Reading the Draft Marks of the Ship

ane.

Introduction

In marine transportation, a draft survey is a means to determine the quantity of bulk cargo. For reducing cargo shortage claims, typhoon reading must exist accurate and fair to a shipper and a receiver. The draft reading has been conventionally conducted by professional surveyors, merely it may not be accurate because it is based on visual observation. Moreover, if the surveyor is not independent of the shipper or the receiver, the draft reading may not be off-white.

The use of automatic typhoon reading systems provides fair draft reading. Automatic and accurate draft reading systems using sensors, such as a laser distance sensor^ane and a liquid level optical sensor,² have been proposed. Still, to read the draft of a ship, the send needs to be equipped with the sensors in advance. Low price and high flexibility are required for ease of draft reading.

Several epitome-based draft reading systems have been thus proposed to accomplish a low toll and high flexibility in draft reading. Okamoto et al.ⁱⁱⁱ proposed an image-based draft reading method using Otsu's binarization method⁴ and a frame differencing technique. The method segments typhoon marks past Otsu's binarization method, assuming that the whole of the observed image can be roughly classified into two classes, i.due east., typhoon marks and a hull. However, it sometimes fails considering nearly all observed images include other regions than draft marks and the hull, such as a sea surface and shadows. The method takes the difference of two consecutive frames, accumulates the binary departure images, so estimates the waterline from the body of water surface segmented by the accumulated binary difference image. Therefore, the estimated waterline tends to be the highest during the observation time. Ran et al.⁵ proposed a method which estimates the waterline using Hough transform after applying Canny border detection.^{half dozen} The method fails when other lines, such every bit underwater projections of the ship's hull, are detected in the sea surface.

In this paper, nosotros suggest an prototype-based typhoon reading method that uses a draft mark segmentation and estimates the waterline for every frame. To improve an accuracy of the typhoon mark segmentation, we detect draft marks with morphological operations and binarize local images around draft marks. Since these local images can exist well classified into draft marks and the ship's hull, draft marks are accurately segmented from the local images by Otsu's binarization method. Next, we detect the waterline using Canny edge detection for every frame. We and so utilize a robust estimation to fit a straight line to the Canny border image in a limited region effectually the typhoon mark for decreasing racket furnishings and regard the straight line as the waterline. Since the waterline is estimated for every frame, we can avert the misestimation due to aggregating of difference images. Moreover, nosotros tin can efficiently remove noisy edges such every bit projections and scars using the property that the noisy edges remain stationary relative to typhoon marks. The proposed method is the first study that provides all the five steps needed for typhoon reading; draft mark segmentation, draft mark recognition, waterline detection, waterline estimation, and typhoon adding. Another salient feature of the proposed method is that the understanding of a shipper and a receiver can be obtained past emulating surveyors' draft reading procedure. The accuracy of draft reading has been evaluated by using a towing tank to show the effectiveness of the proposed method. It is also shown that accurate typhoon reading has been achieved in a real-world scene.

2.

Conventional Method

This section explains 2 types of draft marks and gives a cursory clarification of the conventional draft marker division and waterline estimation,³ which are closely related to the proposed method.

2.one.

Typhoon Marker

Figure 1 shows two typical types of draft marks, representing iii.6, 3.8, and iv.0 m. The size of each draft mark is 10-cm high and ii-cm broad.⁷ Type 1 uses "M" to correspond meter, while type 2 uses only numbers. The bottom of each draft mark indicates the draft. For case, if a waterline touches the bottom of "4M," the draft is 4.0 grand.

Fig. one

Two typical types of draft marks: (a) type i and (b) type two.

2.2.

Draft Mark Segmentation

The conventional typhoon marking segmentation method segments draft marks by Otsu'south binarization method. It may be effective for images consisting only of typhoon marks and a hull, because the grayscale histogram of the image becomes bimodal. However, almost all observed images include other regions than draft marks and the hull, such as body of water surface and shadows.

