大闸蟹外观缺陷识别数据

借助机器视觉技术，可实现大闸蟹的高效分类与价值最大化，精准识别大闸蟹外观完整性。外观完全完整、可纳入高端品类，用于礼盒包装，主打节日礼品市场和高端餐饮渠道，对于断半腿、短腿等残次大闸蟹，通过分级系统标记后，采取差异化处理：轻度残缺的可低价流入大众消费市场，满足家庭日常食用需求；严重残缺的则定向输送至加工企业，经清洗、拆解后制成蟹肉棒、蟹粉等深加工产品，提升原料利用率。 采集残缺的大闸蟹图像，包括断整腿的大闸蟹图像和断半腿的大闸蟹图像。对图像预处理，保留原始分辨率，以保留材料表面的原有特征。采用随机旋转-随机缩放-随机亮度调整-随机对比度调整-随机噪声添加对图像进行增强。使用labelme等标注工具基于视觉特征判断对缺陷的位置进行精确标注，使用COCO包围框(x,y,width,height)进行定位和目标检测标注，其中x为包围框起点x轴坐标，y为包围框起点y轴坐标，width为包围框宽度,height为包围框高度；将缺陷分为断整腿，断半腿2种类别， 将识别缺陷区域的COCO包围框转换为数据格式；转换公式如下：x1=x,y1=y,x2=x+width,y2=y+height,x1为缺陷标记区域起点x轴坐标，y1为缺陷标记区域起点y轴坐标，x2为缺陷标记区域终点x轴坐标，y2为缺陷标记区域终点y轴坐标，缺陷识别类型∈{"断整腿","断半腿"}，（x1,y1,x2,y2）是缺陷标记坐标。Machine vision technology enables efficient classification and value maximization of Chinese mitten crabs (Eriocheir sinensis) by accurately identifying their appearance integrity. Crabs with completely intact appearance are categorized as high-end products, suitable for gift box packaging and targeting holiday gift markets and high-end dining channels. For defective Chinese mitten crabs with half legs missing or short legs, a grading system is employed for tagging, followed by differentiated processing: mildly defective crabs can be sold at low prices to the mass consumer market to meet daily family dining needs; severely defective crabs are directed to processing enterprises, where they are made into deep-processed products such as crab sticks and crab paste after cleaning and dismantling, thereby improving raw material utilization. Image collection of defective Chinese mitten crabs is conducted, including images of crabs with entire legs missing and crabs with half legs missing. Image preprocessing is performed while retaining the original resolution to preserve the original features of the crab surface. Image augmentation is implemented via random rotation, random scaling, random brightness adjustment, random contrast adjustment, and random noise addition. Annotation tools such as LabelMe are used to accurately mark the location of defects based on visual feature judgment. COCO bounding boxes (x, y, width, height) are adopted for localization and object detection annotation, where x is the x-axis coordinate of the bounding box's starting point, y is the y-axis coordinate of the starting point, width is the width of the bounding box, and height is the height of the bounding box. Defects are divided into two categories: entire legs missing and half legs missing. The COCO bounding boxes for identifying defect regions are converted to the target data format using the following conversion formula: x1 = x, y1 = y, x2 = x + width, y2 = y + height, where x1 is the x-axis coordinate of the starting point of the defect marking region, y1 is the y-axis coordinate of the starting point, x2 is the x-axis coordinate of the ending point of the defect marking region, y2 is the y-axis coordinate of the ending point of the defect marking region, and the defect recognition type falls into {"entire legs missing", "half legs missing"}, with (x1, y1, x2, y2) being the defect marking coordinates.

搜集汇总

背景与挑战

背景概述

该数据集包含1462条大闸蟹外观缺陷识别数据，以xlsx格式存储，主要用于通过机器视觉技术识别大闸蟹的断整腿和断半腿等缺陷，实现高效分类和价值最大化。数据集采用图像预处理和增强方法，并使用COCO包围框进行精确标注，支持农业领域的智能分级和加工应用。