demo2

@@ -22,6 +22,7 @@ class OcrController extends Controller
             'customer_name_xy' => 'required|string',
         ]);
         $dataDetail = $request->fields ?? [];
+        $tableColumns = $request->table_columns ?? [];
         try {
             $masterTemplate = MstTemplate::updateOrCreate(
                 ['tpl_name' => $request->template_name],
@@ -49,6 +50,24 @@ class OcrController extends Controller
                 );
 
             }
+
+            // Lưu mapping cột bảng (lưu col index vào field_xy với field_name đặc biệt)
+            if (!empty($tableColumns) && is_array($tableColumns)) {
+                foreach ($tableColumns as $name => $colIdx) {
+                    if ($colIdx === null || $colIdx === '' || $colIdx === false) {
+                        continue;
+                    }
+                    DtTemplate::updateOrInsert(
+                        [
+                            'tpl_id'     => $masterTemplate->id,
+                            'field_name' => '__table_col__' . $name,
+                        ],
+                        [
+                            'field_xy' => (string) $colIdx,
+                        ]
+                    );
+                }
+            }
             return response()->json([
                 'success' => true,
                 'message' => 'Lưu template thành công',
@@ -70,8 +89,8 @@ class OcrController extends Controller
             $templateName = $request->get('template_name', '');
 
             // Giả sử file OCR JSON & ảnh nằm trong storage/app/public/image/
-            $jsonPath = public_path("image/3_1757295841_with_table.json");
-            $imgPath = ("image/3_1757295841.jpg");
+            $jsonPath = public_path("image/nemo_new_1757393338_with_table.json");
+            $imgPath = ("image/nemo_new_1757393338.jpg");
 
             if (!file_exists($jsonPath)) {
                 return response()->json(['error' => 'File OCR JSON không tìm thấy'], 404);
@@ -100,7 +119,13 @@ class OcrController extends Controller
                     // Lấy detail của template
                     $details = DtTemplate::where('tpl_id', $mst->id)->get();
 
+                    $tableColumnMapping = [];
                     foreach ($details as $detail) {
+                        if (strpos($detail->field_name, '__table_col__') === 0) {
+                            $name = substr($detail->field_name, strlen('__table_col__'));
+                            $tableColumnMapping[$name] = is_numeric($detail->field_xy) ? intval($detail->field_xy) : null;
+                            continue;
+                        }
                         $coords = array_map('intval', explode(',', $detail->field_xy));
                         // coords = [x1, y1, x2, y2]
 
@@ -130,6 +155,7 @@ class OcrController extends Controller
                 'pdfImageUrl' => $imgPath,
                 'dataMapping' => $dataMapping,
                 'is_template' => $is_template,
+                'tableColumnMapping' => $tableColumnMapping ?? [], //?? new \stdClass(),
                 'fieldOptions' => [
                     [ 'value' => 'template_name',  'label' => 'Tên Mẫu PDF' ],
                     [ 'value' => 'customer_name',  'label' => 'Tên khách hàng' ],

@@ -11,14 +11,14 @@ from PIL import Image, ImageEnhance
 
 # ==== Config ====
 BASE_DIR = os.path.abspath(os.path.join(os.path.dirname(__file__), "..", "..", ".."))
-# PDF_NAME = 'aaaa'
+PDF_NAME = 'nemo_new'
 
 # PDF path
-pdf_path = Path(BASE_DIR) / "storage" / "pdf" / "3.pdf"
+pdf_path = Path(BASE_DIR) / "storage" / "pdf" / "2.pdf"
 # Output folder
 output_folder = Path(BASE_DIR) / "public" / "image"
 
-PDF_NAME = pdf_path.stem  # Get the stem of the PDF file
+# PDF_NAME = pdf_path.stem  # Get the stem of the PDF file
 #print(PDF_NAME)
 
 os.makedirs(output_folder, exist_ok=True)
@@ -151,9 +151,31 @@ for table in table_info:
             x1, y1, x2, y2 = cell["cell"]
             cell_texts = []
 
