The Orientation Angle Detection of Multiple Insulators in Aerial Image - Advances in Image and Graphics Technologies - page 295

Literature [5] develops an effective algorithm for human face orientation detection
applying the symmetry features of the human face. It is effective and promising for
human face orientation detection, but, it may be not fit for generic images without
symmetry. Literature [6,7] can solve the identification of the orientation of sample
polygons in bizarre circumstance effectively. This method has a good stability and
generality, and high robust for the identification of sample polygon in all cases, but
the accuracy would be reduced significantly for image with curve edges and compli-
cated background. Literature [8] proposes a recognition method to deal with multi-
view targets in infrared images synthetically utilizing image invariant moments
feature and SVM (Support Vector Machine) classification method. But this method
can solve the orientation detection of targets without background, and the division of
view-Angle coverage increases the error in the determination of orientation angle, and
also the complexity of SVM classification method would bring large calculation. In
literature [9], a novel skew correction algorithm is proposed focusing on boundary
line that optimizes speed and accuracy by Hough transform to get the skew corrected
license plate image. But this method achieves high accuracy of orientation angle de-
tection at the expense of large calculation, and for images with complex background,
the calculation would be increasing when the target area is small which causes bad
angle detection. Literature [10] develops a probabilistic approach to image orientation
detection by confidence-Based integration of low-Level and semantic cues within a
Bayesian framework. The computation is increasing caused by Bayesian, especially in
complex background. Literature [11] presents a document skew and orientation detec-
tion technique by white run histograms through scanning documents in horizontal and
vertical directions. This method is capable of detecting arbitrary skew orientation of
documents with high speed and low computation. Literature [12] introduces a generic,
scale-Independent algorithm which is capable of accurately detecting the global skew
angle of document images within the range

[ °

180- . Despite its generality, the
method is very fast and requires no explicit parameters with high accuracy and ro-
bustness. But literature [11,12] would achieve high accuracy because of a large
amount of equidistant interline spacings leading to significant limitations, and would
fail on complex images containing almost exclusively majuscules, digits and mathe-
matical formulae. Literature [13] presents a recognition driven method for page orien-
tation detection. A small subset of text lines are extracted in the four orientations, and
then the outputs of the OCR (Optical Character Recognition) are evaluated to deduce
the right orientation. The method can handle documents containing small lines of text,
and it is able to detect multiple orientations. But it may not realize the detection of
arbitrary angle. Literature [14] proposes a script identification method that works for
unknown orientation for all official Indian scripts. It has high accuracy, but classifier
may result in large computation. Literature [15] uses PCA (Principal Component
Analysis) to realize insulators' tilt correction, and it has a high accuracy for insulator
images without background, but it may be not effective for images with complex
background.

° , 180