Feature selection from sparse and high dimension features using conventional greedy based boosting gives classifiers of poor generalization. We propose a novel “shrink boost” method to address this problem. It solves a sparse regularization problem with two iterative steps. First, a “boosting” step uses weighted training samples to learn a full high dimensional classifier on all features. This avoids over fitting to few features and improves generalization. Next, a “shrinkage” step shrinks least discriminative classifier dimension to zero to remove the redundant features. In our object detection system, we use “shrink boost” to select sparse features from histograms of local binary pattern (LBP) of multiple quantization and image channels to learn classifier of additive lookup tables (LUT). Our evaluation shows that our classifier has much better generalization than those from greedy based boosting and those from SVM methods, even under limited number of train samples. On public dataset of human detection and pedestrian detection, we achieve better performance than state of the arts. On our more challenging dataset of bird detection, we show promising results.

In this paper, an accurate long-distance position measurement system using laser range finders (LRFs), which can be used for surveys in construction fields is proposed. Since the LRF is a sensor which can measure distance to surfaces of objects by radiating laser beams from itself and receiving the reflected ones, data obtained from the LRF are nothing more than the contours of objects. For this reason, we adopted cylindrical shaped objects since the contour, a circular arc, is invariant against rotation. Therefore, this research aims to fit circles to the arc-shaped contours of the cylindrical objects and estimate their accurate center positions by applying the least square method and maximum likelihood estimation. If we know the radius of the cylindrical object in advance, the aforementioned two methods become non-linear problems. For this reason, we applied the Newton-Raphson method to solve these non-linear equations. We improve the angular resolution of the LRF by using a pan unit, and reveal that the maximum likelihood estimation can give us the most accurate center position. Additionally, we implemented proposed position measurement system in an actual construction field.