Shape features of image features

2. shape description features

After an object is separated from an image, the combination of shape description features and size measurement can serve as a basis for distinguishing different objects and plays an important role in the machine vision system.

·Length and width

It is the simplest method to portray the basic shape of an object by the size of its external rectangle when the boundary is known. If you only calculate the external rectangle in the coordinate system direction, you only need to calculate the maximum and minimum coordinate values of the boundary points of the object to obtain the horizontal and vertical spans of the object. However, it is usually necessary to calculate the length of the spindle direction that reflects the shape and features of the object and the width in the vertical direction. Such an external Rectangle is the smallest external Rectangle (MER-Minimum Enclosing Rectangle) of the object ).

One Method for Calculating MER is to rotate an object at intervals of 90 degrees, and record the external rectangular parameters in the direction of the coordinate system each time, take the rectangle with the smallest area as the length and width of the spindle. Generally, the spindle can be obtained through the calculation of the moment (moments), or by the method of finding the best fit line of the object.

　

·Rectangle

The rectangle is characterized by the ratio of the area of the object to the area of the smallest external rectangle, reflecting the degree to which the object is filled with the external rectangle:

 

·Circular Degree

The circular degree is used to describe the complexity of the Object Boundary. They take the minimum value when the circular boundary is reached. The most common circle degree is the ratio of the circumference square to the area.

 

·Moment unchanged

Moment definition: For binary Bounded Functions f (x, y), its (j + k) moment order is:

 

A function has a one-to-one correspondence with its moment set.

In order to describe the shape, assume that f (x, y) has a value of 1 in the object and 0 out of it, so that it has a one-to-one correspondence with the outline of the object, its Moments reflect the contour information of objects.

The center moment has positional independence.

ObjectMain AxisThe direction can be obtained using the following formula:

 

Eccentric heart rate is defined:

 

For normalized center moment, there are seven invariant moment combinations [1], which are unchanged for translation, rotation, and scale changes:

 

However, due to noise and other interference factors in the image, the above-mentioned invariant moment, especially the higher-order invariant moment, is unstable. It is unreliable to use the invariant moment feature only to identify objects.

·Fourier Descriptor of Contour

Set P to any point on the boundary contour. Take point A on the boundary contour as the reference point, record s as the arc length from A to point P, and set the perimeter of the boundary contour to S, the P point can be expressed as a function of the arc length.

 

In this way, the boundary contour of an object corresponds to the periodic function, so we can use its Fourier transform coefficient to characterize its contour features. Because the mold of the Fourier transform coefficient has translation and rotation immutability, the coefficient vectors of the Fourier transform can be used as features to identify objects. During implementation, it is usually necessary to normalize its amplitude, for example, dividing by the maximum amplitude or average amplitude to obtain Shape Recognition features with irrelevant scales.