python使用AprilTag 3
最近想测试一下AprilTag精度，看看能不能用的上。
1 安装
法1：github源码编译安装（放弃）
一开始找到了AprilTag 3的官方github网址https://github.com/AprilRobotics/apriltag，但是按着操作下来不会把python库链接到conda虚拟环境，所以尝试看看有没有别的办法。
法2：pip
在pypi.org官网上找到了两种apriltag安装方法
法2.1：pip install apriltag（放弃）
https://pypi.org/project/apriltag/，看到写的更新时间为2018年，所以应该不是最新版的AprilTag 3，故放弃
法2.2：pip install pyapriltags（可用）
再一番寻找，找到了2023更新的版本，https://pypi.org/project/pyapriltags/，应该比较靠谱，并且有如下描述，可以判断是AprilTag原作者AprilRobotics开发的官方库。
These are Python bindings for the Apriltags 3 library developed by AprilRobotics.
这个包的github链接为https://github.com/WillB97/pyapriltags，可以git clone一下里面有pyapriltags/test/test.py测试例程序。
运行时应当注意yaml包版本，过新版本可能会出现报错。
pip install pyyaml==5.3
2 测试
进入pyapriltags/test/文件夹下，终端中执行
python test.py
得到如下结果，还没有集成到摄像头里面，但应该是跑通了
TESTING WITH A SAMPLE IMAGE
[Detection object:
tag_family = b'tag36h11'
tag_id = 60
tag_size = 0.065
hamming = 0
decision_margin = 24.732816696166992
homography = [[1.12852846e+01 1.35218628e+01 1.50604021e+02]
[4.12931891e-01 1.34012105e+01 9.97275963e+01]
[7.54358942e-03 3.97360209e-02 1.00000000e+00]]
center = [150.60402144 99.72759634]
corners = [[148.07374573 109.20044708]
[167.49220276 108.4158783 ]
[153.30262756 89.62454987]
[132.03965759 90.17698669]]
pose_R = [[ 0.97275506 0.17189059 0.1555674 ]
[ 0.04820706 0.50638753 -0.86095746]
[-0.22676788 0.84500017 0.48430469]]
pose_t = [[-0.4970639 ]
[-0.30799945]
[ 0.9148219 ]]
pose_err = 1.4234227187581564e-06
, Detection object:
tag_family = b'tag36h11'
tag_id = 82
tag_size = 0.065
hamming = 0
decision_margin = 24.040637969970703
homography = [[ 8.02054019e+00 3.10314575e+00 4.66621306e+02]
[-8.49221724e-01 1.06309763e+01 8.20201682e+01]
[-3.96118917e-03 1.89261753e-02 1.00000000e+00]]
center = [466.62130564 82.02016819]
corners = [[451.37316895 91.40827179]
[470.70095825 90.44836426]
[482.58377075 72.19226074]
[462.41769409 73.33588409]]
pose_R = [[ 0.96003341 -0.15463188 -0.2332913 ]
[-0.14024405 0.45556232 -0.87908736]
[ 0.24221365 0.87667095 0.41566885]]
pose_t = [[ 0.37945231]
[-0.37270953]
[ 0.95850859]]
pose_err = 3.8481665025688905e-07
, Detection object:
tag_family = b'tag36h11'
tag_id = 318
tag_size = 0.065
hamming = 0
decision_margin = 49.61515426635742
homography = [[-1.18572593e+01 3.08253464e+01 5.37597981e+02]
[-2.09246950e+01 1.09739259e+01 3.75029839e+02]
[-1.38257708e-02 3.30089272e-02 1.00000000e+00]]
center = [537.59798111 375.02983943]
corners = [[554.31921387 388.7227478 ]
[546.09033203 358.20751953]
[519.23352051 359.99130249]
[528.7734375 392.51019287]]
pose_R = [[-0.08183852 0.99100632 -0.1058722 ]
[-0.92256889 -0.03513853 0.38422901]
[ 0.37705318 0.12911913 0.91714729]]
pose_t = [[0.46192837]
[0.36769891]
[0.76352648]]
pose_err = 5.429954298671284e-06
, Detection object:
tag_family = b'tag36h11'
tag_id = 328
tag_size = 0.065
hamming = 0
decision_margin = 30.794836044311523
homography = [[ 1.15647124e+01 -1.31398782e+00 3.19180294e+02]
[ 5.44862875e-01 6.30841077e+00 9.93377081e+01]
