ISA Transactions 48 (2009) 389–395

Contents lists available at ScienceDirect

ISA Transactions
journal homepage: www.elsevier.com/locate/isatrans

Distance measurement based on pixel variation of CCD images

Chen-Chien Hsu a,∗ , Ming-Chih Lu b , Wei-Yen Wang a , Yin-Yu Lu b
a
Institute of Applied Electronics Technology, National Taiwan Normal University, 162 He-ping East Road, Section 1, Taipei 10644, Taiwan
b
Department of Electronic Engineering, St. John’s University, 499 Tam King Road, Sec. 4, Tam-Sui, Taipei 25135, Taiwan

article info abstract

Article history: This paper presents a distance measurement method based on pixel number variation of CCD images
Received 24 September 2008 by referencing to two arbitrarily designated points in the image frames. By establishing a relationship
Received in revised form between the displacement of the camera movement along the photographing direction and the difference
25 May 2009
in pixel count between reference points in the images, the distance from an object can be calculated
Accepted 26 May 2009
Available online 21 June 2009
via the proposed method. To integrate the measuring functions into digital cameras, a circuit design
implementing the proposed measuring system in selecting reference points, measuring distance, and
Keywords:
displaying measurement results on CCD panel of the digital camera is proposed in this paper. In
Digital cameras comparison to pattern recognition or image analysis methods, the proposed measuring approach is simple
Distance measurement and straightforward for practical implementation into digital cameras. To validate the performance of the
CCD images proposed method, measurement results using the proposed method and ultrasonic rangefinders are also
Image-based measuring systems presented in this paper.
Pixels © 2009 ISA. Published by Elsevier Ltd. All rights reserved.
Video signal

1. Introduction [18–23] was generally not satisfactory because of the speed con-
straint. Based on a triangular relationship, image-based distance
As far as noncontact distance measurement is concerned, measuring systems (IBDMS) [24–29] were proposed to mea-
ultrasonic-based [1–4] and laser-based [5–11] techniques are sure distance and area using two laser projectors and a CCD
among the most commonly-used methods. Unfortunately, mea- camera. Unfortunately, the two laser projectors needed to be
surement accuracy via the laser- and ultrasonic-based methods precisely aligned with the camera, which inevitably imposed a
heavily depends on the surface reflectivity of the object under mea- critical constraint on the calibration of the measuring system.
surement. If the surface reflectivity is poor, the measuring system Furthermore, measurement accuracy of the IBDMS depended on
generally performs poorly or not at all. These methods also have the distance between the laser projectors. Incorporation of the
measuring system into a digital camera might become cumber-
difficulties in recording images of the objects while measuring dis-
some if higher measuring resolution is required [29]. Because of
tance. Alternatively, imaged-based methods have been proposed
the problems and difficulties via the above-mentioned methods,
for distance measurement by using a CCD (digital camera) [12–15].
accurate and reliable measurements were not always guaranteed
These methods, however, generally require two cameras set up
in real-world applications.
at different positions to capture two different pictures for further
To overcome the problems and difficulties encountered via
analysis. As a result, pattern recognition or image analysis of a
existing image-based distance measuring methods, this paper
whole image frame were required [16,17] to extract features from
presents a distance measurement method based on pixel number
the images for obtaining the distance measurement. Thus, a huge
variation in images of digital cameras by referencing to two ar-
amount of storage capacity and high-speed DSP processors are
bitrarily designated points in the image frames, rather than the
required for system so established, inevitably resulting in disad- laser-projected spots in the image. It is apparent that the actual
vantages in terms of system complexity, processing speed and distance between the reference points will not change no mat-
establishment cost. As a result, the performance of real-time mea- ter how the digital camera moves backwards or forwards along
surements via the pattern recognition or image analysis methods the photographing direction. However, objects in the image frame
captured by the camera do vary in size if the camera moves back-
wards or forwards along the photographing direction. That is, pixel
∗
Corresponding author. Tel.: +886 2 2621 5656; fax: +886 2 2620 9814. counts between the reference points in images will be different if
E-mail addresses: [email protected] (C.-C. Hsu), the digital camera moves along the photographing direction. By es-
[email protected] (M.-C. Lu), [email protected] (W.-Y. Wang). tablishing a relationship between the displacement of the camera
0019-0578/$ – see front matter © 2009 ISA. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.isatra.2009.05.005
390 C.-C. Hsu et al. / ISA Transactions 48 (2009) 389–395

