基于特征融合的無(wú)人駕駛多任務(wù)感知算法

孫傳龍　趙紅　崔翔宇　牟亮　徐福良　路來(lái)偉

摘要： 為提高無(wú)人駕駛汽車感知系統(tǒng)硬件資源利用率，構(gòu)建了一種基于特征融合的無(wú)人駕駛多任務(wù)感知算法。采用改進(jìn)的CSPDarknet53作為模型的主干網(wǎng)絡(luò)，通過(guò)構(gòu)建特征融合網(wǎng)絡(luò)與特征融合模塊對(duì)多尺度特征進(jìn)行提取與融合，并以7種常見道路物體的檢測(cè)與可行駛區(qū)域的像素級(jí)分割兩任務(wù)為例，設(shè)計(jì)多任務(wù)模型DaSNet（Detection and Segmentation Net）進(jìn)行訓(xùn)練與測(cè)試。使用BDD100K數(shù)據(jù)集對(duì)YOLOv5s、Faster R-CNN以及U-Net模型進(jìn)行訓(xùn)練，并對(duì)mAP、Dice系數(shù)以及檢測(cè)速度等性能指標(biāo)做出對(duì)比分析。研究結(jié)果表明：DaSNet多任務(wù)模型在道路物體檢測(cè)任務(wù)上，mAP值分別比YOLOv5s和Faster RCNN高出0.5%和4.2%，在RTX2080Ti GPU上達(dá)到121FPS的檢測(cè)速度；在占優(yōu)先權(quán)與不占優(yōu)先權(quán)的可行駛區(qū)域上分割的Dice值相較于U-Net網(wǎng)絡(luò)分別高出了4.4%與6.8%，有較明顯的提升。

關(guān)鍵詞： 無(wú)人駕駛；多任務(wù)；特征融合；道路物體檢測(cè)；可行駛區(qū)域分割

中圖分類號(hào)： TP391文獻(xiàn)標(biāo)識(shí)碼： A

Multi-task Sensing Algorithm for Driverless Vehicle Based on Feature Fusion

SUN Chuanlong1，ZHAO Hong1，CUI Xiangyu2，MU Liang1，XU Fuliang1，LU Laiwei1

Abstract：In order to improve the utilization of hardware resources of driverless vehicle perception system， a multi-task driverless vehicle perception algorithm based on feature fusion is constructed. The improved CSPDarknet53 is used as the backbone network of the model， and multi-scale features are extracted and fused by constructing feature fusion network and feature fusion module. The detection of 7 common road objects and pixel-level segmentation of the driving area are taken as examples. Multi-task DaSNet （Detection and Segmentation Net） is designed for training and testing. In order to compare model performance， BDD100K data set is used to train YOLOv5s， Faster R-CNN and U-NET models， and comparative analysis is made on mAP， Dice coefficient and detection speed and other performance indicators. The results showed that DaSNet multi-task model′s mAP value is 0.5% and 4.2% higher than YOLOv5s and Faster RCNN， respectively， and the detection speed of 121FPS can be achieved on RTX2080Ti GPU. Compared with U-NET network， Dice value of segmentation in priority and non-priority drivable are 4.4% and 6.8% higher， showing an obvious improvement.

Key words： driverless vehicle; multi-task; fature fusion; road object dection;driveable area segmentation

0 引言

目前，無(wú)人駕駛已經(jīng)成為眾多國(guó)家的發(fā)展戰(zhàn)略之一，其中，感知系統(tǒng)［1-3］作為無(wú)人駕駛汽車中不可或缺的部分之一，對(duì)環(huán)境的感知適應(yīng)性、實(shí)時(shí)性直接影響了無(wú)人駕駛汽車的安全性和可靠性，而目標(biāo)檢測(cè)、語(yǔ)義分割作為感知系統(tǒng)中的兩大任務(wù)，其效果將直接影響無(wú)人駕駛汽車決策系統(tǒng)的決策質(zhì)量。

近年來(lái)，深度學(xué)習(xí)［4］在目標(biāo)檢測(cè)、語(yǔ)義分割領(lǐng)域的應(yīng)用取得了重大的突破，這很大程度上歸功于大型數(shù)據(jù)集、計(jì)算機(jī)強(qiáng)大的計(jì)算能力、復(fù)雜的網(wǎng)絡(luò)架構(gòu)和優(yōu)化算法的進(jìn)展。在目標(biāo)檢測(cè)領(lǐng)域，兩階段算法……