基于Faster-RCNN改进的目标检测算法

Improved target detection algorithm based on Faster-RCNN

以Faster-RCNN目标检测算法为基础,用(1×3+3×1+3×3)非对称卷积块替代Faster-RCNN网络模型的3×3卷积核,提出一种基于Faster-RCNN的改进目标检测算法。首先,将残差网络ResNet作为算法骨干,用于提取图像的特征图(Feature map),将Feature map先通过(1×3+3×1+3×3)的卷积核块之后经过两个1×1的卷积核。其次,利用区域建议网络(Regional proposal network,RPN)获得共享特征层的建议框,把建议框映射到卷积的最后一层Feature map上,通过感兴趣区域池化层(Region of interest,RoI)将不同尺寸的锚框进行归一化。最后,利用探测分类概率(Softmax loss)和探测边框回归(Smooth L1 loss)进行训练。本文使用的是PASCAL_VOC数据集,平均查确率(Mean average precision,mAP)结果表明,相比于原始Faster-RCNN算法,mAP值提高了0.38%,相比于RetinaNet算法,mAP值提高了2.68%,相比于YOLOv4算法,mAP值提高了3.41%。

Asymmetric convolution block network is introduced into the Faster-RCNN network model,and it is defined as improved target detection algorithm based on Faster-RCNN.In this algorithm,the convolution kernel of 3×3 in the network model is replaced by the asymmetric convolution block of 1×3+3×1+3×3.Firstly,the residual network ResNet is used as the backbone of the algorithm to extract the feature map of the image.The feature map passes through the convolution kernel block of 1×3+3×1+3×3 and then passes through two convolution kernels of 1×1.Secondly,the regional proposal network(RPN)is used to obtain the suggestion box of shared feature layer,and the suggestion box is mapped to the last feature map of convolution,and the anchor box of different sizes are unified by region of interest(RoI).Finally,the detection classification probability(Softmax loss)and detection border regression(Smooth L1 loss)are used for training.PASCAL_VOC data set is used.The results of mean average precision(mAP)show that the mAP value is increased by 0.38%compared with the original Faster-RCNN algorithm,the mAP value is increased by 2.68%compared with the RetinaNet algorithm,and the mAP value is increased by 3.41%compared with the YOLOv4 algorithm.