movement and the difference in pixel counts between the refer- if the CCD camera moves forwards along the photographing
ence points in the images at different photographing distances, we direction. Alternatively, the photographing distance
can measure the distance of a remote object. One of the advantages
N (h2 )
in using the proposed measuring approach is that a precise dis- h1 = × 1h − hs (6)
tance between the reference points is not required. Two arbitrarily N (h2 ) − N (h1 )
selected points by the user on the CCD panel covered by the view- can be derived by substituting h2 = h1 − 1h into Eq. (4), if the
ing angle of the camera can be adopted as the reference points in CCD camera moves backwards along the photographing direction.
achieving a reliable measurement. Note that 1h in Eqs. (5) and (6) is the displacement along the
photographing direction due to the movement of the camera,
2. Measurement based on variation of pixel counts of CCD which can be fixed and known a priori. For example, a linear slide
images powered by a motor with the help of position sensors or rotary
encoders for precise detection can be used to drive the CCD camera
to obtain a fixed displacement 1h. As long as hs becomes available,
To equip digital cameras with the function of measuring
we can calculate the photographing distance from an object via
distance while recording images, the performance of the existing Eqs. (5) or (6).
IBDMS needs to be improved from two aspects. The first one is
to remove the constraint on generating two laser beams precisely
2.2. Selection of reference points
formed in parallel as required by the IBDMS. The second one is the
determination of a critical parameter between the optical origin
Fig. 2 shows schematic diagrams illustrating that the variation
and the front end of the camera for various kinds of CCD camera so
of image size of an identical object in image frames depends on the
as to achieve an accurate measurement.
photographing distance. When the CCD camera moves forwards
or backwards along the photographing direction, pixel counts
2.1. Relationship between distance and variation of pixel counts between the two reference points in the image frames will be
different. On the basis of the variation of the pixel counts between
Assume that objects or surfaces to be measured are perpendic- reference points in CCD images, we can derive the photographing
ular to the optical axis of the CCD camera. There exists a close re- distance.
lationship between distance and pixel counts (scanning time) on Assume that points Pa and Pb of an object in the image frame
a scan line of an image frame, as revealed in previous researches are chosen as the reference points by the user. When taking
[24–26]. Fig. 1 shows a schematic diagram of a CCD camera captur- pictures at the photographing distance h2 , pixel counts N (h2 )
ing images at different photographing distances h1 and h2 , in which between the reference points Pa (h2 ) and Pb (h2 ) can be calculated
the distance between the optical origin (OP) and the front end of by N (h2 ) = MR (h2 ) − ML (h2 ), by adjusting the left and right
the CCD camera is hs , D(h1 ) and D(h2 ) are the real-world maximal cursor lines on the CCD panel via control switches. Pixel counts
horizontal distances formed by the field of view of the CCD camera N (h1 ) at the photographing distance h1 can be obtained in exactly
at photographing distance h1 and h2 , respectively, ` is the actual the same way, and will not be re-iterated. When the camera is
distance between Pa and Pb , N (h1 ) and N (h2 ) are the pixel counts set to the function to measure distance, a highlighted horizontal
of ` at the photographing distance h1 and h2 , respectively, and Nmax line on the kth scan line perpendicular to these two vertical cursor
is the maximal pixels in a horizontal scan line of an image frame, lines is generated for easier identification of the reference points.
which is fixed and known a priori irrelevant of photographing dis- When the displacement of the camera movement between the
tances. photographing distances, 1h = h2 − h1 , becomes available, we
can calculate the distances h1 and h2 based on the variation of pixel
Assume that actual distance ` between reference points Pa and
counts 1N = N (h2 ) − N (h1 ).
Pb will not change. We have the following relationship between
pixel counts and distances:
3. Determination of the optical origin for CCD cameras
` D (h1 )
= (1) To construct a measuring system suitable for all kinds of CCD
N (h1 ) Nmax
cameras, the distance hs between the optical origin (OP) and the
` D (h2 ) front end of the camera needs to be established. Fig. 3 shows
= . (2)
N (h2 ) Nmax a proposed method for obtaining an accurate hs for a specific
CCD camera, in which an iris limiting the viewing angle of 2θs is
From Eqs. (1) and (2), we have:
imposed on the CCD camera.
D (h1 ) N (h2 ) With reference to Fig. 3, when the horizontal ruler is positioned
= . (3) at (A1 , A2 ) and (B1 , B2 ), the distance between the front end of the
D (h2 ) N (h1 )
CCD camera and (A1 , A2 ) and (B1 , B2 ) is hm1 and hm2 , respectively.
Because of the displacement, 1h = h1 − h2 , resulting from By a triangular relationship, we have:
the movement of the camera along the photographing direction,
1
two similar isosceles triangles having bases D(h1 ) and D(h2 ), hs + hm2 = Dm2 cot θs (7)
respectively, are formed as shown in Fig. 1. We have: 2
1
N (h2 ) D(h1 ) h1 + hs hs + hm1 = Dm1 cot θs . (8)
= = . (4) 2
N (h1 ) D(h2 ) h2 + hs
Subtracting Eq. (7) from Eq. (8), we obtain:
Substituting h1 = h2 + 1h into Eq. (4), the photographing distance 1
h2 can be obtained as: hm1 − hm2 = (Dm1 − Dm2 ) cot θs
2
N (h1 ) 2 (hm1 − hm2 )
h2 = × 1h − hs , (5) cot θs = . (9)
N (h2 ) − N (h1 ) Dm1 − Dm2
 C.-C. Hsu et al. / ISA Transactions 48 (2009) 389–395 391