Figures 2(a) and 2(b) show the observed image and its grayscale histogram, respectively. We captured the image from a wharf using a hand-held photographic camera for adding photographic camera shake effect, because surveyors often read drafts on a pitching and rolling boat. We tin verify that the histogram is not bimodal due to the influence of the sea surface and shadows. Figure two(c) shows the binary image obtained by Otsu's binarization method, where the threshold value is 113. Nosotros see that the typhoon marking "ii" above "2M" is not separated from the hull.

Fig. 2

Conventional typhoon mark segmentation. (a) Original paradigm, (b) grayscale histogram, and (c) binary epitome.

ii.3.

Waterline Interpretation

The conventional waterline estimation detects a moving sea surface by a frame differencing technique. More than concretely, information technology takes the difference of ii consecutive frames, binarizes the deviation image with threshold $\theta$ , and and so accumulates the binary difference images during $n$ frames. And then the conventional method binarizes the horizontal mean of the accumulated binary paradigm to segment the body of water surface and regards the upper edge of the segmented sea surface as the waterline. Nonetheless, if the sea surface moves slowly, we cannot accurately detect the waterline from the frame deviation. Moreover, the estimated waterline tends to be the highest during the ascertainment time. This misestimation is acquired by accumulation of binary difference images.

Figure three shows the result of the waterline interpretation, where we set $\theta =60$ and $n=600$ . Figure iii(a) is the aggregating of binary divergence images of the video sequence captured in a towing tank. The white region in Fig. three(b) denotes the segmented body of water surface. Figure three(c) shows the estimated waterline superimposed on a photograph taken in the absence of a wave. We run into that the estimated waterline is higher than the true ane.

Fig. 3

Conventional waterline interpretation. (a) Accumulation of binary deviation images, (b) segmented bounding main surface, and (c) estimated waterline.

three.

Proposed Method

Figure 4 shows the flowchart of the proposed method. It consists of five steps: (1) typhoon mark segmentation, (two) draft mark recognition, (three) waterline extraction, (4) waterline estimation, and (5) draft calculation.

Fig. four

Flowchart of the proposed method.

three.1.

Draft Mark Segmentation

The procedure of draft mark partition is farther divided into three steps: draft mark detection, local thresholding, and removing noise segments.

3.1.1.

Typhoon marking detection

Top-hat transform^viii is one of the morphological operations for detecting white or black objects that are smaller than a structuring element. We apply the top-lid transform to find draft marks since they are thin white or blackness objects. The white top-chapeau transform for detecting white draft marks is defined by Eq. (ane), and the black peak-lid transform for detecting blackness draft marks is defined by Eq. (2):

Eq. (ane)

$$\{\begin{array}{l}{I}_{\text{erosion}}(x,y)=\underset{\sqrt{m^2+n^{\mathrm{two}}}\le p}{\min }I(\mathrm{ten}+m,y+\mathrm{northward}),\\ I_{\text{opening}}(x,y)=\underset{\sqrt{{\mathrm{1000}}^2+n^2}\le p}{\max }{I}_{\text{erosion}}(x+m,y+n),\\ I_{\text{white}}(x,y)=I(x,y)-I_{\text{opening}}(x,y),\end{array}$$

Eq. (ii)

$$\{\begin{array}{l}{I}_{\text{dilation}}(x,y)=\underset{\sqrt{{\mathrm{thou}}^2+n^2}\le p}{\max }I(x+\mathrm{yard},y+\mathrm{north}),\\ I_{\text{closing}}(x,y)=\underset{\sqrt{m^2+n^{\mathrm{two}}}\le p}{\min }{I}_{\text{dilation}}(x+m,y+\mathrm{north}),\\ I_{\text{black}}(x,y)=I_{\text{closing}}(x,y)-I(\mathrm{10},y),\end{array}$$

where we use a circular structuring element of a radius $p$ in pixels, $I(\mathrm{ten},y)$ is an input epitome, $I_{\text{white}}(x,y)$ is the white peak-hat image, and $I_{\text{blackness}}(x,y)$ is the blackness top-hat paradigm. The radius $p$ must exist chosen to be larger than the stroke width of draft marks, i.e., 2 cm. The image $I_{\text{opening}}(x,y)$ is the upshot of the opening operation, which has the effect of filling small and sparse white objects. The white top-lid image is the difference between the input image and its opening image. The image $I_{\text{closing}}(x,y)$ is the result of the closing operation, which has the effect of filling pocket-sized and thin black objects. The black top-hat prototype is the difference betwixt the input image and its closing prototype. Since some ships may have both white and black draft marks, nosotros use both the white and black top-lid transforms to detect draft marks.

