techniques such as Artificial Intelligence/ deep learning
based UAV geotagged images classification can be considered for
identification of crops and trees with enhance efficiency & transparency
for faster and more accurate assessment
Way Forward

12. Existing Remote sensing platforms for Earth Observation
Remote Sensing Technologies Advantages Disadvantages
Satellite Very Large area coverage Very low spatial resolution
Could Sensitivity
High Cost
Data not available all time
Manned aircraft Low area coverage High Cost
Low spatial resolution
Affected by weather conditions
Arial Vehicle(Unmanned) Low Cost
High spatial resolution
Data available all the time
Not sensitive to cloud
Accurate position
Difficult area accessibility
Medium area coverage
Affected by weather condition
Controlled by flying laws and regulation
Ground Vehicle(Manned/ Unmanned) Very high spatial resolution Very low area coverage
Require very long time to cover even small
area(s)

13. CNN Architecture for Image Classification

14. UAV Image
Acquisition
Image Pre
Processing
Crop
Dataset
Fine tuned
CNN
Training
Crop CNN
Model
Testing
Image
Crop CNN
Model
Classification
Probabilities
Counting of
Detection
Refined
bounding
boxes
Output
Steps involved in Proposed Work

15. Advantages of UAV Crop Images Classification/ Identification
Parameter Conventional UAV/ Drones
Time Saving It takes months to capture the
data. Cause to time overrun of
the transmission project
Less time for gathering the data. Survey can be completed
within hours and processed within few days. Survey can be
done for multi-season crops- identification for enhanced
accuracy.
Cost Effective Not cost effective The acquisition of data can be done in a timely, accurate
and safe way and thus minimising cost over run.
Accessibility Not easy to access the locations. survey can be done in the areas which are difficult or
impossible to undertake by other means. For example
hazardous terrain, cliffs and valleys with inaccessible steep
slopes, where conventional surveying is challenging and
time consuming and thus minimising time overrun
Safety Traditional survey is the risky job
as it is difficult to explore
unknown terrain / lands and
high risk places such as railway
lines and rocky terrains
Drone survey improves safety. It eliminates the risk
involved while surveying tough/ unknown terrains.

16. Conclusion
• Automatic Crop Classification/ Identification during preliminary Survey
can reduce the overall Project Cost Estimate towards vegetation
evaluation with accuracy- Cost Overrun ↓
• Less Time taking as compared to manual process- Time Overrun ↓
• Less damage to Environment
• Good Social Impact (Transparency in system)
• enhance Integrity in the system (corruption free)