[ 8.41156717e-03 -4.81280377e-03 1.00000000e+00]]
center = [319.18029385 99.33770814]
corners = [[310.40652466 106.50978088]
[328.24972534 105.81019592]
[327.72503662 92.35285187]
[310.04534912 92.81846619]]
pose_R = [[ 9.99085975e-01 4.27458458e-02 -8.01169633e-05]
[-3.87858311e-02 9.07314404e-01 4.18660042e-01]
[ 1.79686689e-02 -4.18274269e-01 9.08143030e-01]]
pose_t = [[-0.0521486 ]
[-0.41622112]
[ 1.23252283]]
pose_err = 3.00796680270929e-07
, Detection object:
tag_family = b'tag36h11'
tag_id = 387
tag_size = 0.065
hamming = 0
decision_margin = 39.92521286010742
homography = [[-1.31081417e+01 6.81016759e+00 1.79359887e+02]
[ 3.18874693e-01 -6.31109107e+00 3.22140884e+02]
[-1.26834699e-03 2.44129945e-02 1.00000000e+00]]
center = [179.35988748 322.14088375]
corners = [[194.28860474 307.61105347]
[169.14706421 308.99703979]
[163.6441803 337.43667603]
[190.05665588 335.90756226]]
pose_R = [[-0.99045956 -0.03036217 0.13441724]
[ 0.08804465 -0.88980218 0.4477725 ]
[ 0.10600941 0.45533528 0.88398631]]
pose_t = [[-0.36597334]
[ 0.26055146]
[ 0.79989467]]
pose_err = 2.613317301680534e-07
]
TESTING WITH ROTATION IMAGES
Testing image test_image_rotation_0.png
[[ 0.00855086]
[-0.03378784]
[ 0.20410712]] 0.3 0.0 0.05
[[ 0.99958323 -0.02017469 0.02064826]
[ 0.02462415 0.96918519 -0.2450994 ]
[-0.01506718 0.2455057 0.96927805]] 0.0 0.0 0.0
Testing image test_image_rotation_10.png
[[ 0.00899355]
[-0.03396286]
[ 0.20464217]] 0.3 0.0 0.05
[[ 0.99553509 -0.01814129 -0.09263244]
[-0.00525816 0.96917553 -0.24631514]
[ 0.09424557 0.24570244 0.96475286]] 0.0 0.0 10.0
Testing image test_image_rotation_20.png
[[ 0.01026933]
[-0.03436307]
[ 0.20649642]] 0.3 0.0 0.05
[[ 0.96053902 -0.00980882 -0.27797227]
[-0.05970563 0.96881099 -0.24050054]
[ 0.27166161 0.24760666 0.92999511]] 0.0 0.0 20.0
Testing image test_image_rotation_30.png
[[ 0.01173215]
[-0.03442039]
[ 0.20700676]] 0.3 0.0 0.05
[[ 0.9001428 -0.01497421 -0.43533746]
[-0.09601522 0.96800711 -0.23182602]
[ 0.42488117 0.25047555 0.86990689]] 0.0 0.0 30.0
Testing image test_image_rotation_40.png
[[ 0.012491 ]
[-0.03455821]
[ 0.20762498]] 0.3 0.0 0.05
[[ 0.81339466 -0.00986884 -0.58162852]
[-0.13883209 0.96766979 -0.21057264]
[ 0.56490245 0.25202736 0.78572732]] 0.0 0.0 40.0
Testing image test_image_rotation_50.png
[[ 0.01392287]
[-0.03490017]
[ 0.20890351]] 0.3 0.0 0.05
[[ 0.68644151 -0.01404759 -0.72704932]
[-0.1685925 0.96949743 -0.17790814]
[ 0.70737163 0.2446986 0.66313496]] 0.0 0.0 50.0
Testing image test_image_rotation_60.png
[[ 0.01452253]
[-0.0352799 ]
[ 0.2101661 ]] 0.3 0.0 0.05
[[ 0.55337696 -0.00760598 -0.83289621]
[-0.19765949 0.9701953 -0.14018492]
[ 0.80913822 0.24220495 0.53538033]] 0.0 0.0 60.0
Testing image test_image_rotation_70.png
[[ 0.01504444]
[-0.03561829]
[ 0.211431 ]] 0.3 0.0 0.05
[[ 0.39452178 -0.00315769 -0.91888116]
[-0.22194234 0.97005917 -0.09862454]
[ 0.89168053 0.24284816 0.38200865]] 0.0 0.0 70.0
Testing image test_image_rotation_-10.png
[[ 0.00737225]
[-0.03395985]