Throughout this newspaper, nosotros set $p=Y/\mathrm{thirty}$ , where $Y$ is the height of the input image in pixels. When nosotros choose the input image then that more than two typhoon marks are included, $Y$ is equivalent to more than than xxx cm, and $Y/30$ is equivalent to more than than 1 cm. For this reason, we can brand the radius of the circular structuring element larger than the stroke width of draft marks past putting $p=Y/30$ .

We accept applied the draft marking detection method to the epitome of Fig. 2(a). Figure 5 shows the issue of the draft mark detection. We evidence only the white elevation-hat image, because there are only white typhoon marks in Fig. ii(a). We see that draft marks are roughly segmented, but they are corrupted by racket. We therefore segment the draft marks conspicuously past local thresholding in Sec. iii.ane.2.

Fig. five

Result of the draft mark detection.

3.1.2.

Local thresholding

The assumption that the whole of an observed epitome can exist classified into simply two classes, i.east., draft marks and the send'due south hull, is not exactly true in practical applications, while local images around typhoon marks can be roughly classified into the ii classes. We thus introduce a local thresholding for draft marker segmentation. More than concretely, nosotros select bounding boxes of the segments in the resulting image of the draft marker detection, and and then binarize each of the local images enclosed by the bounding box with Otsu's binarization method. The edges of the bounding box are parallel to the coordinate axes and laissez passer through the topmost, bottommost, rightmost, and leftmost points of the segment.

Figure 6(a) shows the bounding boxes of the segments obtained in Fig. 5. Figure 6(b) shows the outcome of the local thresholding. We meet that draft marks are clearly segmented, simply also that there are still noise segments, such as under h2o draft marks.

Fig. 6

Local thresholding: (a) bounding boxes of the segments shown in Fig. five and (b) result of the local thresholding.

3.1.3.

Removing noise segments

We judge whether segments in the resulting prototype of the local thresholding are racket or not past checking if the segment satisfies the post-obit inequalities:

Eq. (3)

$$T_1\le \mathrm{west}/h\le T_2,$$

Eq. (iv)

$$T_{\mathrm{iii}}\le \mathrm{due\; south},$$

where $h$ and $w$ are the height and width of the segment in pixels, respectively, and $\mathrm{westward}/h$ is a ratio of sides. The ratio of sides of draft marking "1" is 0.2, which is the smallest of all draft marks. Nosotros thus set up $T_1=\mathrm{0.i}$ . We gear up $T_{\mathrm{two}}=\mathrm{ane}$ , considering that the height of draft marks is longer than the width in almost all cases, and the top of underwater draft marks is contracted by refraction. Meanwhile, $s$ is an area of the segment. Nosotros set $T_3=10$ , considering that also small segments are manifestly noise.

We have used the judging method to remove noise segments, equally shown in Fig. 6(b). Effigy 7 shows the outcome. We see that virtually all noise segments are well removed.

Fig. vii

Consequence of removing racket segments.

iii.2.

Draft Mark Recognition

Effigy 8 shows the flowchart of the proposed draft mark recognition. Nosotros shall call a string of draft marks as the draft mark string. In Fig. 7, "2M" consisting of the two draft marks "2" and "Grand" is the draft marker string. Recognition of the draft mark string is indispensable for draft reading. Nosotros thus distinguish the typhoon mark cord and the unmarried draft mark according to character connectivity⁹ and the predetermined size of the draft marker. More concretely, we distinguish them based on the following rules:

• 1. Ratio of the heights of the two typhoon marks in a typhoon marking string is between 0.9 and 1.1, considering that heights of all draft marks are the aforementioned.

