Morphological geodesic active contour algorithm for the segmentation of the histogram‐equalized welding bead image edges

Abstract

Assessment and evaluation are the essential processes of industrially manufactured products for the determination of the quality and quantity of products. They give justifications in a practical way about whether the machine is perfect or imperfect, which can lead to a better or poorer production. In this study, the authors propose an algorithm that uses morphological geodesic active contour and image processing techniques to perform segmentation and assess the performance of a robot used to manufacture welding beads. The algorithm has four parameters which are pre-processed images, balloon force, smoothing parameter, and number of iterations. To pre-process the images, the algorithm uses an inverse Gaussian gradient operator for edge detection and applies the histogram equalization method to level the distribution. To detect the external contour of the bead, the level set is initialized as the region of interest whereby a balloon force can inflate or deflate towards the edges. To smoothen the contour, a smoothing parameter is applied to convert the jagged lines into a curve over a reasonable number of iterations. Based on the experimental results, the authors' algorithm used a fixed balloon force of −2, a smoothing parameter value of 4, and 40 iterations to segment images obtained from three different environments. The computation time for the segmentation and evaluation of one image was 0.70, 0.61, and 0.67 s for datasets with high brightness, low brightness, and normal brightness, respectively. Additionally, the authors' proposed algorithm achieved an outstanding performance of 0.9954, 0.9843, 0.9892, and 0.9435 in terms of recall, precision, F-measure, and IOU, respectively. To justify the performance of the authors' proposed algorithm, the authors compared it with the existing algorithms and found that it worked better than all the others for segmentation, although it lagged behind the entropy-based algorithm in terms of speed.

Classification Algorithm of 3D Pattern Film Using the Optimal Widths of a Histogram

“…In addition, several studies were proposed using a morphological snake that performs object detection using morphological operators such as expansion and erosion as a segmentation method [6]. The morphological geodesic active contour method in [7,8] that combines morphological snakes and geodesic active contour in [9] is used to accurately detect patterns by widening the gradually developing contour to a local minimum; this prevents the detection of other reduced minimum contours. Recently, Medeiros et al [8] proposed a Fast Morphological Geodesic Contour (FGAC) method of segmentation without prior training using a new Fuzzy Border Detector.…”

“…Since the imaginary part is nearly zero and an image is always a real-valued function, we utilize the real part of the processed image. We determine the horizontal and vertical index of white pixel values to obtain the horizontal I x (x, y), vertical I y (x, y), as shown in Equations ( 6) and (7).…”

“…To analyze 3D pattern films, a range of inspection methods such as shape, texture, similarity, and contrast, as found in existing algorithms, can be employed [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18]. Segmentation algorithms may be considered for analyzing the shape of 3D pattern films, while edge detection algorithms can be applied to study texture.…”

Classification of a 3D Film Pattern Image Using the Optimal Height of the Histogram for Quality Inspection

“…In addition, the classification accuracy of each image histogram was calculated for different height intervals, ranging from 1/10 to 9/10, as shown in Table 1. The classification of the pattern images was obtained using Equation (7), with β set to 600. Table 1 shows the number of overlapping good and bad pattern images, enabling us to determine the number of incorrectly classified images.…”

“…Analysis of misclassification images and number of overlapping good pattern and bad pattern images, along with classification accuracy, sensitivity, and specificity of 3D film images. To evaluate the performance of the proposed algorithm, we analyzed the results of the existing algorithms using the Abs-based difference method, Otsu thresholding, Canny edge detection, the CNN with Canny [4], morphological geodesic active contour [7], the Michelson contrast [22], the Canny with HSV [3], and the SVM with Canny [14]. For the CNN with Canny method proposed by Mlyahilu et al [4], we used 32 and 64 nodes in the convolution layer, 10 epochs, the ReLu activation function, and Adam for the optimization function, as in the paper.…”