智能检测日常刀具算法模型的训练数据

本数据集包含多角度、多场景下的日常刀具的X光安检图像，通过对图像的标注、抠图、增强、融合等处理，可作为优质样本训练生成智能检测日常刀具的算法模型，实现对复杂环境下的日常刀具器精准识别。1、数据来源：应用X射线光源多角度、多场景下透射日常刀具物品，采集并建立其原始的X光数据图例库。 2、数据深度处理：对采集到的原始X光图像预标注坐标位置和品项类别，并对各类日常刀具的X光图像进行抠图处理。将抠出的图像与多场景图像分别进行几何变换、像素变换等增广处理。 3、检测模型生成算法规则：将处理后的日常刀具的X光安检图像和场景X光图像通过密度统计（像素值代表实物密度值）依据区域匹配原则进行融合，融合区域掩模作为数据标签与融合后的图像作为深度学习样本数据。还可通过调整抠图区域在场景图像区域的位置，获得不同的平均密度差值，训练生成可精准定位、精准识别压缩和液化气体容器的智能检测模型。区域匹配原则按照Mask*(α*ρ抠图图像+β*ρ场景图像)，融合后的图像处理公式按照Mask*(α*ρ抠图图像+β*ρ抠图图像)+（1-Mask）*ρ场景图像。（所述公式中：Mask为图像掩膜，图像目标区域值为1，目标区域外值为4，ρ为密度值，α、β指系数）检测模型可对多场景下的日常刀具精准识别，同时将目标物的位置及所在X光图像信息记录标出。进一步的还可根据目标物位置信息推算目标物尺寸信息。

This dataset comprises X-ray security inspection images of daily knives captured under multiple angles and scenarios. Through processing operations including image annotation, matting, enhancement and fusion, it can serve as high-quality samples for training intelligent detection algorithms for daily knives, enabling accurate recognition of such items in complex environments.
1. Data Source: X-ray sources are utilized to transmit daily knife items under multiple angles and scenarios, and an original X-ray image database is established through collection.
2. In-depth Data Processing: Pre-labels for coordinate positions and item categories are added to the collected original X-ray images, and matting processing is performed on the X-ray images of various daily knives. The extracted matting images and multi-scenario images are then respectively subjected to augmentation processing such as geometric transformation and pixel transformation.
3. Rules for Detection Model Generation Algorithm: The processed X-ray security inspection images of daily knives and scenario X-ray images are fused according to the regional matching principle via density statistics, where pixel values represent physical density values. The fused regional mask is used as the data label, and the fused image serves as the deep learning sample data. Additionally, by adjusting the position of the matting-extracted region within the scenario image region, different average density differences can be obtained to train an intelligent detection model capable of accurately locating and recognizing compressed and liquefied gas containers. The regional matching principle follows the formula: $ ext{Mask} imes (alpha imes
ho_{ ext{matting}} + eta imes
ho_{ ext{scenario}})$, and the fused image processing formula follows: $ ext{Mask} imes (alpha imes
ho_{ ext{matting}} + eta imes
ho_{ ext{matting}}) + (1- ext{Mask}) imes
ho_{ ext{scenario}}$. (In the above formula: Mask is the image mask, with a value of 1 within the target region and 4 outside the target region; $
ho$ represents the density value, and $alpha$ and $eta$ are adjustable coefficients).
The trained detection model can accurately identify daily knives across multiple scenarios, while recording and marking the position of the target object and the information of the X-ray image where it resides. Furthermore, the size information of the target object can be inferred based on its position information.

搜集汇总

数据集介绍

特点

该数据集包含1221条多角度、多场景下的日常刀具X光安检图像，经过深度处理后用于训练智能检测算法模型，适用于复杂环境下的日常刀具精准识别。