• ii. Vertical distance between the centers of the bounding boxes of draft marks in a draft marking string is less than the quarter of their mean pinnacle, because that draft marks in the draft mark string are at the aforementioned vertical position.

• 3. Horizontal distance between the centers is less than twice equally long as their mean height, considering that the distance between draft marks in a draft mark string is close.

Fig. 8

Flowchart of the draft marker recognition.

Although plural white or black draft marker strings may be extracted by using the above rules, only the lowest draft marker string is used for draft reading. Draft marks in the lowest draft mark string are recognized past template matching using sum of squared differences. The matching templates are the images "0" to "9" and "1000," each of which height and stroke width are 10 and ii cm, respectively. Merely when the rightmost draft mark in the lowest draft mark cord is "M," typhoon marks "8," "6," "4," and "two" are recursively searched below until no typhoon marker is plant or "2" is found, according to the following rules:

• ane. Ratio of the acme of the draft mark to that of the above one is between 0.8 and 1.two, considering that heights of all draft marks are the same.

• 2. Vertical distance between the centers of the draft marker and the above 1 is 1.5 to 2.v times the height of the above draft mark, considering that draft marks are placed 10 cm autonomously from each other.

• 3. Results of the template matching satisfy the positional relation of draft marks. For instance, "viii" must be located beneath "Grand," and "half-dozen" must be located below "8."

iii.3.

Waterline Extraction

Canny edge detection^half-dozen is one of the most widely used edge detection algorithms considering of its sensitivity and high point-to-racket ratio. We utilise the Canny border detection to detect the waterline. However, the straightforward application also detects noisy edges, such as scars and projections on the hull. Noticing the property of the noisy edges, the edges on the hull remain stationary relative to draft marks. Nosotros extract the stationary edges by taking the logical conjunction of neighbour frames equally follows:

Eq. (5)

$$P(x,y,t)=\underset{1\le i}{\wedge }{I}_{\text{dilation}}[x-(a_{t-i}-a_0),y-(b_{t-i}-b_0),t-i],$$

Eq. (6)

$$F(x,y,t)=\underset{1\le i}{\wedge }{I}_{\text{dilation}}[\mathrm{ten}-(a_{t+i}-a_0),y-(b_{t+i}-b_0),t+i].$$

Here, $I(x,y,t)$ is the result of the Canny edge detection applied to the $t$ 'th frame image of the video sequence, $I_{\text{dilation}}(\mathrm{ten},y,t)$ is the dilation epitome of $I(x,y,t)$ , and $a_t$ and $b_t$ are the $x$ -coordinate and $y$ -coordinate of the reference typhoon mark to align $I_{\text{dilation}}(x,y,t)$ , respectively. Nosotros set $n$ to the half of a frame rate, considering that the waterline vanishes if $n$ is besides small. When the alignment is succeeded, $P(x,y,t)$ and $F(x,y,t)$ consist mainly of the stationary edges of the past and future frames, respectively, and the marriage of $P$ and $F$ expresses the stationary edges. Therefore, nosotros can remove the noisy edges on the hull past taking the intersection of $I$ and $P\cup F$ as follows:

Eq. (7)

$$O=I\cap (P\cup F),$$

where $O$ is the result of removing the stationary edges. Since the waterline is represented by a long edge, we further remove very small edges included in $O$ . More than concretely, we remove the edges, each of which width is smaller than 10 in pixels.

Figure 9 shows the result of Canny edge detection applied to the image of Fig. 2(a), where the thresholds of Canny edge detection are 15 and 30. Figure 10 shows the issue of waterline extraction obtained by removing noisy edges. We shall call white pixels corresponding to a waterline every bit waterline pixels. Figure 10 contains waterline pixels, but it however contains noisy pixels, such every bit edges in the bounding main surface. We thus use the to the lowest degree median of squares (LMedS) method¹⁰ robust against noisy points to estimate the waterline.

Fig. 9

Consequence of Canny edge detection.

Fig. 10

Event of waterline extraction.

3.four.