Fig. 1. Schematic diagram depicting the relationship between distance and pixel counts at different photographing distances.

a b
Fig. 2. Schematic diagram showing the variation of image size of an identical object in image frames at different photographing distance h1 and h2 , respectively.

Vertical
ruler
CCD camera
OP

Horizontal
ruler
B1 B2
Dm2

A1 A2
Dm1

Fig. 3. Mechanism for obtaining an accurate hs for a CCD camera.

392 C.-C. Hsu et al. / ISA Transactions 48 (2009) 389–395

Fig. 4. Block diagram depicting the hardware design incorporated into a digital
camera to measure and display distance. Fig. 6. Timing sequence in generating the kth scan line control signal VHK .

Alternatively, we can obtain: into the added-on circuit implementing the proposed measuring
hs + hm2 Dm2 system for generating the cursor lines on the CCD panel of the
= digital camera and calculating the distance under measurement.
hs + hm1 Dm1
Thanks to the design of the added-on circuit, extra functions are
by dividing Eq. (7) by Eq. (8). Thus, the critical parameter of the provided for measuring distance without making changes to the
distance between the optical origin and front end of the CCD original processing circuits of the digital camera. Measurement re-
camera for a specific CCD can be obtained as: sults merged into the video image can be simultaneously displayed
hm1 Dm2 − hm2 Dm1 on the CCD panel of the digital camera and stored in a memory card
hs = . (10) together with the video image. The original functionalities of the
Dm1 − Dm2
digital camera are not affected.

4. The proposed measurement system based on pixel variations 4.2. Added-on circuit in implementing the proposed distance measur-
of CCD images ing system

