from paddleocr import PaddleOCRimport pandas as pdimport osimport cv2import numpy as npfrom openpyxl import Workbookfrom openpyxl.styles import Font, Alignment, Border, Side, PatternFillfrom openpyxl.utils import get_column_letterimport reimport warningswarnings.filterwarnings('ignore')# ===================== 配置参数 =====================class Config: # 文件路径 MODEL_DIR = r"C:\Users\admin\.paddleocr" IMG_PATH = r"C:\Users\admin\Desktop\gongzi.png" OUTPUT_EXCEL = r"C:\Users\admin\Desktop\工资表识别结果.xlsx" # OCR参数 LINE_THRESHOLD = 15 # 同行判定阈值(像素) COLUMN_GAP = 20 # 列分割阈值 # 图片预处理参数 ENHANCE_CONTRAST = True # 是否增强对比度 DESKEW = True # 是否校正倾斜 REMOVE_NOISE = True # 是否去噪 BINARIZE = True # 是否二值化 # 表格识别参数 MIN_CELL_AREA = 100 # 最小单元格面积 MERGE_HEADER = True # 是否合并表头 AUTO_CORRECT = True # 是否自动纠错 USE_STRUCTURE_ANALYSIS = True # 是否使用表格结构分析cfg = Config()# ===================== 图片预处理模块 =====================class ImagePreprocessor: @staticmethod def enhance_contrast(img): """增强对比度""" lab = cv2.cvtColor(img, cv2.COLOR_BGR2LAB) l, a, b = cv2.split(lab) clahe = cv2.createCLAHE(clipLimit=3.0, tileGridSize=(8, 8)) l = clahe.apply(l) enhanced = cv2.merge([l, a, b]) return cv2.cvtColor(enhanced, cv2.COLOR_LAB2BGR) @staticmethod def remove_noise(img): """去除噪点""" return cv2.fastNlMeansDenoisingColored(img, None, 10, 10, 7, 21) @staticmethod def correct_skew(img): """校正图片倾斜""" gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) edges = cv2.Canny(gray, 50, 150, apertureSize=3) lines = cv2.HoughLines(edges, 1, np.pi/180, 200) if lines is not None: angles = [] for line in lines: rho, theta = line[0] angle = theta * 180 / np.pi - 90 if -45 < angle < 45: angles.append(angle) if angles: median_angle = np.median(angles) if abs(median_angle) > 0.5: h, w = img.shape[:2] center = (w // 2, h // 2) matrix = cv2.getRotationMatrix2D(center, median_angle, 1.0) img = cv2.warpAffine(img, matrix, (w, h), borderMode=cv2.BORDER_REPLICATE) return img @staticmethod def binarize_image(img): """二值化处理""" gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) # 自适应阈值 binary = cv2.adaptiveThreshold(gray, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 11, 2) return binary @classmethod def preprocess(cls, img_path): """完整的预处理流程""" img = cv2.imread(img_path) if img is None: raise ValueError(f"无法读取图片:{img_path}") print("原始图片尺寸:", img.shape) if cfg.REMOVE_NOISE: img = cls.remove_noise(img) print("✓ 去噪完成") if cfg.DESKEW: img = cls.correct_skew(img) print("✓ 倾斜校正完成") if cfg.ENHANCE_CONTRAST: img = cls.enhance_contrast(img) print("✓ 对比度增强完成") # 保留原始彩色图像用于结构分析 color_img = img.copy() if cfg.BINARIZE: binary_img = cls.binarize_image(img) print("✓ 二值化完成") else: binary_img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) return color_img, binary_img# ===================== 表格结构分析模块 =====================class TableStructureAnalyzer: @staticmethod def detect_lines(binary_img): """检测表格线""" # 检测横线 horizontal_kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (40, 1)) horizontal_lines = cv2.morphologyEx(binary_img, cv2.MORPH_OPEN, horizontal_kernel, iterations=2) # 检测竖线 vertical_kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (1, 40)) vertical_lines = cv2.morphologyEx(binary_img, cv2.MORPH_OPEN, vertical_kernel, iterations=2) # 合并横竖线 table_structure = cv2.add(horizontal_lines, vertical_lines) return horizontal_lines, vertical_lines, table_structure @staticmethod def find_cells(table_structure): """查找单元格""" contours, _ = cv2.findContours(table_structure, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE) cells = [] for contour in contours: x, y, w, h = cv2.boundingRect(contour) area = w * h if area > cfg.MIN_CELL_AREA: cells.append({ 'x': x, 'y': y, 'width': w, 'height': h, 'center_x': x + w//2, 'center_y': y + h//2 }) return cells# ===================== OCR识别模块 =====================class OCRProcessor: def __init__(self): self.ocr = PaddleOCR( use_angle_cls=True, lang="ch", show_log=False, use_gpu=False, det_db_thresh=0.3, det_db_box_thresh=0.3, det_db_unclip_ratio=1.8, det_model_dir=os.path.join(cfg.MODEL_DIR, "ch_PP-OCRv4_det_infer"), rec_model_dir=os.path.join(cfg.MODEL_DIR, "ch_PP-OCRv4_rec_infer"), cls_model_dir=os.path.join(cfg.MODEL_DIR, "ch_ppocr_mobile_v2.0_cls_infer") ) def