Waterline Interpretation

We set a search region in the resulting image of the waterline extraction, and so we use the LMedS method to fit a directly line to a prepare of waterline pixels in the search region. Figure 11 shows the search region. We search for white pixels downward for getting waterline pixels in the search region. The region below the lowest typhoon mark is not searched then as non to misrecognize the typhoon marker's edge as waterline pixels.

Fig. 11

Search region for waterline pixels.

We have applied the waterline estimation to the resulting image of the waterline extraction, as shown in Fig. 10. Figure 12 shows the detected waterline pixels, and Fig. 13 shows the estimated waterline superimposed on a photograph. We run into that the estimated waterline agrees with the true one.

Fig. 12

Waterline pixels.

Fig. 13

Estimated waterline.

iii.5.

Draft Calculation

We gauge the waterline for every frame and summate the typhoon from the estimated waterlines past the following steps:

• one. Compute the altitude between the center of the estimated waterline and the lesser of the draft mark, and calculate the draft reading for every frame.

• two. Apply a median filter to the typhoon readings for reducing the influence of outliers caused by failures of draft marking sectionalisation and waterline estimation.

• 3. Observe local maxima and local minima within two standard deviations from the mean of all typhoon readings.

• four. Calculate the mean of the mean local minimum and maximum.

The steps 3 and 4 are employed to emulate the professional surveyors' typhoon reading process^eleven for getting the understanding of a shipper and a receiver, although the resulting draft may be virtually equal to the mean of typhoon readings.

4.

Experiments

iv.1.

Quality of Draft Mark Segmentation

We have used two images of Figs. 14(a) and 15(a) to test the operation of the draft mark segmentation. Figure fourteen(a) is a broad angle paradigm, and Fig. 15(a) is a typhoon marker image of type ii, as shown in Fig. 1. Figures 14(b) and 14(c) show the results of Otsu's binarization method and the proposed method, respectively. Few draft marks are segmented by the conventional method, while all typhoon marks are well segmented by the proposed method. Figures 15(b) and xv(c) show the corresponding results for Fig. 15(a). No typhoon mark is segmented by the conventional method, while all draft marks are well segmented past the proposed method.

Fig. xiv

Wide angle: (a) original image, (b) Otsu'southward binarization method, and (c) proposed method.

Fig. 15

Other type draft mark paradigm: (a) original image, (b) Otsu'southward binarization method, and (c) proposed method.

four.2.

Waterline Extraction in the Rain

Nosotros test the operation of the proposed waterline extraction in the rain. Figure 16(a) shows a photograph taken in the heavy rain of about 10 mm/h, and Fig. sixteen(b) shows the result of Canny edge detection, where the thresholds are xv and 30. We run across that raindrops are besides detected in add-on to the waterline. Figure sixteen(c) shows the result of removing very small edges in the waterline extraction. We see that raindrops are removed, thus the proposed waterline extraction is effective even in the rain.

Fig. 16

Rainy condition: (a) observed epitome, (b) Canny edge, and (c) dissonance reduction.

4.3.

Reading in a Towing Tank

4.3.1.

Experimental set-up

We have evaluated the proposed method in terms of the accuracy of draft reading using a towing tank. Nosotros used a board with full-scale typhoon marks for imitating a hull. Figure 3(c) shows the lath and the water surface in the absence of a wave. The height of draft mark "six" is 50 pixels, and the altitude between the bottom of the draft marker and the waterline is 25 pixels. Nosotros thus discover that the truthful draft is 3.55 m since the height of a draft mark is 10 cm. We generated regular waves of amplitude about 3 cm and menses of 2 s and captured the scene with $410\times 440$ pixels resolution at 29.97 fps. The conventional method accumulates the binary departure of 2 consecutive frames during 600 frames, where nosotros fix the binarization threshold to fifteen. In the proposed method, nosotros used 1800 frames (most 60 s) for draft reading, and we gear up the thresholds of Canny edge detection to 15 and 30, and prepare the window size of a median filter for waterline detection to $3\times 3$ pixels.

4.3.2.

Experimental results

Figure 17(a) shows the draft readings of the first 450 frames estimated past the proposed method. Figure 17(b) shows the draft readings after one-dimensional median filtering of window size 9, where "square" represents local maxima and "triangle" represents local minima. Nosotros call a frame in which draft reading is maximum/minimum inside 4 frames from the frame as the local maximum/minimum frame.

