Automatic detection of bends in louvers using range imaging
Shah, Utpal D.
MetadataShow full item record
Shutters have become an important part of home decor. Quality of shutters depends on how much light it lets through after it is closed. Shutters let light through because the louvers used to construct them are not straight. To this end, it is important to detect bends in louvers before they are used in constmction of shutters. Current methods of detecting bends are subjective and inadequate. So, we have proposed a laser-based imaging system that utilizes a smart camera to acquire 3-D range images of the louver samples. In this system, we use range images to detect the 'dome points,' which follows the shape of the louver accurately, using a peak detection algorithm. The system also incorporates methods based on polynomial curve-fitting and scale-space filtering that ultimately detects the location of the bends. First, we fit a polynomial curve to reduce the noise in the dome points, followed by scale-space filtering to locate the extrema of the curve. Finally, we relate the extrema of the curve to the locations of the bends in the louver. We tested our method on 39 samples of louvers with varied lengths and compared the results with the results collected from ground tmth. Results are comparable as they resulted in high correlation coefficient between the two methods.