recognize(self, img): """执行OCR识别""" # 对于二值化图像,转换为3通道 if len(img.shape) == 2: img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR) result = self.ocr.ocr(img, cls=True) return self._parse_result(result) def _parse_result(self, ocr_result): """解析OCR结果""" text_info = [] if ocr_result and ocr_result[0]: for item in ocr_result[0]: box = item[0] text = item[1][0] confidence = item[1][1] # 计算文本框的中心和范围 x_coords = [point[0] for point in box] y_coords = [point[1] for point in box] text_info.append({ 'text': text, 'confidence': confidence, 'x_min': min(x_coords), 'x_max': max(x_coords), 'y_min': min(y_coords), 'y_max': max(y_coords), 'center_x': sum(x_coords) / 4, 'center_y': sum(y_coords) / 4, 'box': box }) return text_info# ===================== 表格重建模块 =====================class TableReconstructor: @staticmethod def group_by_rows(text_info, threshold=None): """将文本按行分组""" if threshold is None: threshold = cfg.LINE_THRESHOLD if not text_info: return [] # 按y坐标排序 sorted_texts = sorted(text_info, key=lambda x: x['center_y']) rows = [] current_row = [sorted_texts[0]] for text in sorted_texts[1:]: if abs(text['center_y'] - current_row[-1]['center_y']) < threshold: current_row.append(text) else: # 按x坐标排序当前行 current_row.sort(key=lambda x: x['center_x']) rows.append(current_row) current_row = [text] # 添加最后一行 current_row.sort(key=lambda x: x['center_x']) rows.append(current_row) return rows @staticmethod def align_columns(rows): """列对齐处理""" if not rows: return [], [] # 收集所有文本的x坐标 all_x_positions = [] for row in rows: for cell in row: all_x_positions.append(cell['center_x']) # 使用聚类方法找到列边界 all_x_positions.sort() column_boundaries = TableReconstructor._cluster_positions(all_x_positions) # 构建表格矩阵 table_matrix = [] for row in rows: row_data = [''] * len(column_boundaries) for cell in row: col_idx = TableReconstructor._find_column(cell['center_x'], column_boundaries) if col_idx is not None and col_idx < len(row_data): if row_data[col_idx]: row_data[col_idx] += ' ' + cell['text'] else: row_data[col_idx] = cell['text'] table_matrix.append(row_data) return table_matrix, column_boundaries @staticmethod def _cluster_positions(positions, gap_threshold=None): """聚类x坐标以确定列""" if gap_threshold is None: gap_threshold = cfg.COLUMN_GAP if not positions: return [] clusters = [[positions[0]]] for pos in positions[1:]: if pos - clusters[-1][-1] < gap_threshold: clusters[-1].append(pos) else: clusters.append([pos]) # 返回每个聚类的中心 return [np.mean(cluster) for cluster in clusters] @staticmethod def _find_column(x_pos, boundaries): """找到x坐标对应的列索引""" min_dist = float('inf') col_idx = 0 for i, boundary in enumerate(boundaries): dist = abs(x_pos - boundary) if dist < min_dist: min_dist = dist col_idx = i # 设置合理的距离阈值 if min_dist > 100: # 如果距离太远,可能是新列 return None return col_idx @staticmethod def merge_headers(rows, max_header_rows=3): """智能合并表头""" if len(rows) <= max_header_rows: return rows # 检测表头行(通常前几行包含特殊格式) header_rows = rows[:max_header_rows] data_rows = rows[max_header_rows:] # 合并表头 merged_header = [] num_columns = max(len(row) for row in header_rows) for col in range(num_columns): header_texts = [] for row in header_rows: if col < len(row) and row[col].strip(): header_texts.append(row[col].strip()) merged_header.append('_'.join(header_texts) if header_texts else f'列{col+1}') return [merged_header] + data_rows @staticmethod def clean_table(table_matrix): """清理表格数据""" cleaned = [] for row in table_matrix: # 过滤空行 if any(cell.strip() for cell in row): # 清理每个单元格 cleaned_row = [TableReconstructor._clean_text(cell) for cell in row] cleaned.append(cleaned_row) return cleaned @staticmethod def _clean_text(text): """清理文本""" # 去除多余空格 text = re.sub(r'\s+', ' ', text.strip()) # 修正常见OCR错误 text = text.replace('O', '0').replace('l', '1').replace('I', '1') text = text.replace(':', ':').replace(',', ',') return text# ===================== Excel导出模块 =====================class ExcelExporter: @staticmethod def export_to_excel(table_data, output_path, headers=None): """导出为格式化的Excel文件""" wb = Workbook() ws = wb.active # 设置样式 header_font = Font(name='微软雅黑', size=11, bold=True, color='FFFFFF') header_fill = PatternFill(start_color='4472C4', end_color='4472C4', fill_type='solid') header_alignment = Alignment(horizontal='center', vertical='center', wrap_text=True) cell_font = Font(name='微软雅黑', size=10) cell_alignment = Alignment(horizontal='center', vertical='center', wrap_text=True) thin_border = Border( left=Side(style='thin'), right=Side(style='thin'), top=Side(style='thin'), bottom=Side(style='thin') ) # 写入表头 if headers: for col, header in enumerate(headers, 1): cell = ws.cell(row=1, column=col, value=header) cell.font = header_font cell.fill = header_fill cell.alignment = header_alignment cell.border = thin_border start_row = 2 else: start_row = 1 # 写入数据 for row_idx, row_data in enumerate(table_data, start_row): for col_idx, value in enumerate(row_data, 1): cell = ws.cell(row=row_idx, column=col_idx, value=value) cell.font = cell_font cell.alignment = cell_alignment cell.border = thin_border # 数值列右对齐 if value and value.replace('.', '').replace('-', '').isdigit(): cell.alignment = Alignment(horizontal='right', vertical='center') # 自动调整列宽 for col in ws.columns: max_length = 0 column = col[0].column_letter for cell in col: try: if len(str(cell.value)) > max_length: max_length = len(str(cell.value)) except: pass adjusted_width = min(max_length + 2, 30) ws.column_dimensions[column].width = adjusted_width # 冻结首行 if headers: ws.freeze_panes = 'A2' wb.save(output_path) print(f"✓ Excel文件已保存至: {output_path}")# ===================== 主程序 =====================def main(): print("=" * 60) print("图片表格识别系统 v1.0") print("=" * 60) # 1. 图片预处理 print("\n【步骤1】图片预处理...") preprocessor = ImagePreprocessor() color_img, binary_img = preprocessor.preprocess(cfg.IMG_PATH) # 2. 表格结构分析 print("\n【步骤2】表格结构分析...") if cfg.USE_STRUCTURE_ANALYSIS: analyzer = TableStructureAnalyzer() h_lines, v_lines, table_structure = analyzer.detect_lines(binary_img) cells = analyzer.find_cells(table_structure) print(f"✓ 检测到 {len(cells)} 个单元格") # 根据检测到的单元格动态调整阈值 if cells: avg_cell_height = np.mean([c['height'] for c in cells]) cfg.LINE_THRESHOLD = max(int(avg_cell_height * 0.5), 10) print(f"✓ 自动调整行阈值: {cfg.LINE_THRESHOLD}") # 3. OCR识别 print("\n【步骤3】OCR文字识别...") ocr_processor = OCRProcessor() # 对原始彩色图像和预处理图像分别识别,取最佳结果 text_info_color = ocr_processor.recognize(color_img) text_info_binary = ocr_processor.recognize(binary_img) # 合并识别结果,优先使用高置信度结果 text_info = text_info_color if len(text_info_color) > len(text_info_binary) else text_info_binary print(f"✓ 识别到 {len(text_info)} 个文本区域") if len(text_info) < 10: print("⚠ OCR识别文本过少,尝试调整参数...") # 使用更激进的预处理 enhanced_img = cv2.convertScaleAbs(color_img, alpha=1.5, beta=30) text_info = ocr_processor.recognize(enhanced_img) print(f"✓ 增强后识别到 {len(text_info)} 个文本区域") # 4. 表格重建 print("\n【步骤4】表格重建...") reconstructor = TableReconstructor() # 按行分组 rows = reconstructor.group_by_rows(text_info) print(f"✓ 分组为 {len(rows)} 行") # 列对齐 table_matrix, column_boundaries = reconstructor.align_columns(rows) print(f"✓ 识别到 {len(column_boundaries)} 列") # 清理数据 table_matrix = reconstructor.clean_table(table_matrix) # 合并表头 if cfg.MERGE_HEADER and len(table_matrix) > 1: table_matrix = reconstructor.merge_headers(table_matrix) print("✓ 表头合并完成") # 5. 数据处理和导出 print("\n【步骤5】数据导出...") if len(table_matrix) > 1: headers = table_matrix[0] data = table_matrix[1:] print(f"✓ 表头: {len(headers)} 列") print(f"✓ 数据行: {len(data)} 行") else: headers = None data = table_matrix print("⚠ 未检测到明确表头,使用自动列名") # 预览前几行数据 print("\n【数据预览】") preview_rows = min(5, len(data)) if headers: print("表头:", headers[:10], "..." if len(headers) > 10 else "") for i in range(preview_rows): print(f"第{i+1}行:", data[i][:10], "..." if len(data[i]) > 10 else "") # 导出Excel ExcelExporter.export_to_excel(data, cfg.OUTPUT_EXCEL, headers) print("\n" + "=" * 60) print("识别完成!") print(f"输出文件: {cfg.OUTPUT_EXCEL}") print("=" * 60)if __name__ == "__main__": try: main() except Exception as e: print(f"\n❌ 程序运行出错: {str(e)}") import traceback traceback.print_exc()