computer-vision-warehouse

当用户想要在仓库中应用计算机视觉，检测缺陷、跟踪包裹、计数库存或自动化视觉检查时。或者当用户提到“计算机视觉”、“图像识别”、“物体检测”、“条形码读取”、“包裹跟踪”、“质量检查”、“YOLO”、“RCNN”、“图像分类”或“基于视觉的仓库自动化”时。对于一般的机器学习，请参阅ml-supply-chain。

Computer Vision for Warehouse Operations

You are an expert in applying computer vision to warehouse and supply chain operations. Your goal is to implement object detection, classification, tracking, and inspection systems using CNNs, YOLO, and other vision models.

Applications

Package Detection & Tracking: YOLO, tracking algorithms
Quality Inspection: CNN classification, defect detection
Inventory Counting: Object counting, OCR
Barcode/QR Reading: Traditional + ML methods
Safety Monitoring: Person detection, PPE detection

YOLO for Package Detection

import cv2
from ultralytics import YOLO

class PackageDetector:
    """
    Real-time package detection using YOLO
    """
    
    def __init__(self, model_path='yolov8n.pt'):
        self.model = YOLO(model_path)
    
    def detect_packages(self, image_path):
        """Detect packages in warehouse image"""
        
        results = self.model(image_path)
        
        detections = []
        for result in results:
            boxes = result.boxes
            for box in boxes:
                x1, y1, x2, y2 = box.xyxy[0]
                confidence = box.conf[0]
                class_id = box.cls[0]
                
                detections.append({
                    'bbox': (x1, y1, x2, y2),
                    'confidence': confidence,
                    'class': class_id
                })
        
        return detections

CNN for Quality Inspection

from tensorflow import keras
from tensorflow.keras import layers

class QualityInspector:
    """
    CNN for defect detection
    """
    
    def build_model(self, image_size=(224, 224), num_classes=2):
        model = keras.Sequential([
            layers.Conv2D(32, 3, activation='relu',
                         input_shape=(*image_size, 3)),
            layers.MaxPooling2D(2),
            layers.Conv2D(64, 3, activation='relu'),
            layers.MaxPooling2D(2),
            layers.Conv2D(128, 3, activation='relu'),
            layers.MaxPooling2D(2),
            layers.Flatten(),
            layers.Dense(256, activation='relu'),
            layers.Dropout(0.5),
            layers.Dense(num_classes, activation='softmax')
        ])
        
        model.compile(optimizer='adam',
                     loss='categorical_crossentropy',
                     metrics=['accuracy'])
        
        return model

Object Tracking

from sort import Sort

class PackageTracker:
    """
    Multi-object tracking for packages
    """
    
    def __init__(self):
        self.tracker = Sort()
        self.package_trajectories = {}
    
    def track(self, detections, frame_id):
        """
        Track packages across frames
        """
        
        # Convert detections to format for tracker
        dets = np.array([[d['bbox'][0], d['bbox'][1],
                         d['bbox'][2], d['bbox'][3],
                         d['confidence']] for d in detections])
        
        # Update tracker
        tracked_objects = self.tracker.update(dets)
        
        # Store trajectories
        for obj in tracked_objects:
            obj_id = int(obj[4])
            bbox = obj[:4]
            
            if obj_id not in self.package_trajectories:
                self.package_trajectories[obj_id] = []
            
            self.package_trajectories[obj_id].append({
                'frame': frame_id,
                'bbox': bbox
            })

Tools & Libraries

OpenCV: image processing
YOLOv8: object detection
TensorFlow/PyTorch: deep learning
Roboflow: dataset management
Detectron2: Facebook detection

Related Skills

ml-supply-chain: general ML
warehouse-automation: automation systems
quality-management: inspection processes