The target detection method was first proposed by PaulViola[Viola0 1] and improved by RainerLienhart[Lienhart02]. The basic steps of this method are as follows: firstly, the harr features of samples (about hundreds of sample pictures) are used for classifier training to get the cascade lifting classifier. "Cascade" in the classifier means that the final classifier is composed of several simple classifiers. In image detection, the detection window passes through each classifier in turn, thus excluding most candidate regions in the first few layers, and the regions detected by each classifier are all target regions. After the classifier is trained, it can be used to detect the region of interest in the input image (the same size as the training sample). The output of the classifier that detects the target area (car or face) is 1, otherwise the output is 0. In order to detect the whole image, we can move the search window in the image and detect each position to determine the possible target. In order to search for target objects with different sizes, the classifier is designed to change the size, which is more effective than changing the size of the image to be detected. Therefore, in order to detect the target object with unknown size in the image, the scanning program usually needs to scan the image several times with search windows of different scales.