Gains:
- Ability to set up and run workflow automation with AI support with PyQGIS and model designer
- Ability to manage personal data privacy, copyright, license and engineer liability limits in an end-to-end project
- Ability to design a geomatics project in an AI-enabled but verification-centric flow from data collection to delivery
Module Exam
1. Which of the following is the basic principle that determines the risk of an AI output in survey engineering?
- A) The risk of an output is equal to the harm it would cause if that output were faulty; validation scales accordingly ✔
- B) AI output is generally safe as it is written fluently and confidently
- C) Validation is unnecessary when the most current model is used
- D) If the output refers to a standard, no additional control is required
Explanation: The risk of an output is equal to the harm it will cause if that output is faulty. While a minor error in a literature summary is largely harmless, an error in a cadastral boundary or elevation value can lead to property infringement, improper construction application and property damage; so verification intensity scales with potential harm.
2. Which task is most safely and appropriately assigned to AI in quality control of a GNSS measurement set?
- A) To finally accept the accuracy of the point coordinates and declare them ready for registration
- B) Editing raw observation tables and marking possible outliers/inconsistencies ✔
- C) Confirming application values directly without looking at measurement principles
- D) Considering the result valid without calculating any closing errors
Explanation: AI's safest contribution is ancillary tasks that the engineer can control, such as organizing scattered raw data and annotations, flagging inconsistencies and outliers. The final acceptance of the accuracy of the point coordinate remains with the engineer's approval, including closing errors and measurement principles; AI saying 'the point is absolutely correct' is not evidence.
3. Which information must be clearly stated to the AI in the coordinate transformation request?
- A) Names of points only
- B) Just the name of the target projection, datum is unimportant
- C) Source and target coordinate system (datum + projection/EPSG) and transformation parameters ✔
- D) Name of the person making the measurement
Explanation: Coordinate transformation only makes sense if the source system, target system (datum + projection, e.g. EPSG code) and the required transformation parameters are specified. Just saying 'convert this to UTM' leaves it unclear which datum is being moved from to which datum and may cause a datum shift of meters.
4. Which of the following is the most practical way to verify the output of a transformation?
- A) Looking at the number of decimal places in the output
- B) Looking at how quickly the model gives the answer
- C) Transferring the result to CAD without checking it at all
- D) Putting the control point whose coordinates are known through the same transformation and comparing it with the expected value ✔
Explanation: Passing one or several control points whose coordinates are known through the same transformation and comparing them with the expected value quickly reveals both parameter and datum errors. Systematic difference of tens of meters is typically the wrong datum/parameter sign. The fact that the output appears smooth is not proof of correctness.
5. What is the most critical prerequisite for correct buffer and area calculations in GIS?
- A) The layer is in the appropriate metric (projection) coordinate system ✔
- B) Color of the layer
- C) Column order in the attribute table
- D) The file name of the layer should be short
Explanation: Distance and area calculations are only meaningful in a metric (projected) coordinate system. Direct buffering in geographic coordinates (latitude/longitude in degrees) gives inaccurate results due to degree-meter confusion. Therefore, the layer must be placed in the appropriate projection before analysis.
6. What is the standard way to measure the accuracy of a land cover map in satellite image classification?
- A) Looking at how vibrant the colors of the map are
- B) Establishing an error matrix with independent ground truth samples and calculating overall accuracy/Kappa ✔
- C) Relying on the large number of classes
- D) Looking at the resolution of the image and assuming accuracy
Description: Classification accuracy is evaluated by establishing a confusion matrix with independent ground truth samples and reporting it with metrics such as overall accuracy and Kappa. The 'good looking' of the map or the vividness of class colors is not evidence of accuracy.
