智能检测烟花爆竹算法模型的图像训练数据

本数据集包含种类丰富、多样环境下的烟花爆竹的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烟花爆竹的算法模型，实现对藏匿在其他物品中或伪装成其他物品等复杂环境下的烟花爆竹的精准识别。1、数据来源：应用X射线光源多角度、多场景下透射多类别烟花爆竹物品，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，并对采集的烟花爆竹的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：将处理后的烟花爆竹的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、精准识别烟花爆竹的智能检测模型。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7，ρ为密度值，α、β指系数）检测模型可对各种环境下的烟花爆竹精准识别，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 fireworks and firecrackers captured across diverse categories and complex environments. Through processing steps including image annotation, matting, enhancement, and fusion, it can serve as high-quality samples to train intelligent fireworks and firecrackers detection models, enabling accurate identification of fireworks and firecrackers hidden in or disguised as other items in complex scenarios.

1. Data Source: Multiple types of fireworks and firecrackers were scanned using X-ray sources from multiple angles and scenarios, and the original X-ray image database was collected and established.

2. In-depth Data Processing: Pre-label the coordinate positions and item categories for the collected original X-ray images, and perform matting on the X-ray images of the collected fireworks and firecrackers. Subsequently, conduct data augmentation operations such as geometric transformation and pixel transformation on the matting-extracted images and multi-scenario images respectively.

3. Algorithm Rules for Detection Model Generation: Fuse the processed X-ray security inspection images of fireworks and firecrackers and the scenario X-ray images based on the regional matching principle, using density statistics (where pixel values represent physical density values). The fusion region mask is used as the data label, and the fused images are used as deep learning sample data. Additionally, by adjusting the position of the matting area within the scenario image area, different average density differences can be obtained to train an intelligent detection model capable of accurately positioning and identifying fireworks and firecrackers. The regional matching principle follows the formula: $Mask*(alpha*
ho_{matting} + eta*
ho_{matting})$, and the image processing formula for the fused result follows: $Mask*(alpha*
ho_{matting} + eta*
ho_{matting}) + (1-Mask)*
ho_{scenario}$. In the aforementioned formulas: Mask is the image mask, with a value of 1 in the target area of the image and 7 outside the target area, ρ represents the density value, and α and β are coefficients. The detection model can accurately identify fireworks and firecrackers in various environments, while recording and marking the position of the target object and its corresponding X-ray image information. Furthermore, the size information of the target object can be inferred based on its position information.