[ 0.20490716]] 0.3 0.0 0.05
[[ 0.98644522 -0.02056676 0.16279692]
[ 0.05954953 0.96934681 -0.23837075]
[-0.15290416 0.24483417 0.95743227]] 0.0 0.0 -10.0
Testing image test_image_rotation_-20.png
[[ 0.00605699]
[-0.03420006]
[ 0.20600093]] 0.3 0.0 0.05
[[ 0.94790959 -0.01801473 0.31802968]
[ 0.09644923 0.96776511 -0.23265478]
[-0.30358682 0.25120942 0.91909133]] 0.0 0.0 -20.0
Testing image test_image_rotation_-30.png
[[ 0.00486003]
[-0.03435319]
[ 0.20670285]] 0.3 0.0 0.05
[[ 0.87250798 -0.01693225 0.48830639]
[ 0.14092693 0.96564692 -0.21832458]
[-0.46783483 0.25930546 0.84492085]] 0.0 0.0 -30.0
Testing image test_image_rotation_-40.png
[[ 0.00344044]
[-0.03460087]
[ 0.20756462]] 0.3 0.0 0.05
[[ 0.76359383 -0.01288116 0.64556838]
[ 0.17905511 0.9648149 -0.19253954]
[-0.62037386 0.26261432 0.73903315]] 0.0 0.0 -40.0
Testing image test_image_rotation_-50.png
[[ 0.00257899]
[-0.0345974 ]
[ 0.20762641]] 0.3 0.0 0.05
[[ 0.62773692 -0.01039825 0.77835611]
[ 0.21986308 0.96156606 -0.16447172]
[-0.7467306 0.27437674 0.6058967 ]] 0.0 0.0 -50.0
Testing image test_image_rotation_-60.png
[[ 0.00175676]
[-0.0346137 ]
[ 0.20774601]] 0.3 0.0 0.05
[[ 0.47993701 -0.00589266 0.87728316]
[ 0.25294356 0.9584421 -0.13194053]
[-0.84004763 0.28522627 0.46148234]] 0.0 0.0 -60.0
Testing image test_image_rotation_-70.png
[[ 0.00084808]
[-0.03494523]
[ 0.20876179]] 0.3 0.0 0.05
[[ 0.31758472 -0.00758214 0.94819959]
[ 0.27671098 0.9571842 -0.08502607]
[-0.90695699 0.28938022 0.30608514]] 0.0 0.0 -70.0
AVG time per detection: 0.01741302808125814
TESTING WITH MULTIPLE TAGS IMAGES
Testing image test_image_multiple_01.png
6 tags found: [22, 24, 58, 85, 144, 198]
Testing image test_image_multiple_02.png
5 tags found: [22, 24, 85, 144, 198]
Testing image test_image_multiple_03.png
6 tags found: [22, 24, 58, 85, 144, 198]
Testing image test_image_multiple_04.png
6 tags found: [22, 24, 58, 85, 144, 198]
Testing image test_image_multiple_05.png
4 tags found: [22, 24, 85, 198]
Testing image test_image_multiple_06.png
6 tags found: [22, 24, 58, 85, 144, 198]
Testing image test_image_multiple_07.png
6 tags found: [22, 24, 58, 85, 144, 198]
Testing image test_image_multiple_08.png
3 tags found: [85, 144, 198]
Testing image test_image_multiple_09.png
5 tags found: [22, 24, 58, 85, 144]
Testing image test_image_multiple_10.png
4 tags found: [24, 58, 85, 198]
AVG time per detection: 0.01161332130432129
贴一下测试例程：
from pyapriltags import Detector
import numpy
import os
test_images_path = 'test_files'
visualization = True
try:
import cv2
except:
raise Exception('You need cv2 in order to run the demo. However, you can still use the library without it.')
try:
from cv2 import imshow
except:
print("The function imshow was not implemented in this installation. Rebuild OpenCV from source to use it")
print("VIsualization will be disabled.")
visualization = False
try:
import yaml
except:
raise Exception('You need yaml in order to run the tests. However, you can still use the library without it.')
