In the previous chapter, we learned about processing videos and how to perform the operations and algorithms from all of the previous chapters on frames read from cameras or video files. We learned that each video frame can be treated as an individual image, so we can easily use algorithms, such as filtering, on videos in almost the same way as we did with images. After learning how to process videos using algorithms that work on single individual frames, we moved on to learn about video processing algorithms that require a set of consecutive video frames to perform object detection, tracking, and so on. We learned about how to use the magic of the Kalman filter to improve object-tracking results, and ended the chapter by learning about background and foreground extraction.

The object detection (and tracking) algorithms that we learned about in the previous chapter rely heavily on the color of an object, which has proven not to be too reliable, especially if the object and the environment we are working with are not controlled in terms of lighting. We all know that the brightness and color of an object can easily (and sometimes extremely) change under sunlight and moonlight, or if a light of a different color is near the object, such as a red traffic light. These difficulties are the reason why the detection of objects is more reliable when their physical shape and features are used as a basis for object detection algorithms. Obviously, the shape of an image is independent of its color. A circular object will remain circular during the day or night, so an algorithm that is capable of extracting the shape of such an object would be more reliable to be used for detecting that object.

In this chapter, we're going to learn about computer vision algorithms, functions, and classes that can be used to detect and recognize objects using their features. We'll learn about a number of algorithms that can be used for shape extraction and analysis, and then we'll proceed to learning about key-point detection and descriptor-extraction algorithms. We'll also learn how to match descriptors from two images to detect objects of known shapes in an image. In addition to the topics that we just mentioned, this chapter will also include the required functions for proper visualization of key points and matching results.

In this chapter, you'll learn about the following:

• Template matching for object detection
• Detecting contours and using them for shape analysis
• Calculating and analyzing contours
• Extracting lines and circles using the Hough transformation
• Detecting, descripting, and matching features