Euclidean distances as well as the polarities of the are recorded as R according to Equation (3). In traces in order to judge the similarity of two addition, do the same operation on d ' and d ' descriptors, we can determine whether the two and we can also get R ' . Assume the angle o f descriptors match or not. Firstly, if they do not  P  P  and  P  P  is  as well as the angle of have the same type of contrast, this feature point is P ' P ' is ' . Then the difference value treated as a false one and it will not be taken into be tw een R and R ' and the difference between consideration. Conversely, if they have the same  and  ' accordi n g to (4) are achieved. In case type of contrast, we go on calculating the that the difference absolute value is in the range of Euclidean distance of them respectively according θ and R and R to: same. It is the same 'would be considered to be the with  and  ' . :   For each feature point, compute the ratio of the shortest Euclidean distance and the second shortest one. Next, the ratio is compared with the default There is a proposed criteria that if R is equal threshold value (usually 0.7 in experiment). If the is also equal to ' , P is matched ratio is in the range of 0.7, we think that this point

doi:10.1201/b18660-69

Chapter

Euclidean distances as well as the polarities of the are recorded as R according to Equation (3). In traces in order to judge the similarity of two addition, do the same operation on d ' and d ' descriptors, we can determine whether the two and we can also get R ' . Assume the angle o f descriptors match or not. Firstly, if they do not  P  P  and  P  P  is  as well as the angle of have the same type of contrast, this feature point is P ' P ' is ' . Then the difference value treated as a false one and it will not be taken into be tw een R and R ' and the difference between consideration. Conversely, if they have the same  and  ' accordi n g to (4) are achieved. In case type of contrast, we go on calculating the that the difference absolute value is in the range of Euclidean distance of them respectively according θ and R and R to: same. It is the same 'would be considered to be the with  and  ' .

Chapter

Euclidean distances as well as the polarities of the are recorded as R according to Equation (3). In traces in order to judge the similarity of two addition, do the same operation on d ' and d ' descriptors, we can determine whether the two and we can also get R ' . Assume the angle o f descriptors match or not. Firstly, if they do not  P  P  and  P  P  is  as well as the angle of have the same type of contrast, this feature point is P ' P ' is ' . Then the difference value treated as a false one and it will not be taken into be tw een R and R ' and the difference between consideration. Conversely, if they have the same  and  ' accordi n g to (4) are achieved. In case type of contrast, we go on calculating the that the difference absolute value is in the range of Euclidean distance of them respectively according θ and R and R to: same. It is the same 'would be considered to be the with  and  ' .

ABSTRACT

In order to verify the robustness of the improved algorithm against the changes of illumination, background, angle, and scale, letter “V” is taken as an example, and experiments are conducted in the cases of strong and weak illumination, complex background, angle, and scale variations. As shown in Figure 5, they are RGB image, depth image, and result image respectively in 5 cases above. The experiment results show that the method proposed can accurately acquire good recognition results in different environmental cases. As is written before, fusion method of

hand gesture segmentation is adopted to overcome the influences on the changes of illumination and complex background. In addition, feature extraction based on SURF has strong robustness against angle and scale variations.