at_detector = Detector(families='tag36h11',
nthreads=1,
quad_decimate=1.0,
quad_sigma=0.0,
refine_edges=1,
decode_sharpening=0.25,
debug=0)
with open(test_images_path + '/test_info.yaml', 'r') as stream:
parameters = yaml.load(stream)
#### test WITH THE SAMPLE IMAGE ####
print("\n\nTESTING WITH A SAMPLE IMAGE")
img = cv2.imread(test_images_path+'/'+parameters['sample_test']['file'], cv2.IMREAD_GRAYSCALE)
cameraMatrix = numpy.array(parameters['sample_test']['K']).reshape((3,3))
camera_params = ( cameraMatrix[0,0], cameraMatrix[1,1], cameraMatrix[0,2], cameraMatrix[1,2] )
if visualization:
cv2.imshow('Original image',img)
tags = at_detector.detect(img, True, camera_params, parameters['sample_test']['tag_size'])
print(tags)
color_img = cv2.cvtColor(img, cv2.COLOR_GRAY2RGB)
for tag in tags:
for idx in range(len(tag.corners)):
cv2.line(color_img, tuple(tag.corners[idx-1, :].astype(int)), tuple(tag.corners[idx, :].astype(int)), (0, 255, 0))
cv2.putText(color_img, str(tag.tag_id),
org=(tag.corners[0, 0].astype(int)+10,tag.corners[0, 1].astype(int)+10),
fontFace=cv2.FONT_HERSHEY_SIMPLEX,
fontScale=0.8,
color=(0, 0, 255))
if visualization:
cv2.imshow('Detected tags', color_img)
k = cv2.waitKey(0)
if k == 27: # wait for ESC key to exit
cv2.destroyAllWindows()
#### test WITH THE ROTATION IMAGES ####
import time
print("\n\nTESTING WITH ROTATION IMAGES")
time_num = 0
time_sum = 0
test_images_path = 'test_files'
image_names = parameters['rotation_test']['files']
for image_name in image_names:
print("Testing image ", image_name)
ab_path = test_images_path + '/' + image_name
if(not os.path.isfile(ab_path)):
continue
groundtruth = float(image_name.split('_')[-1].split('.')[0]) # name of test_files image should be set to its groundtruth
parameters['rotation_test']['rotz'] = groundtruth
cameraMatrix = numpy.array(parameters['rotation_test']['K']).reshape((3,3))
camera_params = ( cameraMatrix[0,0], cameraMatrix[1,1], cameraMatrix[0,2], cameraMatrix[1,2] )
img = cv2.imread(ab_path, cv2.IMREAD_GRAYSCALE)
start = time.time()
tags = at_detector.detect(img, True, camera_params, parameters['rotation_test']['tag_size'])
time_sum+=time.time()-start
time_num+=1
print(tags[0].pose_t, parameters['rotation_test']['posx'], parameters['rotation_test']['posy'], parameters['rotation_test']['posz'])
print(tags[0].pose_R, parameters['rotation_test']['rotx'], parameters['rotation_test']['roty'], parameters['rotation_test']['rotz'])
print("AVG time per detection: ", time_sum/time_num)
#### test WITH MULTIPLE TAGS IMAGES ####
print("\n\nTESTING WITH MULTIPLE TAGS IMAGES")
time_num = 0
time_sum = 0
image_names = parameters['multiple_tags_test']['files']
for image_name in image_names:
print("Testing image ", image_name)
ab_path = test_images_path + '/' + image_name
if(not os.path.isfile(ab_path)):
continue
cameraMatrix = numpy.array(parameters['multiple_tags_test']['K']).reshape((3,3))
camera_params = ( cameraMatrix[0,0], cameraMatrix[1,1], cameraMatrix[0,2], cameraMatrix[1,2] )
img = cv2.imread(ab_path, cv2.IMREAD_GRAYSCALE)
start = time.time()
tags = at_detector.detect(img, True, camera_params, parameters['multiple_tags_test']['tag_size'])
time_sum+=time.time()-start
time_num+=1
tag_ids = [tag.tag_id for tag in tags]
print(len(tags), " tags found: ", tag_ids)
color_img = cv2.cvtColor(img, cv2.COLOR_GRAY2RGB)
for tag in tags:
for idx in range(len(tag.corners)):
cv2.line(color_img, tuple(tag.corners[idx-1, :].astype(int)), tuple(tag.corners[idx, :].astype(int)), (0, 255, 0))
cv2.putText(color_img, str(tag.tag_id),
org=(tag.corners[0, 0].astype(int)+10,tag.corners[0, 1].astype(int)+10),
fontFace=cv2.FONT_HERSHEY_SIMPLEX,
fontScale=0.8,
color=(0, 0, 255))
if visualization:
cv2.imshow('Detected tags for ' + image_name , color_img)
k = cv2.waitKey(0)
if k == 27: # wait for ESC key to exit
cv2.destroyAllWindows()
print("AVG time per detection: ", time_sum/time_num)
扫一扫
3974
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