4.1. Hardware design for realizing the proposed distance measuring Fig. 5 shows the schematic diagram of the added-on circuit
system implementing the proposed measuring system in generating the
left and right cursor lines on the CCD panel of the digital camera
To obtain the distances h1 and h2 , images captured by the and calculating the distance under measurement. The video signal
camera at photographing distances h1 and h2 can be stored in a is processed by a Synchronization Separation Circuit to obtain a
memory card for off-line processing in a computer. N (h1 ) and horizontal synchronization signal Hsycn . With suitable selection on
N (h2 ) are then calculated via image processing software based the horizontal synchronization signal Hsycn via a Scan Line Setting
on the selected reference points Pa and Pb . This approach, though circuit, a particular scan line, for example, the kth scan line can
workable, is very inconvenient for practical implementation. In be chosen to produce a kth scan line control signal VHK as shown
this paper, we will propose a hardware design for incorporation in Fig. 6. Signal VHK is then fed into a Level Converter via an OR
into a digital camera to measure distance and display measured gate GA to adjust voltage levels to the largest intensity suitable for
results on CCD panel of the digital camera in addition to full camera merging with the original video signal. As a result, a highlighted
functionalities. horizontal line on the kth scan line will be generated on the CCD
Fig. 4 shows the functional block of an added-on circuit for in- panel of the camera. Because of the Scan Line Setting circuit, users
corporation into a digital camera to measure and display distance. can arbitrarily designate the kth scan line. Therefore, reference
The original video signal is disconnected from the camera and fed points Pa and Pb can be selected anywhere in the CCD panel.

Fig. 5. Schematic diagram of the added-on circuit for incorporation into a digital camera to measure and display distance.
 C.-C. Hsu et al. / ISA Transactions 48 (2009) 389–395 393

Fig. 7. Timing sequence of signals from the comparators and OR gate.

300.83cm 330.83cm

l = 50cm, h2 = 300cm l = 50cm, h1 = 330cm

Fig. 8. Images showing the measured distance on the CCD panel of the digital camera via the proposed method.

With reference to Fig. 5, Vfc is the clock signal for the video touches another position switch, interrupt pin INT1 of the micro-
signal of the digital camera. As a result, the output from the Pixel processor is activated, triggering the interrupt subroutine to read
Counter, Np = 0 ∼ Nmax , is synchronized with the video signal. MR (h1 ) − ML (h1 ) = N (h1 ) from the subtractor into the micro-
The frequency of the input clock to the Left and Right Cursor processor. When both N (h1 ) and N (h2 ) become available, the mi-
Line Counters, Vfm , is far smaller than Vfc to the Pixel Counter. By croprocessor can calculate the photographing distances h1 and h2
adjusting switches SL and SR of the Left and Right Cursor Line via Eqs. (5) and (6). For an NTSC system, MR − ML can be obtained
Counters, clock counts ML and MR from the Left and Right Cursor in every 1/60 s. Thus, photographing distances h1 and h2 can be
Line Counters can be accurately set in such a way that pressing UP promptly calculated.
increases the count, pressing DN decreases the count, and pressing To simultaneously display the measurement on the CCD panel
ST stops the count, respectively. of the digital camera, a Character Storage Memory is used to store
The output from the Pixel Counter, Np , starts counting from 0 the measured results as bitmapped characters. In the beginning of
in the beginning of every scan line and ends up with NP = Nmax . every scan line, contents in the Character Storage Memory are read
Assume that the Left Cursor Line Counter has a clock count ML , out according to the synchronization clock frequency Vfc . Because
Right Cursor Line Counter has a clock count MR and the Pixel the Character Storage Memory has a fixed scanning interval
Counter repeatedly scans from 0 to Nmax . When NP = ML and NP = corresponding onto the CCD panel, characters representing the
MR , outputs from comparators A and B activate high, i.e., VEQ 1 = measurement results will appear on the same positions of the CCD
High and VEQ 2 = High , respectively. That is, output signal from the panel, as will be shown in Figs. 8 and 9.
OR gate GA , Vdk = High , whenever the Left Cursor Line Counter
outputs ML or Right Cursor Line Counter outputs MR for all scan 5. Experimental results
lines. The positive pulses Vdk after level conversion are added into
the original video signal. Therefore, a tiny spot of high intensity In this section, we present experimental results to demonstrate
is generated on the CCD panel of the digital camera for each scan the effectiveness of the proposed measuring system based on pixel
line at horizontal positions ML and MR . The aggregation of the tiny number variation of CCD images. A CCD camera with a resolution
spots for all scan lines at horizontal positions ML and MR will then of 2048 × 1536 pixels is used to conduct the experiment. The
generate two vertical fine lines with highlighted intensity on the critical parameter between the optical origin and the front end
CCD panel, perpendicular to the highlighted horizontal line on the of the camera for the CCD camera is hs = 2.5 cm, which is
kth scan line, as demonstrated in Fig. 7. experimentally obtained by the proposed mechanism in Fig. 3
When the digital camera is moved forwards along the pho- using different sets of Dm1 , Dm2 , Hm1 and Hm2 to obtain a reliable
tographing direction and touches a position switch, the interrupt parameter for hs . Displacement of camera movement, 1h = 30 cm,
pin INT0 of the microprocessor in Fig. 5 is activated, triggering an is adopted for all the experiments. Experiments are conducted
interrupt subroutine to read MR (h2 ) − ML (h2 ) = N (h2 ) from the adapting to practical situations in the use of digital cameras.
subtractor into the microprocessor via I/O port A. When the digital Different reference points are chosen depending on photographing
camera is moved backward along the photographing direction and the ranges under consideration, because a broader range is covered
394 C.-C. Hsu et al. / ISA Transactions 48 (2009) 389–395

