随着深度学习的进步，卷积神经网络（convolutional neural network, CNN）在目标检测领域的应用取得了显著的提升。相比传统人工特征构造的方法，CNN算法具备自动特征提取和强大的泛化能力，在复杂场景下，目标检测任务得到更好的解决。此外，该算法具有更好的鲁棒性，使其在实际应用中具备更高的可靠性和效率。首先回顾卷积神经网络的发展历程，重点介绍其在计算机视觉领域的应用，并突出与传统算法相比在图像处理方面的明显优势。接着，对分类与分割算法、目标检测算法和基于区域与回归检测的3种主流算法进行介绍，特别强调各类算法在解决目标定位、分类和语义分割等关键问题上的创新和突破。然后，对现有的公开数据集进行归纳。最后，展望目标检测算法未来的发展趋势。

With the advancement of deep learning, the application of convolutional neural network(CNN) in the field of target detection has made significant improvements. Compared with the traditional manual feature construction methods, the CNN algorithm has automatic feature extraction and strong generalization ability, which makes the target detection task better solved in complex scenarios. In addition, the algorithm has better robustness, making it more reliable and efficient in practical applications. Firstly, the development history of CNN was reviewed, focusing on its application in the field of computer vision and highlighting the obvious advantages in image processing compared with traditional algorithms. Then, three mainstream algorithms, namely, classification and segmentation algorithms, target detection algorithms, and region and regression based detection, were introduced, with special emphasis on the innovations and breakthroughs of each type of algorithm in solving the key problems of target localization, classification, and semantic segmentation. Then, the existing public datasets were summarized. Finally, the future development trend of target detection algorithms was envisioned.