+            # Helper: compute overlap ratio of bbox against cell
+            def overlap_ratio(bbox, cell_box):
+                ix1 = max(bbox[0], cell_box[0])
+                iy1 = max(bbox[1], cell_box[1])
+                ix2 = min(bbox[2], cell_box[2])
+                iy2 = min(bbox[3], cell_box[3])
+                iw = max(0, ix2 - ix1)
+                ih = max(0, iy2 - iy1)
+                inter = iw * ih
+                bbox_area = max(1, (bbox[2] - bbox[0]) * (bbox[3] - bbox[1]))
+                return inter / float(bbox_area)
+
+            # Helper: check center inside cell
+            def center_inside(bbox, cell_box):
+                cx = (bbox[0] + bbox[2]) / 2.0
+                cy = (bbox[1] + bbox[3]) / 2.0
+                return (cx >= cell_box[0] and cx <= cell_box[2] and
+                        cy >= cell_box[1] and cy <= cell_box[3])
+
+            cell_box = [x1, y1, x2, y2]
             for item in ocr_data_list:
                 bx1, by1, bx2, by2 = item["bbox"]
-                if bx1 >= x1 and by1 >= y1 and bx2 <= x2 and by2 <= y2:
+                bbox = [bx1, by1, bx2, by2]
+                # Accept if bbox is largely inside the cell, or its center lies inside the cell
+                if overlap_ratio(bbox, cell_box) >= 0.3 or center_inside(bbox, cell_box):
                     cell_texts.append(item["text"])
 
             # thêm vào cell gốc

@@ -2,46 +2,73 @@ import cv2
 import numpy as np
 import os
 
-def detect_tables(image_path):
+def filter_horizontal_lines(lines_h, img_width, min_h_len_ratio=0.8, tol_y=10):
+    if lines_h is None:
+        return [], []
+
+    ys_candidates = []
+    for l in lines_h:
+        x1, y1, x2, y2 = l[0]
+        if abs(y1 - y2) <= 3:  # ngang
+            line_len = abs(x2 - x1)
+            y_mid = int(round((y1 + y2) / 2))
+            ys_candidates.append((y_mid, line_len, x1, x2))
+
+    ys_candidates.sort(key=lambda x: x[0])
+    filtered_lines, line_segments, current_group = [], [], []
+
+    for y, length, x1, x2 in ys_candidates:
+        if not current_group:
+            current_group.append((y, length, x1, x2))
+        else:
+            if abs(y - current_group[-1][0]) <= tol_y:
+                current_group.append((y, length, x1, x2))
+            else:
+                longest = max(current_group, key=lambda x: x[1])
+                if longest[1] >= min_h_len_ratio * img_width:
+                    filtered_lines.append(longest[0])
+                    line_segments.append((longest[2], longest[3], longest[0]))
+                else:
+                    break
+                current_group = [(y, length, x1, x2)]
+
+    if current_group:
+        longest = max(current_group, key=lambda x: x[1])
+        if longest[1] >= min_h_len_ratio * img_width:
+            filtered_lines.append(longest[0])
+            line_segments.append((longest[2], longest[3], longest[0]))
+
+    total_rows = max(0, len(filtered_lines) - 1)
+    print(f"Detected {total_rows} rows")
+    return filtered_lines, line_segments
+
+
+def detect_tables(image_path, gap_threshold=50):
     img = cv2.imread(image_path)
     if img is None:
         raise FileNotFoundError(f"Không đọc được ảnh: {image_path}")
 
     gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
     blur = cv2.GaussianBlur(gray, (3, 3), 0)
-
-    # Edge detection
     edges = cv2.Canny(blur, 50, 150, apertureSize=3)
 
     # --- Horizontal lines ---
     lines_h = cv2.HoughLinesP(edges, 1, np.pi/180, threshold=120,
                               minLineLength=int(img.shape[1] * 0.6), maxLineGap=20)
-    ys_candidates, line_segments = [], []
-    if lines_h is not None:
-        for l in lines_h:
-            x1, y1, x2, y2 = l[0]
-            if abs(y1 - y2) <= 3:  # ngang
-                y_mid = int(round((y1 + y2) / 2))
-                ys_candidates.append(y_mid)
-                line_segments.append((x1, x2, y_mid))
-
-    # gom nhóm y
-    ys, tol_y = [], 10
-    for y in sorted(ys_candidates):
-        if not ys or abs(y - ys[-1]) > tol_y:
-            ys.append(y)
-
+    img_height, img_width = img.shape[:2]
+    ys, line_segments = filter_horizontal_lines(lines_h, img_width, min_h_len_ratio=0.8, tol_y=10)
     total_rows = max(0, len(ys) - 1)
 
