(a) (b)
Figure 6: Grayscale Histogram of Absolute Value Difference Grayscale Map, Grayscale Histogram of Difference Grayscale Map
Use the Otsu algorithm to segment two grayscale images. Calculate the maximum grayscale level K according to formula (2), which is the threshold of Otsu.
In the formula, when the grayscale level is set as the threshold, the probability of a pixel being classified as A is , is the grayscale value of the entire image, and is the variance.
The two-dimensional grayscale histogram in Figure 7a presents a multimodal distribution, and the Otsu algorithm cannot find the optimal threshold in the multimodal distribution. Therefore, many backgrounds are segmented, resulting in poor segmentation performance. Figure 7b is a binary image for constructing a differential gray-scale image, highlighting the gray-scale characteristics of the target area, so that the gray-scale distribution of the target and the background presents a double peak feature, and the segmentation effect is good. To remove the influence of paint peeling on the wall surface of the crane’s grab boom, this article adopts morphological opening operation to eliminate white impurities in the black area, and then performs morphological closing operation to eliminate black pixels in the white area, as shown in Figure 7c.