7. What is the basic element that ensures positional accuracy in photogrammetric orthophoto and DSM production?
- A) Just take lots of photos
- B) UAV is an expensive model
- C) Accuracy control with accurately measured ground control points and independent control points ✔
- D) The flight should be made on a cloudy day
Description: Ground control points (GCP) connect the photogrammetric model to the real-world coordinate system and determine scale/location accuracy. Additionally, accuracy is tested by calculating RMSE with independent control points that do not enter the model. Simply increasing the number of images or making the UAV expensive does not guarantee accuracy.
8. Which step is mandatory to generate digital terrain model (DTM) from LiDAR point cloud?
- A) Turning all points into a surface as it is
- B) Using only the highest points
- C) Deleting density values
- D) Separating non-ground (building, tree) points with ground filtering ✔
Description: DTM represents bare ground surface; Therefore, non-ground points such as buildings, trees, vehicles need to be separated (ground filtering/classification). If this step is skipped, the top surface (DSM) is obtained, not the DTM. AI can speed up classification, but the result must be validated with profile cross-sections.
9. What is one of the most common and riskiest mistakes in automatic map generation with AI?
- A) Skipping mandatory elements such as scale, legend, coordinate system and accuracy rating ✔
- B) Using too many checkpoints
- C) Grouping layers
- D) Adding a title to the sheet
Explanation: Automatic symbolization and labeling is fast, but if essential cartographic elements such as scale, legend, north arrow, coordinate system and accuracy note are missing, the map becomes misleading and officially unusable. Visual appeal is no substitute for these essential elements.
10. What should be the status of AI output when working with cadastral and land registry data?
- A) It is considered as definitive data that can be directly recorded in the title deed.
- B) Taken as a preliminary check/draft; Final validity depends on measurement, legislation and authority approval ✔
- C) Solve property disputes alone
- D) It becomes an official document without the need for any control.
Explanation: Cadastral border, area and ownership information has legal consequences; AI produces the most preflights, inconsistency scans, and drafts. Final validity depends on measurement, legislation and authorized engineer/institution approval. Declaring AI output as 'final' creates legal and professional risk.
11. What is the most important check to make before using an AI-generated GeoPandas/PyProj script?
- A) Assuming it is correct because the code is long
- B) Considering that it is enough for it to work without errors
- C) Run it with small, known test data and verify CRS and rank ✔
- D) Approving variable names as nice
Explanation: The produced code should be run with a small test data whose result is known and the expected output and rank should be checked. In particular, whether the CRS (coordinate system) definition is assigned correctly is a critical error; false CRS silently produces meters of drift. Just because the code 'works' doesn't mean it's correct.
12. What is the correct approach when giving a cadastral/address dataset containing location data and personal information to AI?
- A) Send all real owner and identity information as is
- B) Ignoring privacy because it would be beneficial
- C) Pasting data into a public forum
- D) Anonymizing personal data or working with representative values/approved environment ✔
Explanation: Data containing personal data (owner name, TR ID, sensitive location) should not be sent to an external model in violation of corporate policy and legislation. Data should be anonymized, sample/representative values should be used, or an approved/in-house environment should be used. The justification 'useful anyway' does not eliminate the obligation of confidentiality.
13. What is the main purpose of using the NDVI index in a land cover classification?
- A) Measuring vegetation viability/density from near infrared and red bands ✔
- B) Measuring the height of the land in meters
- C) Determining the parcel owner
- D) Transforming the coordinate system
Description: NDVI (Normalized Difference Vegetation Index) measures the vigor/density of green vegetation using the difference of infrared and red bands. Vegetation is a powerful marker in distinguishing between water and bare ground. Height alone does not measure or determine ownership.
14. Which statement is most accurate for the role of AI output in a safety-critical application or cadastral job?
- A) If the AI output is fluent enough, there is no need for engineer approval
- B) AI is an accelerator; Does not replace competent engineer approval and responsibility ✔
- C) Responsibility passes to the company that produces the model
- D) No approval required when using the most expensive model
Statement: In engineering and map production, AI output does not replace the control and approval of a competent expert in work that has security and legal consequences; It is merely an accelerator and draft generating tool. Final responsibility and signature remain with the authorized engineer.