     # --- Vertical lines ---
     lines_v = cv2.HoughLinesP(edges, 1, np.pi/180, threshold=100,
-                              minLineLength=int(img.shape[0] * 0.5), maxLineGap=20)
-    xs = []
+                              minLineLength=int(img.shape[0] * 0.4), maxLineGap=20)
+    v_lines, xs = [], []
     if lines_v is not None:
         for l in lines_v:
             x1, y1, x2, y2 = l[0]
             if abs(x1 - x2) <= 3:
                 xs.append(int(round((x1 + x2) / 2)))
+                v_lines.append((int(round((x1 + x2) / 2)), min(y1, y2), max(y1, y2)))
 
     # gom nhóm x
     x_pos, tol_v = [], 10
@@ -50,26 +77,66 @@ def detect_tables(image_path):
             x_pos.append(v)
 
     total_cols = max(0, len(x_pos) - 1)
-
     tables = []
+
     if total_rows > 0 and total_cols > 0:
         y_min, y_max = ys[0], ys[-1]
         x_min, x_max = x_pos[0], x_pos[-1]
         table_box = (x_min, y_min, x_max, y_max)
 
-        # build cells
         rows_data = []
         for i in range(total_rows):
             row_cells = []
-            for j in range(total_cols):
+            j = 0
+            while j < total_cols:
                 cell_box = (x_pos[j], ys[i], x_pos[j+1], ys[i+1])
+                row_height = cell_box[3] - cell_box[1]
+
+                # Check vertical line coverage (>=70% chiều cao hàng)
+                has_left = any(
+                    abs(x - cell_box[0]) <= tol_v and
+                    (min(y_end, cell_box[3]) - max(y_start, cell_box[1])) >= 0.7 * row_height
+                    for x, y_start, y_end in v_lines
+                )
+                has_right = any(
+                    abs(x - cell_box[2]) <= tol_v and
+                    (min(y_end, cell_box[3]) - max(y_start, cell_box[1])) >= 0.7 * row_height
+                    for x, y_start, y_end in v_lines
+                )
+
+                if has_left and has_right:
+                    col_start = j
+                    col_end = j
+                    # nếu cột tiếp theo không có line → merge
+                    while col_end + 1 < total_cols:
+                        next_box = (x_pos[col_end+1], ys[i], x_pos[col_end+2], ys[i+1])
+                        has_next_left = any(
+                            abs(x - next_box[0]) <= tol_v and
+                            (min(y_end, next_box[3]) - max(y_start, next_box[1])) >= 0.7 * row_height
+                            for x, y_start, y_end in v_lines
+                        )
+                        if not has_next_left:  # merge tiếp
+                            col_end += 1
+                        else:
+                            break
+
+                    merged_box = (x_pos[col_start], ys[i], x_pos[col_end+1], ys[i+1])
+                    if col_start == col_end:
+                        col_id = col_start
+                    else:
+                        col_id = f"{col_start}-{col_end}"
+
                     row_cells.append({
-                    "cell": cell_box,
+                        "cell": merged_box,
                         "row_idx": i,
-                    "col_idx": j
+                        "col_idx": col_id
                     })
-                # Vẽ ô
-                cv2.rectangle(img, (cell_box[0], cell_box[1]), (cell_box[2], cell_box[3]), (0, 255, 255), 1)
+                    cv2.rectangle(img, (merged_box[0], merged_box[1]),
+                                  (merged_box[2], merged_box[3]), (0, 255, 255), 1)
+                    j = col_end + 1
+                else:
+                    j += 1  # skip ô lỗi (không có line đầy đủ)
+
             rows_data.append(row_cells)
 
         tables.append({
@@ -78,11 +145,9 @@ def detect_tables(image_path):
             "table_box": table_box,
             "cells": rows_data
         })
-
-        # vẽ viền bảng
         cv2.rectangle(img, (x_min, y_min), (x_max, y_max), (255, 0, 0), 2)
 
-    debug_path = os.path.splitext(image_path)[0] + "_debug.jpg"
+    debug_path = os.path.splitext(image_path)[0] + "_fix_debug.jpg"
     cv2.imwrite(debug_path, img)
 
     return tables