文字识别准确率

import easyocr
import cv2
import os
import numpy as np

def calculate_ocr_accuracy(image_path, ground_truth_text=None):
    reader = easyocr.Reader(['ch_sim', 'en'])
    
    try:
        image = cv2.imread(image_path)
        if image is None:
            raise ValueError(f"无法读取图片: {image_path}")
        
        results = reader.readtext(image)
        
        recognized_text = ""
        confidence_scores = []
        
        for (bbox, text, confidence) in results:
            recognized_text += text + " "
            confidence_scores.append(confidence)
        
        recognized_text = recognized_text.strip()
        
        print("=" * 50)
        print("OCR识别结果:")
        print("=" * 50)
        
        for i, (bbox, text, confidence) in enumerate(results):
            print(f"文本块 {i+1}: '{text}' (置信度: {confidence:.4f})")
        
        print(f"\n完整识别文本: {recognized_text}")
        
        avg_confidence = np.mean(confidence_scores) if confidence_scores else 0
        print(f"\n平均置信度: {avg_confidence:.4f}")
        
        if ground_truth_text:
            accuracy = calculate_text_accuracy(ground_truth_text, recognized_text)
            print(f"文本准确率: {accuracy:.2f}%")
            
            return {
                'recognized_text': recognized_text,
                'ground_truth': ground_truth_text,
                'accuracy': accuracy,
                'avg_confidence': avg_confidence,
                'details': results
            }
        else:
            return {
                'recognized_text': recognized_text,
                'avg_confidence': avg_confidence,
                'details': results
            }
            
    except Exception as e:
        print(f"处理图片时出错: {e}")
        return None

def calculate_text_accuracy(ground_truth, recognized):
    gt_clean = ''.join(ground_truth.split()).lower()
    rec_clean = ''.join(recognized.split()).lower()
    
    distance = levenshtein_distance(gt_clean, rec_clean)
    
    max_len = max(len(gt_clean), len(rec_clean))
    if max_len == 0:
        return 100.0
    
    accuracy = (1 - distance / max_len) * 100
    return accuracy

def levenshtein_distance(s1, s2):
    if len(s1) < len(s2):
        return levenshtein_distance(s2, s1)
    
    if len(s2) == 0:
        return len(s1)
    
    previous_row = range(len(s2) + 1)
    for i, c1 in enumerate(s1):
        current_row = [i + 1]
        for j, c2 in enumerate(s2):
            insertions = previous_row[j + 1] + 1
            deletions = current_row[j] + 1
            substitutions = previous_row[j] + (c1 != c2)
            current_row.append(min(insertions, deletions, substitutions))
        previous_row = current_row
    
    return previous_row[-1]

def batch_ocr_accuracy_test(image_folder, ground_truths):
    total_accuracy = 0
    total_confidence = 0
    count = 0
    
    print("开始批量OCR准确率测试...")
    print("=" * 60)
    
    for filename, truth_text in ground_truths.items():
        image_path = os.path.join(image_folder, filename)
        if os.path.exists(image_path):
            print(f"\n处理图片: {filename}")
            result = calculate_ocr_accuracy(image_path, truth_text)
            
            if result and 'accuracy' in result:
                total_accuracy += result['accuracy']
                total_confidence += result['avg_confidence']
                count += 1
    
    if count > 0:
        print("\n" + "=" * 60)
        print("批量测试结果汇总:")
        print(f"测试图片数量: {count}")
        print(f"平均准确率: {total_accuracy/count:.2f}%")
        print(f"平均置信度: {total_confidence/count:.4f}")
        print("=" * 60)

def main():
    image_path = r"C:\Users\18306\Desktop\picture\test_image.jpg"
    
    if not os.path.exists(image_path):
        print(f"错误: 图片文件不存在 - {image_path}")
        return
    
    print("开始OCR识别...")
    
    result = calculate_ocr_accuracy(image_path)
    
    if result:
        print("\n识别完成！")
    else:
        print("识别失败！")

if __name__ == "__main__":
    main()