What Is Blob Tracking?

Blob in computer vision refers to a connected region in an image. Blob analysis is to extract and label the connected domain of the binary image after foreground / background separation. Each blob that is labeled represents a foreground target, and then some relevant features of the blob can be calculated. The advantage is that through the Blob extraction, the information of the relevant area can be obtained, but the speed is slow and the analysis is difficult.

Object tracking is an important research task in computer vision and has a wide range of applications. It is to solve the problem of estimating the trajectory of a moving object on the image plane. The tracker needs to assign a unique label to a specific tracking target in different frames of the video sequence. In addition, according to the tracking results, it can also provide information related to moving targets, such as the shape and appearance of the target, the state of movement, and statistical information. ^[1]

The core idea of the Blob algorithm is to find out the range where "gray abrupt changes" occur in a region. Determine its size, shape, and area. The "convolution algorithm" in the algorithm for edge search, for example, the matrix of the "convolution algorithm" is set to 2 * 4 scanning from left to right from top to bottom. Each time the matrix is moved, the algorithm calculates the difference between the sum of the gray values in the first half (columns 1 and 2) of the matrix and the sum of the gray values in the second half (columns 3 and 4) of the matrix. If the color of this image in the matrix is the same, that is to say, the gray values of the 8 pixels in the matrix are very close, then the difference between the sum of the gray values at the front and the back of the matrix should be close to zero. The program runs step by step, and the total difference between the gray values at the front and the back of the matrix is always close to zero. If the matrix scans to the edge of a foreground, the sum of the gray values at the front and the back of the matrix suddenly changes, and the gray values are no longer close, then it is a blob. When the algorithm scan is complete, the system records the coordinates of all points where this sudden change occurred. After a series of analysis of these edge points, we can get the size, shape and area of this blob. ^[2]

An important step in Blob analysis is the determination of connected regions. According to different processing methods, Blob analysis is basically divided into two types: pixel scanning-based processing methods and

Blob disadvantages

The current Blob algorithm has several disadvantages:

1. The speed is too slow. Scan the entire area point by point.

2. Blob analysis is difficult. This is a purely geometric problem. For an irregular shape, there is no simple and easy algorithm to calculate its area and size. It is too complicated, the calculation time is long, and the speed is even slower. In addition, in practical applications, there is no simple surface detection. In a scene, several area blocks are usually detected. Surface detection is only one of them. This makes the limitations of the Blob analysis algorithm more obvious. For example, we do a printed matter inspection: one is to check whether the words printed on the printed matter are clear; the other is to check whether the surface of the printed matter is scratched. Therefore, the Blob algorithm must be able to distinguish which is text and which is scratch.

3. Practical application. The Blob algorithm is very dependent on light sources in practical applications. It can almost be said that if the Blob algorithm leaves a reliable light source design, it will not work at all. There are various colors in the scene. The difference in these colors is the change in gray value when viewed from a black and white camera. Because the performance of colors has a great relationship with light sources, a stable light source is necessary. ^[2]

blob advantages

The advantages of Blob in target tracking are:

(1) Through the blob extraction, the information of the relevant area can be obtained, and this information can be used as supplementary information of the edge detector or corner detector. In target recognition, Blob can provide local statistical information and appearance information, which can provide basis for target recognition and tracking;

(2) Blob can be used for peak detection of the histogram;

(3) Blob can also be used as a primitive for texture analysis and texture recognition;

(4) Through Blob analysis, the number of targets and their area can be obtained. When performing target matching, there is no need to search the global image. ^[1]