Update app.py

app.py

@@ -10,22 +10,44 @@ import time
 # Initialize PaddleOCR once with updated parameters
 ocr_model = PaddleOCR(use_textline_orientation=True, lang='en')
 
-def analyze_uv_coverage(img, brightness_threshold=150):
+def analyze_uv_coverage(img, brightness_threshold=150, kernel_size=5, apply_blur=True, adaptive_thresh=False):
+    """
+    Analyze UV sterilization coverage by thresholding the grayscale image.
+    Optional adaptive thresholding and Gaussian blur for noise reduction.
+    Morphological operations clean the mask for better accuracy.
+    """
     gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
-    _, binary_mask = cv2.threshold(gray, brightness_threshold, 255, cv2.THRESH_BINARY)
 
-    kernel = np.ones((5, 5), np.uint8)
-    binary_mask = cv2.dilate(binary_mask, kernel, iterations=1)
+    if apply_blur:
+        gray = cv2.GaussianBlur(gray, (5, 5), 0)
+
+    if adaptive_thresh:
+        binary_mask = cv2.adaptiveThreshold(
+            gray, 255,
+            cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
+            cv2.THRESH_BINARY,
+            11, 2)
+    else:
+        _, binary_mask = cv2.threshold(gray, brightness_threshold, 255, cv2.THRESH_BINARY)
+
+    # Morphological opening (erosion followed by dilation) to remove noise
+    kernel = np.ones((kernel_size, kernel_size), np.uint8)
+    binary_mask = cv2.morphologyEx(binary_mask, cv2.MORPH_OPEN, kernel, iterations=1)
+
+    # Morphological closing (dilation followed by erosion) to close small holes inside foreground
+    binary_mask = cv2.morphologyEx(binary_mask, cv2.MORPH_CLOSE, kernel, iterations=1)
 
     total_pixels = binary_mask.size
     sterilized_pixels = cv2.countNonZero(binary_mask)
     coverage_percent = (sterilized_pixels / total_pixels) * 100
 
+    # Create overlay for visualization: Green = sterilized, Red = unsterilized
     overlay = img.copy()
-    overlay[binary_mask == 255] = [0, 255, 0]  # Green sterilized
-    overlay[binary_mask == 0] = [0, 0, 255]    # Red unsterilized
+    overlay[binary_mask == 255] = [0, 255, 0]  # Green
+    overlay[binary_mask == 0] = [0, 0, 255]    # Red
 
     annotated_img = cv2.addWeighted(img, 0.6, overlay, 0.4, 0)
+
     return annotated_img, coverage_percent
 
 def create_pdf_report(coverage_percent, extracted_texts, annotated_image_path, output_path):
@@ -44,7 +66,9 @@ def create_pdf_report(coverage_percent, extracted_texts, annotated_image_path, o
     pdf.set_font("Arial", size=10)
     if extracted_texts:
         for text in extracted_texts:
-            pdf.multi_cell(0, 8, f"- {text}")
+            # Filter out very short or empty OCR texts to improve clarity
+            if len(text.strip()) > 1:
+                pdf.multi_cell(0, 8, f"- {text}")
     else:
         pdf.cell(0, 8, "No text detected.", ln=True)
 
@@ -57,7 +81,7 @@ def create_pdf_report(coverage_percent, extracted_texts, annotated_image_path, o
 def process_image(input_img, brightness_threshold=150):
     img = cv2.cvtColor(np.array(input_img), cv2.COLOR_RGB2BGR)
 
-    # Resize large images for faster processing
+    # Resize large images for faster processing, preserving aspect ratio
     max_dim = 640
     h, w = img.shape[:2]
     if max(h, w) > max_dim:
@@ -72,24 +96,29 @@ def process_image(input_img, brightness_threshold=150):
     for line in ocr_result:
         if line:
             for word_info in line:
-                extracted_texts.append(word_info[1][0])
+                # Filter short strings and whitespace only
+                text = word_info[1][0].strip()
+                if len(text) > 1:
+                    extracted_texts.append(text)
 
     annotated_img, coverage_percent = analyze_uv_coverage(img, brightness_threshold)
 
     with tempfile.NamedTemporaryFile(delete=False, suffix=".jpg") as temp_img_file:
         cv2.imwrite(temp_img_file.name, annotated_img)
+        annotated_img_path = temp_img_file.name
 
-        temp_pdf_file = tempfile.NamedTemporaryFile(delete=False, suffix=".pdf")
-        temp_pdf_file.close()
-        create_pdf_report(coverage_percent, extracted_texts, temp_img_file.name, temp_pdf_file.name)
+    temp_pdf_file = tempfile.NamedTemporaryFile(delete=False, suffix=".pdf")
+    temp_pdf_file.close()
+    create_pdf_report(coverage_percent, extracted_texts, annotated_img_path, temp_pdf_file.name)
 
     annotated_img_rgb = cv2.cvtColor(annotated_img, cv2.COLOR_BGR2RGB)
 
     report_text = f"UV Sterilization Coverage: {coverage_percent:.2f}%"
-    # Optionally include OCR time:
-    # report_text += f"\n\nOCR Processing Time: {ocr_time:.2f} seconds"
+    # Optionally include OCR time for debugging:
+    # report_text += f"\nOCR Processing Time: {ocr_time:.2f} seconds"
 
-    os.unlink(temp_img_file.name)
+    # Clean up temp image file after PDF generation
+    os.unlink(annotated_img_path)
 
     return annotated_img_rgb, report_text, temp_pdf_file.name