Organizes post-flight video and flight logs by capture position, and runs data quality inspection and object detection.
DX-Sky AI is post-flight analysis software that takes the video and flight logs (position, attitude and gimbal) from a completed drone flight and prepares them for analysis. Each frame extracted from the video is linked to its capture time, position, attitude and gimbal data, and target objects found in captured frames and orthomosaics are organized together with their capture positions.
Data quality inspection runs in the same process. Segments where the position could not be recovered or where blur is suspected are compiled into re-capture areas, and the decision to re-capture is made after operator approval.
Video and flight logs are collected, checked for position anomalies and linked to capture data frame by frame. Object detection and quality assessment then run on the prepared data.
Collects video together with position, attitude and gimbal records. Both aircraft flight logs and our own collection format are supported, so data can be loaded without separate conversion.
Detects and flags position jumps, dropouts and drift, as well as altitude and timestamp anomalies. Short gaps are interpolated from the surrounding records.
Positions that pass the quality criteria are selected in a fixed order of priority. Real-time corrected positions (RTK) are used first, post-processed positions (PPK) when RTK is unavailable, and standard GPS positions when neither is available. Each selected position is marked with its confidence level.
Extracts frames from the video and links each one to its capture time, position, attitude and gimbal data in a single table. Frames with suspected blur are flagged, and the final decision is made after operator review.
Detects target objects in captured frames and orthomosaics and organizes the results with their capture positions. Orthomosaics produced separately are taken as input for analysis.
Assesses data quality for each flight and records the assessment results and run history every time it runs.
Each check item is assessed for fitness for analysis, and segments with confirmed problems are compiled into re-capture areas.
Each check item is assessed as normal, abnormal or undetermined. Undetermined items are recorded with the reason, making clear what needs further review.
Segments where the position could not be recovered and segments where the operator confirmed blur are grouped and exported as a re-capture coordinate list. Each coordinate specifies its coordinate system and altitude reference, and invalid coordinates are excluded before export.
Proposed actions such as re-capture proceed only after operator approval. Re-captured data is inspected again, and every step from proposal to re-check is recorded.
Because analysis results are the basis for deciding whether to fly again, each result is managed so that its source can be traced.