Algorithm 1. Automated Scale Bar Detection
Input: grayscale or RGB image P after preprocessing
Output: scale ratio R = Lphys/Lpx, real-world dimensions of image Wreal, Hreal
1:	Image ← convert P to RGB if grayscale;
2:	$(H_{\mathrm{px}}$$, W_{px}$) ← Image.shape;
3:	$\mathrm{ROI} \leftarrow \mathrm{Image}[H_{\mathrm{px}}$//2:, :];//Define ROI as the lower half of P
4:	HSV ← Convert ROI to HSV color space;
5:	mask ← inRange(hsv, lower = (0, 0, 0), upper = (180, 255, 30));//Create black mask using HSV thresholds
6:	contours ← findContours(mask);//Extract contours from the mask
7:	Initialize Lpx ← 0;
8:	for each contour do
9:	edge_img ← Canny(drawFilledContour(mask.shape, contour), 50, 150);//Apply Canny edge detection //Use probabilistic Hough transform to detect lines
10:	lines ← HoughLinesP(edge_img, rho = 1, theta = π/180, threshold = 50,         minLineLength = 50, maxLineGap = 10);
11:	for each detected line (x1, y1), (x2, y2) do
12:	$\mathrm{Compute} \mathrm{length} l=\sqrt{{\left(x_2-x_1\right)}^2+{\left(y_2-y_1\right)}^2}$;
13:	$\mathbf{if} l>L_{\mathrm{px}}$ then
14:	Lpx ← l;//Update Lpx
15:	Lphys ← CnOCR_detect_numeric_value(ROI);
16:	if Lphys = None then
17:	Lphys ← ask_user_input(“Please enter the actual length of scale bar (µm):”);//Prompt user to input physical length Lphys
18:	$R \leftarrow L_{\mathrm{phys}}/L_{\mathrm{px}}$;//Compute scale ratio R
19:	$W_{\mathrm{real}} \leftarrow W_{\mathrm{px}}·R$;//Compute real-world width of image
20:	$H_{\mathrm{real}} \leftarrow H_{\mathrm{px}}·R$;//Compute real-world height of image
21:	Return R, Wreal, Hreal;