![\"\"](\"https:\/\/mscvprojects.ri.cmu.edu\/2025team5\/wp-content\/uploads\/sites\/132\/2025\/12\/image-1-1024x683.jpeg\")
<\/figure>\n\n\n\nThe aircraft fuselage we experiment on<\/p>\n\n\n\n
Rig Setup<\/h2>\n\n\n\n
To set up the two-camera stereo system, we built a rig that holds both cameras and the laser (see images below). This ensures that the cameras and laser share a common baseline and maintain fixed relative positions. Note that this configuration is intended only as a convenient test platform for evaluating our pipeline. For practical deployment, the system could be mounted on a drone carrying the cameras and the laser.<\/p>\n\n\n\n Rig setup that holds cameras and the laser<\/p>\n\n\n\n The results of the first-stage pipeline are shown below. For images captured at a distance of 1.5 m, we achieve a recall of 93.6%. However, when the camera is moved farther from the aircraft\u2014approximately 2.5 m\u2014the recall drops sharply to 76.2%. This decline occurs because, at greater distances, many laser-line deformations fall below 1 px and become difficult for computer vision algorithms to detect. It is also worth noting that the precision at this stage is low: various artifacts such as screws, holes, and rivets can also distort the laser line, leading to numerous false positives. These are addressed in the second-stage pipeline.<\/p>\n\n\n\n Images taken with the cameras. The under-exposed image (right) is the input to the first stage pipeline. <\/p>\n\n\n\n Result of the first-stage pipeline<\/p>\n\n\n\n For the second stage, we use the patches extracted in the first stage as our dataset. We evaluate the performance of the ML classifier using 5-fold cross-validation. The resulting confusion matrix is shown below, with the classifier achieving an accuracy of 89.6%. Note that this accuracy is somewhat limited by the relatively small size of our dataset (approximately 1,000 patches). We expect that expanding the dataset would substantially improve the classifier\u2019s performance.<\/p>\n\n\n Input patches to the second stage pipeline<\/p>\n\n\n\n Confusion Matrix of the Second Stage ML classifier<\/p>\n\n\n\n The table below shows the performance of the whole pipeline. The 2nd stage pipeline will improve the precision by ~24%<\/strong> with the cost of slightly lowering the recall rate by ~8%.<\/strong><\/p>\n\n\n\n![\"\"](\"https:\/\/mscvprojects.ri.cmu.edu\/2025team5\/wp-content\/uploads\/sites\/132\/2025\/12\/image-6.png\")
<\/figure>\n\n\n\nFirst Stage Results<\/h2>\n\n\n\n
![\"\"](\"https:\/\/mscvprojects.ri.cmu.edu\/2025team5\/wp-content\/uploads\/sites\/132\/2025\/12\/image-7-1024x342.png\")
<\/figure>\n\n\n\nDistance<\/td> TP<\/td> FN<\/td> FP<\/td> Precision<\/td> Recall<\/strong><\/td><\/tr> 1.5m<\/td> 366<\/td> 25<\/td> 131<\/td> 73.6%<\/td> 93.6%<\/strong><\/td><\/tr> 2.5m<\/td> 298<\/td> 93<\/td> 134<\/td> 69.0%<\/td> 76.2%<\/strong><\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n Second Stage Results<\/h2>\n\n\n\n
![\"\"](\"https:\/\/mscvprojects.ri.cmu.edu\/2025team5\/wp-content\/uploads\/sites\/132\/2025\/12\/image-8-1024x921.png\")
<\/figure>\n<\/div>\n\n\n<\/td> Predicted Negative<\/td> Predicted Positive<\/td><\/tr> Actual Negative<\/td> 0.89<\/td> 0.10<\/td><\/tr> Actual Positive<\/td> 0.11<\/td> 0.90<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n Final Results<\/h2>\n\n\n\n