-
原文链接 halcon帮助文档
-
What is classification?
Classifying an object means to assign an object to one of several available classes. When working with images, the objects usually are pixels or regions. Objects are described by features, which comprise, e.g., the color or texture for pixel objects, and the size or specific shape features for region objects. To assign an object to a specific class,the individual class boundaries have to be known. These are built in most cases by a training using the features of sample objects for which the classes are known. Then, when classifying an unknown object, the class with the largest correspondence between the feature values used for its training and the feature values of the unknown object is returned.
classification把事物归为已有的种类。在机器视觉中,往往根据像素和区域的特征分类,而这建立在已知种类间边界的前提之下。
-
What can you do with classification?
Classification is reasonable in all cases where objects have similarities, but within unknown variations. If you search for objects of a certain fixed shape, and the points of a found contour may not deviate from this shape more than a small defined distance, a template matching will be faster and easier to apply. But if the shapes of your objects are similar, but you can not define exactly what the similarities are and what distinguishes these objects from other objects in the image, you can show a classifier some samples of known objects (with a set of features that you roughly imagine to describe the characteristics of the object types) and let the classifier find the rules to distinguish between the object types. Classification can be used for a lot of different tasks. You can use classification, e.g., for
classification尤其适用于那些具有相似特征又有所区别的物体。在图像处理中,有固定形状、轮廓的使用模板匹配会更加适合、高效。
• image segmentation, i.e., you segment images into regions of similar color or texture,
• object recognition, i.e., you find objects of a specific type within a set of different object types,
• quality control, i.e., you decide if objects are good or bad,
• novelty detection, i.e., you detect changes or defects of objects, or
• optical character recognition (OCR).
classification可用于图像分割、物体识别、品质管控、异常(changes)检测、光学字符识别
使用classifier的步骤:
-
Step 1:create classifier创建
-
Step 2: Add training samples to the classifier添加样例
-
Step 3: Train the classifier训练
-
Step 4: Classify new objects分类未知的物