In order to solve the problem that it is difficult to leverage the performances of effect and efficiency, an image matching algorithm based on the Histogram of Gradient Angle (HGA) was proposed. After obtaining the key points by Features from Accelerated Segment Test (FAST), the block gradient and the new structure as dartboards were introduced to descript the local structure feature. The image matching algorithm based on HGA can work against the rotation, blur and luminance and overcome the affine partly. The experimental results, compared with Speeded Up Robust Feature (SURF), Scale Invariant Feature Transform (SIFT) and ORB (Oriented FAST and Rotated Binary Robust Independent Elementary Features (BRIEF)) in the complex scenes, demonstrate that the performance of HGA is better than other descriptors. Additionally, HGA achieves an accuracy of over 94.5% with only 1/3 of the time consumption of SIFT.
As an important component of the surface meteorological observation, the daily observation of surface frost still relies on manual labor. Therefore, a new method for detecting frost based on computer vision was proposed. First, a k-nearest neighbor graph model was constructed by incorporating the manually labeled frosty image samples and the test samples which were acquired during the real-time detection. Second, the candidate frosty regions were extracted by rating those test samples using a graph-based manifold learning procedure which took the aforementioned frosty samples as the query nodes. Finally, those candidate frosty regions were identified by an on-line trained classifier based on Support Vector Machine (SVM). Some experiments were conducted in a standardized weather station and the manual observation was taken as the baseline. The experimental results demonstrate that the proposed method achieves an accuracy of 87% in frost detection and has a potential applicability in the operational surface observation.
It is very important to ascertain rationally the number and positions of split points for discretization of continuous variables. To improve the efficiency of unsupervised discretization, an entropy-based algorithm was proposed for discretization of continuous variables. It made use of the characteristics of the information content(entropy) of a continuous variable, and partitioned the continuous variable by itself for minimizing both the loss of entropy and the number of partitions, in order to find the best balance between the information loss and a low number of partitions, so then obtained an optimal discretization result. The experiments show this approach effective.