基于旋转框定位的拆垛箱体目标检测

doi:10.3778/j.issn.1002-8331.2310-0102

摘要/Abstract

摘要： 针对物流领域箱体拆垛场景中，箱体密集堆叠、排列杂乱且来料随机等导致难以精确定位的问题，提出了一种基于旋转框定位的轻量级箱体精确检测方法。在单阶段目标检测网络YOLOv7-tiny的基础上，针对箱体目标密集排列的问题，引入融合多尺度可变形卷积的特征选择模块改善箱体形变建模，提升上下文特征提取能力；融合坐标注意力（coordinate attention，CA）机制以增强目标位置特征表达；采用旋转框长边表示法精确表示箱体边界，并结合高斯分布的KL散度（Kullback-Leibler divergence，KLD）作为损失函数，实现箱体旋转框的高精度回归。实验结果表明，改进方法在箱体数据集上检测精度AP75达到85.25%，与原模型相比提升了7.85个百分点，箱体中心点平均像素距离小于3个像素，平均偏移角度小于4°，在小幅增加模型参数量的基础上，检测速度达到26.52?FPS，所提方法可以有效提高拆垛箱体的定位精度，满足实际拆垛需求。

关键词: 旋转目标检测, 箱体定位, YOLOv7-tiny, 特征融合, 注意力机制

Abstract: In order to solve the problems of dense stacking, messy arrangement and random incoming materials in the logistics field, which makes it difficult to accurately locate the depalletizing box, a lightweight and precise positioning method for box based on rotating frame location is proposed. Based on the one-stage detection network YOLOv7-tiny, a feature selection module integrating deformable convolution is used to improve the box deformation modeling and enhance the extraction of contextual features, specifically addressing the problem of densely arranged box targets. Coordinate attention (CA) mechanism is integrated to enhance the expression of target features. The long-edge representation of the rotating box is used to accurately represent the box target frame, and the Kullback-Leibler divergence (KLD) between Gaussian distributions is utilized for the loss function to realize the exact regression of the box boundary. The experimental results show that the proposed method achieves 85.25% AP75 on the box dataset, which is 7.85 percentage points higher than the baseline model, the average pixel distance of the center point of the box is less than 3 pixels, and the average offset angle is less than 4 degrees, and the detection speed reaches 26.52 FPS on the basis of slightly increasing the number of model parameters. The proposed method can effectively improve the positioning accuracy of the depalletizing box and meet the actual destacking needs.

Key words: rotating object detection, box location, YOLOv7-tiny, feature fusion, attention mechanism

张相胜, 程嘉宝, 顾斌杰. 基于旋转框定位的拆垛箱体目标检测[J]. 计算机工程与应用, 2025, 61(4): 323-330.

ZHANG Xiangsheng, CHENG Jiabao, GU Binjie. Object Detection of Depalletizing Box Based on Rotating Frame Location[J]. Computer Engineering and Applications, 2025, 61(4): 323-330.

