玉米质量

在图像预处理阶段，为了更好地识别单个玉米粒的缺陷，我们处理了包含大量玉米的原始图像。首先，我们对图像进行花岗岩化，并通过高斯滤波，以提高图像的清晰度和质量。接下来，我们应用阈值对图像进行二值化，并提取轮廓以获得玉米粒的边缘信息。通过进一步处理，我们获得了每个玉米粒的中心点坐标，并根据最小的外部矩形裁剪出单个玉米图像。最终，我们总共处理了 7,361 张玉米粒图像。

During the image preprocessing stage, to better identify defects on individual corn kernels, we processed raw images containing a large number of corn kernels. First, we converted the images to grayscale and applied Gaussian filtering to enhance their clarity and quality. Next, we used thresholding to binarize the images and extracted contours to obtain the edge information of the corn kernels. Through further processing, we acquired the central coordinates of each corn kernel, and cropped individual corn kernel images based on their minimum bounding rectangles. Finally, we processed a total of 7,361 corn kernel images.

搜集汇总

背景与挑战

背景概述

该数据集包含7,361张经过预处理的玉米粒图像，适用于深度学习和神经网络研究，采用CC BY 4.0许可协议。预处理步骤包括图像增强、二值化和轮廓提取，旨在提高玉米粒缺陷识别的准确性。