RandomBoost: simplified multiclass boosting through randomization

Cascade Learning by Optimally Partitioning

Cascaded AdaBoost classifier is a well-known efficient object detection algorithm. The cascade structure has many parameters to be determined. Most of existing cascade learning algorithms are designed by assigning detection rate and false positive rate to each stage either dynamically or statically. Their objective functions are not directly related to minimum computation cost. These algorithms are not guaranteed to have optimal solution in the sense of minimizing computation cost. On the assumption that a strong classifier is given, in this paper, we propose an optimal cascade learning algorithm (iCascade) which iteratively partitions the strong classifiers into two parts until predefined number of stages are generated. iCascade searches the optimal partition point of each stage by directly minimizing the computation cost of the cascade. Theorems are provided to guarantee the existence of the unique optimal solution. Theorems are also given for the proposed efficient algorithm of searching optimal parameters . Once a new stage is added, the parameter for each stage decreases gradually as iteration proceeds, which we call decreasing phenomenon. Moreover, with the goal of minimizing computation cost, we develop an effective algorithm for setting the optimal threshold of each stage. In addition, we prove in theory why more new weak classifiers in the current stage are required compared to that of the previous stage. Experimental results on face detection and pedestrian detection demonstrate the effectiveness and efficiency of the proposed algorithm.

A New Waters Hole Detection and Tracking Method for UGV in Cross-Country Environment

Environment perception is one of the important issues for unmanned ground vehicle (UGV). It is necessary to develop waters hole detection and tracking method in cross-country environment. This paper is related to the waters hole detection and tracking by using visual information. Image processing strategies based on support vector machine (SVM) and speeded up robust feature (SURF) methods are employed to detect and track waters hole. It focuses on how to extract the waters feature descriptor by exploring the machine learning algorithm. Based on the S/V color features and Gray Level Co-occurrence Matrix, the waters feature descriptor is extracted. The radial basis function (RBF) kernel function and the sampling-window size are determined by using the SVM classifier. The optimal parameters are obtained under the cross-validation conditions by the grid method. In terms of waters tracking, SURF feature matching method is applied to extract the remarkable feature points, then to observe the relation between feature point movement of adjacent frames and scale change ratio. Experiments show that SURF algorithm can still be effective to detect and match the remarkable feature points, against the negative effects of waters scale transformation and affine transform. The conclusion is that the computing speed of SURF algorithm is about three times faster than that of scale-invariant feature transform (SIFT) algorithm, and the comprehensive performance of SURF algorithm is better.