Inspection device for inspecting a building or structure

A method for processing two sets of 3D data captured by one or more depth sensors onboard a drone, each set of 3D data being captured at a respective point in time and comprising information that indicates a distance of the drone from one or more surfaces in the drone's surroundings, the method comprising:

• executing a scan matching algorithm so as to determine a relative shift in position of the drone between capturing the two sets of 3D data, wherein the scan matching algorithm comprises an Normal Distribution Transform (NDT) algorithm and comprises:
• for each set of 3D data:
• generating a respective point cloud from the set of 3D data, wherein the point cloud comprises the coordinates of points on the surfaces as measured by the one or more depth sensors;
• subdividing the 3D field of view of the depth sensor into one or more cells;
• and for each cell:
• identifying the coordinates of points in the point cloud located within the cell;
• determining, based on the identified coordinates, a probability distribution that defines the probability of points in the point cloud being located at different positions throughout the volume of the cell;
• wherein the volume of each cell is set to ensure that the number of points in the point cloud whose coordinates lie within the cell is above a threshold.

A drone for inspecting a building or structure, the drone comprising:

• a propulsion system configured to maneuver the drone relative to the building or structure;
• one or more inspection sensors configured to capture, from each one of a plurality of regions of the building or structure, a set of inspection data for evaluating the condition of the building or structure in the respective regions, the inspection sensor being configured to capture each set of inspection data at a respective point in time as the drone maneuvers relative to the building or structure;
• one or more depth sensors for capturing sets of 3D data of the drone's surroundings at respective points in time, wherein each set of 3D data comprises information indicating a distance of the drone from one or more surfaces of the building or structure;
• a processor configured to combine the sets of 3D data to obtain a 3D map of the surroundings;
• the processor being configured to log the respective points in time at which each set of inspection data is captured, log the respective points in time at which each set of 3D data is captured and for each set of inspection data, compare the time of capture of the respective set of inspection data with the time of capture of different ones of the sets of 3D data, thereby to identify a set of 3D data whose time of capture corresponds to that of the respective set of inspection data, and to use the identified set of 3D data to determine a location in the 3D map where the drone was located when the respective set of inspection data was captured;
• the drone being configured to tag each set of inspection data by associating the respective set of inspection data with the position of the drone in the 3D map at which the set of inspection data was captured, thereby to provide tagged inspection data, whereby the tagged inspection data serves to indicate the location of the drone within the 3D map at the time the respective set of inspection data was captured;
• and
• wherein the processor is configured to generate a 3D view of an environment based on the 3D map, wherein the 3D view of the environment shows the location of the drone in the 3D environment at the time each set of inspection data was captured.

A method for processing data captured by a drone during inspection of a building or structure, the method comprising:

• receiving sets of inspection data captured by one or more inspection sensors onboard the drone, each set of inspection data being captured from a different region of the building or structure, and wherein each set of inspection data is descriptive of a condition of the building or structure in the respective region, each set of inspection data being captured at a respective point in time as the drone maneuvers relative to the building or structure;
• receiving sets of 3D data of the drone's surroundings, each set of 3D data being captured by one or more depth sensors onboard the drone at a respective point in time, wherein each set of 3D data comprises information indicating a distance of the drone from one or more surfaces of the building or structure at the time of capture of the set of 3D data, wherein for each point in time at which a set of 3D data is captured, the set of 3D data collected at that point in time is compared with a preceding set of 3D data using a scan matching algorithm, to determine a relative shift in position of the drone between capturing the two sets of 3D data, wherein determining the position of the drone comprises performing a loop closure, wherein the loop closure comprises:
• selecting a first set of 3D data and a second set of 3D data from the sets of 3D data;
• determining a transform that provides a mapping between the point cloud of the first set of 3D data and a point cloud of the second set of 3D data;
• determining an updated point cloud by applying the transform to the first set of 3D data or the second set of 3D data;
• and
• updating an estimated position of the drone based on the transform;
• wherein updating the estimated position of the drone is carried out subject to
• the condition that the number of points in the updated point cloud that are deemed to coincide with the point cloud of whichever one of the first set of 3D data and the second set of 3D data the transform was not applied to is above a threshold;
• combining the sets of 3D data to obtain a 3D map of the device drone's surroundings when carrying out the inspection;
• tagging each set of inspection data by associating the respective set of inspection data with a position of the drone in the 3D map at which the set of inspection data was captured, thereby to provide tagged inspection data, whereby the tagged inspection data serves to indicate the location of the drone within the 3D map at the time the respective set of inspection data was captured.