547.81cm 577.81cm

l = 150cm, h2 = 540cm l = 150cm, h1 = 570cm

Fig. 9. Images showing the measured distance on the CCD panel of the digital camera via the proposed method.

Table 1
Distance measurement at various photographing distances.
Measured distance (h2 ) 120 cm 180 cm 240 cm 300 cm 360 cm 420 cm 480 cm 540 cm
N (h2 ) 993 667 502 400 334 287 252 225
N (h1 ) 797 573 447 364 309 268 237 213
Estimated distance (cm) (proposed method) 119.49 180.37 241.32 300.83 368.3 420.66 471.5 530
Error % (proposed method) −0.43 0.21 0.55 0.28 2.31 0.16 −1.77 −1.85
Estimated distance (cm) (ultrasonic rangefinder) 122 178 238 298 356 414 474 534
Error % (ultrasonic rangefinder) 1.67 −1.11 −0.83 −0.67 −1.11 −1.43 −1.25 −1.11
Note: Distance between Pa and Pb is 50 cm.

Table 2
Distance measurement at various photographing distances.
Measured distance (h2 ) 300 cm 360 cm 420 cm 480 cm 540 cm 600 cm 660 cm 720 cm
N (h2 ) 754 632 542 475 423 380 348 318
N (h1 ) 690 584 506 447 401 362 333 305
Estimated distance (cm) (proposed method) 305.56 362.5 419.17 476.43 544.32 600.83 663.5 701.35
Error % (proposed method) 1.85 0.69 −0.20 −0.74 0.8 0.14 0.53 −2.59
Estimated distance (cm) (ultrasonic rangefinder) 298 356 414 474 534 594 654 714
Error % (ultrasonic rangefinder) −0.67 −1.11 −1.43 −1.25 −1.11 −1.00 −0.91 −0.83
Note: Distance between Pa and Pb is 100 cm.

Table 3
Distance measurement at various photographing distances.
Measured distance (h2 ) 360 cm 420 cm 480 cm 540 cm 600 cm 660 cm 720 cm 780 cm
N (h2 ) 925 794 695 619 557 508 468 431
N (h1 ) 855 742 655 587 531 486 449 415
Estimated distance (cm) (proposed method) 363.93 425.58 476.77 547.81 610.19 660.23 706.45 775.63
Error % (proposed method) 1.09 1.33 −0.67 1.45 1.69 0.03 −1.89 −0.56
Estimated distance (cm) (ultrasonic rangefinder) 356 414 474 534 594 654 714 772
Error % (ultrasonic rangefinder) −1.11 −1.43 −1.25 −1.11 −1.00 −0.91 −0.83 −1.03
Note: Distance between Pa and Pb is 150 cm.