参考文献

[1] 张芹, 孙永明, 夏水斌, 等. 自动化智能仓储系统的设计与可行性分析[J]. 计算机与数字工程, 2019, 47(10): 2451-2454.
ZHANG Q, SUN Y M, XIA S B, et al. Design and feasibility analysis of automated intelligent storage system[J]. Computer & Digital Engineering, 2019, 47(10): 2451-2454.
[2] 杜宇, 姜伟. 智能机器人仓储物流系统设计[J]. 组合机床与自动化加工技术, 2020(5): 164-168.
DU Y, JIANG W. Design of intelligent robot warehousing logistics system[J]. Modular Machine Tool & Automatic Manufacturing Technique, 2020(5): 164-168.
[3] 赵永良, 付鑫, 吴尚远, 等. 基于计算机视觉的智能仓储图像识别系统设计与实现[J]. 电力信息与通信技术, 2019, 17(12): 31-36.
ZHAO Y L, FU X, WU S Y, et al. Design and implementation of intelligent warehouse image recognition system based on computer vision[J]. Electric Power Information and Communication Technology, 2019, 17(12): 31-36.
[4] 张晓芳, 许佳音, 王文博, 等. 箱体智能识别定位与引导方法研究[J]. 工程机械, 2022, 53(12): 1-7.
ZHANG X F, XU J Y, WANG W B, et al. Research on method of intelligent identification, positioning and guidance method for box[J]. Construction Machinery and Equipment, 2022, 53(12): 1-7.
[5] 路鹭, 杨炯, 梁杰, 等. 一种快速高精度的矩形检测算法[J]. 激光与光电子学进展, 2020, 57(18): 64-70.
LU L, YANG J, LIANG J, et al. Fast and high-precision rectangle detection algorithm[J]. Laser & Optoelectronics Progress, 2020, 57(18): 64-70.
[6] 金卓鑫. 基于轮廓特征的仓储箱体识别与定位算法研究[D]. 武汉: 华中科技大学, 2020: 28-46.
JIN Z X. Research on recognition and location algorithm of warehouse box based on contour features[D]. Wuhan: Huazhong University of Science and Technology, 2020: 28-46.
[7] 王照华. 面向堆叠矩形物体的识别与抓取方法研究[D]. 哈尔滨: 哈尔滨工业大学, 2019: 21-35.
WANG Z H. Research on recognition and grasping method of stacked rectangular objects[D]. Harbin: Harbin Institute of Technology, 2019: 21-35.
[8] 李天剑, 黄斌, 刘江玉, 等. 卷积神经网络物体检测算法在物流仓库中的应用[J]. 计算机工程, 2018, 44(6): 176-181.
LI T J, HUANG B, LIU J Y, et al. Application of convolution neural network object detection algorithm in logistics warehouse[J]. Computer Engineering, 2018, 44(6): 176-181.
[9] 李映松, 杨爱英, 刘轩, 等. 基于Transformer改进的Faster-RCNN仓储箱体检测算法[J]. 自动化与仪器仪表, 2022(8): 1-6.
LI Y S, YANG A Y, LIU X, et al. Storage box detection method based on transformer improved Faster-RCNN[J]. Automation & Instrumentation, 2022(8): 1-6.
[10] SHI P, ZHAO Z, FAN X, et al. Remote sensing image object detection based on angle classification[J]. IEEE Access, 2021, 9: 118696-118707.
[11] 韩兴, 刘晓平, 王刚, 等. 基于深度神经网络复杂场景下的机器人拣选方法[J]. 北京邮电大学学报, 2019, 42(5): 22-28.
HAN X, LIU X P, WANG G, et al. Robotic sorting method in complex scene based on deep neural network[J]. Journal of Beijing University of Posts and Telecommunications, 2019, 42(5): 22-28.
[12] 邹汶材, 刘宝临. 基于图像实例分割的机器人箱体拆垛方法[J]. 计算机工程与应用, 2024, 60(10): 209-216.
ZOU W C, LIU B L. Robot box depalletizing method based on image instance segmentation[J]. Computer Engineering and Applications, 2024, 60(10): 209-216.
[13] YANG X, YANG J, YAN J, et al. SCRDet: towards more robust detection for small, cluttered and rotated objects[C]//Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2019: 8231-8240.
[14] QIAN W, YANG X, PENG S, et al. Learning modulated loss for rotated object detection[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2021, 35(3): 2458-2466.
[15] YANG X, YANG X, YANG J, et al. Learning high-precision bounding box for rotated object detection via Kullback-Leibler divergence[C]//Advances in Neural Information Processing Systems 34, 2021: 18381-18394.
[16] WANG C Y, BOCHKOVSKIY A, LIAO H M. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]//Proceedings of the 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2023: 7464-7475.
[17] PAN X, REN Y, SHENG K, et al. Dynamic refinement network for oriented and densely packed object detection[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 11204-11213.
[18] HOU Q, ZHOU D, FENG J. Coordinate attention for efficient mobile network design[C]//Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 13708-13717.
[19] MA N, ZHANG X, ZHENG H T, et al. ShuffleNet V2: practical guidelines for efficient CNN architecture design[C]// Proceedings of the 15th European Conference on Computer Vision. Cham: Springer, 2018: 122-138.
[20] LIU S, QI L, QIN H, et al. Path aggregation network for instance segmentation[C]//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 8759-8768.
[21] ZHU X, HU H, LIN S, et al. Deformable ConvNets V2: more deformable, better results[C]//Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2019: 9300-9308.
[22] XU Y, FU M, WANG Q, et al. Gliding vertex on the horizontal bounding box for multi-oriented object detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 43(4): 1452-1459.
[23] YANG X, YAN J, FENG Z, et al. R3Det: refined single-stage detector with feature refinement for rotating object[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2021, 35(4): 3163-3171.
[24] ZHU M, HU G, ZHOU H, et al. Arbitrary-oriented ship detection based on RetinaNet for remote sensing images[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2021, 14: 6694-6706.
[25] XIE X, CHENG G, WANG J, et al. Oriented R-CNN for object detection[C]//Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2021: 3500-3509.
[26] BOLYA D, ZHOU C, XIAO F, et al. YOLACT: real-time instance segmentation[C]//Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2019: 9156-9165.