when shooting at a longer distance while it is narrower when the proposed measuring method. Disadvantages of conventional
shooting at a shorter distance. The rationale is that we choose measuring methods which required prior knowledge of camera
suitable reference points to allow easier identification of the parameters, such as focus of the camera and horizontal and vertical
objects in the image frames. view angles, are also avoided. Furthermore, problems due to poor
Table 1 shows measurement results at different photographing reflection surfaces when using traditional laser- and ultrasonic-
distances h2 ranging from 120 cm to 540 cm using the proposed based techniques are prevented via the proposed measuring
method by moving the camera forward with a displacement 1h = method. As the measuring distances increase, suitable reference
points, PA and PB , need to be selected. Tables 2 and 3 show the
30 cm. The actual distance between the reference points, PA and PB ,
measurement results at larger photographing distances ranging
is 50 cm. The first row of the table is the measured distance using a
from 300 cm to 780 cm, with an actual distance of 100 cm and 150,
ruler and the fourth row of the table shows the estimated distance
respectively, between the reference points. As the photographing
via the proposed measuring method. N (h1 ) and N (h2 ) are the pixel distance increases, differentiality on variation of pixel counts
counts between the reference points at photographing distances h1 between reference points reduces due to the resolution of the
and h2 , respectively. As a comparison, measurement results using CCD camera. As a result, the measurement error becomes la-
an ultrasonic rangefinder are also shown in this table. As shown in rger for long-distance measurements. Generally speaking, the
Table 1, the measuring errors for various photographing distances proposed method is suitable for applications with short-distance
fall within 2.31% via the proposed method, which are comparative measurements, such as obstacle avoidance or localization in
to 1.67% via the ultrasonic rangefinder. Laser projectors, which robotic applications. As evidence to show the advantage of the
were essential in the IBDMS [18–22,29], are no longer required via proposed method to merge measurement results into the video
 C.-C. Hsu et al. / ISA Transactions 48 (2009) 389–395 395

image for displaying on the CCD panel, images measuring distances [7] Shin HT. Vehicles crashproof laser radar. M.S. thesis. Taiwan: Optical Science
of 300 cm and 540 cm in practice are demonstrated in Figs. 8 Center, Nation Central Univ.; 2000.
[8] Peng CC. A compact digital image sensing distance and angle measuring
and 9, respectively. device. M.S. thesis. Taoyuan County (Taiwan): Optical Science Center, Nation
Central Univ.; 2001.
[9] Chavand F, Colle E, Chekar Y, Ni FC. 3D measurements using a video camera
6. Conclusions
and a ranger finder. IEEE Transactions on Instrumentation and Measurement
1997;46:1229–35.
In this paper, a measuring system based on pixel number vari- [10] Tiedeke J, Schable P, Rille E. Vehicle distance sensor using a segmented IR
ation of CCD images is presented, equipping digital cameras with laser beam. In: IEEE 40th vehicular technology conference. CH2846-4. 1990.
p. 107–12.
distance measurement functions while recording images. Because [11] Culshaw B, Pierce G, Pan J. Non-contact measurement of the mechanical
of the design of the added-on circuits, which generate cursor lines properties of materials using an all-optical technique. IEEE Sensors Journal
on the CCD panel for adjustment to calculate the distance under 2003;3:62–70.
[12] Miwa M, Ishii M, Koike Y, Sato M. Screen projection camera for ranging far
measurement, the proposed method is capable of measuring dis-
away objects. In: Proceedings of 15th international conference on pattern
tance and displaying the measured results simultaneously on the recognition. 2000. p. 4744–7.
CCD panel without changing the original processing circuits of the [13] Egami T, Oe S, Terada K, Kashiwagi T. Three dimensional measurement using
digital camera. To provide better accuracy for distance measure- color image and movable CCD system. In: The 27th annual conference of the
IEEE industrial electronic society. 2001. p. 1932–6.
ment, the distance between the optical origin and the front end [14] Cano-Garcia A, Lazaro JL, Fernaindez PR. Simplified method for radiometric
of the CCD camera is also determined by a proposed mechanism calibration of an array camera. In: Proceedings of the IEEE international
for taking into consideration during the measurement. As a result, symposium on intelligent signal processing. 2007. p. 1–5.
[15] Mataix C, Lazaro JL, Gardel A, Mateos R. Sensor for environment wide capture
any CCD cameras can be used to establish the proposed measuring
with linear response. In: Emerging technologies and factory automation, 7th
system. Experimental results have shown that the measurement IEEE international conference. 1999. p. 571–8.
accuracy using the proposed method is comparable to that of an ul- [16] Kanade T, Kano H, Kimuram S. Development of a video-rate stereo machine.
trasonic rangefinder for short-distance measurement, overcoming In: Proc. 1995 IEEE/RSJ int. conf. on intelligent robots and systems. vol. 3. 1995.
p. 95–100.
the problems and difficulties encountered in conventional image- [17] Tanaka Y, Gofuku A, Nagai I, Mohamed A. Development of a compact video-
based distance measuring methods. As shown in the paper, the rate range finder and its application. In: Proc. 3rd int. conf. on advanced
proposed measuring system has demonstrated itself as a simple mechatronics. 1998. p. 97–102.
[18] Mertzios BG, Tsirikolias K. Applications of coordinate logic filters in image
way in measuring distances while simultaneously recording im-
analysis and pattern recognition. In: Proc. int. symp. image and signal
ages for applications which require short-distance measurement. processing and analysis. 2001. p. 125–30.
[19] Hong Y. Image analysis for digital media applications. IEEE Computer Graphics
Acknowledgements and Applications 2001;21:18–26.
[20] Cucchiara R, Piccardi M, Mello P. Image analysis and rule-based reasoning for
a traffic monitoring system. IEEE Transactions on Intelligent Transportation
The authors wish to express their gratitude to the anonymous Systems 2000;1:119–30.
reviewers for their valuable recommendations and comments to [21] Paschalakis S, Lee P. Pattern recognition in grey level images using moment
based invariant features. In: International conference on image processing and
improve this manuscript. This work was partially supported by
its applications. vol. 1, 465. 1999. p. 245–9.
the National Science Council, Taiwan, under Grant NSC 94-2218- [22] Katsoulas D, Werber A. Edge detection in range images of piled box-like
E-129-002. objects. In: Proceedings of the 7th international conference on pattern
recognition. vol. 2. 2004. p. 80–4.
References [23] Garcia MA, Solanas A. Estimation of distance to planar surfaces and type
of material with infrared sensors. In: Proceedings of the 7th international
[1] Yasuda A, Kuwashima S, Kanai Y. A shipborne-type wave-height meter for conference on pattern recognition. vol. 1. 2004. p. 745–8.
oceangoing vessels, using microwave Doppler radar. IEEE Journal of Oceanic [24] Lu MC, Wang WY, Lan HH. Image-based height measuring systems for liquid
Engineering 1985;10(2):138–43. or particles in tanks. In: IEEE International conference on networking, sensing
[2] Carullo A, Parvis M. An ultrasonic sensor for distance measurement in and control. 2004. p. 24–9.
automotive applications. IEEE Sensors Journal 2001;1:143–7. [25] Lu MC, Wang WW, Chu CY. Optical-based distance measuring system (ODMS).
[3] Carullo A, Ferraris F, Graziani S. Ultrasonic distance sensor improvement In: The eighth international conference on automation technology. 2005.
using a two-level neural network. IEEE Transactions on Instrumentation and p. 282–3.
Measurement 1966;45:667–82. [26] Lu MC, Hsu YT, Tu JF. Automobile SSD alert system. In: Proceedings of the
[4] Song KT, Tang WH. Environment perception for a mobile robot using ninth international conference on distributed multimedia systems. 2003.
double ultrasonic sensor and a CCD camera. IEEE Transactions on Industrial p. 806–9.
Electronics 1996;43:372–9. [27] Lu MC. Image-based height measuring system for liquid or particles in tanks.
[5] KlimKov YM. A laser polarmetric sensor for measuring angular displacement ROC patent of invention; 2004. Patent number 201536.
of objects. In: Conference on lasers and electro-optics Europe. CLEO/Europe. [28] Lu MC. BUNDESREPUBLIK DEUTSCHLAND, Vorrichtung zum Messen des
1996. p. 190. Fullstands von Lagergut. Nr.203 19 293.1, IPC: G01F 23/292, 12.03.2003, TW
[6] Svirdov SA, Sterlyagov MS. Sea surface slope statistics measured by laser 92105320.
sensor. In: Proceedings of oceans engineering for today’s technology and [29] Lu MC, Wang WY, Chu CY. Image-based distance and area measuring systems.
tomorrow’s preservation. vol. 1. 1994. p. 900–5. IEEE Sensors Journal 2006;6:495–503.