Fig. 17

Draft reading of the proposed method in a towing tank: (a) typhoon readings and (b) after median filtering.

The drafts estimated by the conventional and proposed methods were 3.59 and iii.55 m, respectively. The true draft is three.55 g. The draft reading error of the proposed method is $<\mathrm{one}\mathrm{cm}$ , and it is smaller than that of the conventional method, because the estimated waterline of the conventional method tends to exist the highest during the observation time. The uncomplicated mean of draft readings is likewise 3.55 m, which is equal to the effect of the proposed method. However, nosotros can get the understanding of a shipper and a receiver past emulating surveyors' reading process. The full processing time of the proposed method was near 670 s on a Core i5 clocked at 3.20 GHz. The computation of the draft marker partition and recognition was ascendant, and it took about 540 s.

4.iv.

Reading in a Real-World Scene

We have applied the proposed method to two video sequences captured in a real-earth scene, every bit shown in Figs. 2(a) and 18. Both sequences are in $640\times 360$ pixels resolution at 29.97 fps. Nosotros captured the sequences from a wharf using a handheld camera for adding camera milkshake effect, because surveyors often read drafts on a pitching and rolling boat. We have calculated the typhoon for 1800 frames (about threescore s). Nosotros set the binarization threshold to 50 in the conventional method. The other parameters are the aforementioned as those of the experiment in Sec. 4.iii.

Fig. 18

Existent-world scene.

Figures 19(a) and 20(a) prove the draft readings of the first 450 frames estimated by the proposed method. Figures 19(b) and 20(b) prove the draft readings afterward median filtering. We encounter from Figs. 19 and xx that, although some outliers are caused by the failure of draft mark segmentation and waterline interpretation, the outliers are removed past median filtering.

Fig. nineteen

Proposed method applied to Fig. 2(a): (a) draft readings and (b) later median filtering.

Fig. 20

Proposed method applied to Fig. 18: (a) draft readings and (b) after median filtering.

The drafts of Fig. 2(a) estimated by the conventional and proposed methods were 1.90 and one.89 m, respectively. The true waterline in the video sequence was moving around the top of "8," i.e., i.ix m. Meanwhile, the drafts of Fig. 18 estimated by the conventional and proposed methods were 5.77 and five.60 m, respectively. Although the waterline depicted in Fig. 18 is around 5.iv chiliad, it represents merely i scene of the moving waterline. The waterline in the video sequence was moving around the top of "5," i.eastward., 5.six thousand. Nosotros thus see that the true draft is about 5.six m, and the proposed method is likewise effective in the real-world scene. The total processing time of the proposed method applied to Fig. ii(a) was nigh 1970 s. The draft marker sectionalisation and recognition took about 1390 due south. Meanwhile, the full processing time of the proposed method applied to Fig. 18 was about 1310 s. The draft mark sectionalisation and recognition took about 1010 s. The processing fourth dimension for Fig. 2(a) is longer than that for Fig. 18, considering Fig. 2(a) includes larger typhoon marks in pixels and more noisy edges than Fig. 18.

5.

Determination

In this paper, nosotros have presented an image-based typhoon reading method for improving the accuracy of epitome-based typhoon reading. To segment draft marks, nosotros have detected draft marks with morphological operations and binarized the local images around draft marks. Moreover, the accuracy of waterline estimation has been improved by using Canny edge detection. In addition, nosotros can get the agreement of a shipper and a receiver past emulating surveyors' reading procedure. We have tested the accuracy of the typhoon reading using a towing tank and accept shown that the draft reading error of the proposed method was $<1\mathrm{cm}$ . The proposed method was as well satisfactory for reading in the real-world scene. Further research volition focus on increasing robustness confronting skewed draft marks.

References

ane. 

M. Tsujimoto and H. Sawada, "Typhoon or like measuring device of hull," J. P. Patent, 2007–333530 (2011).

5. 

X. Ran et al., " Draft line detection based on epitome processing for ship typhoon survey ," in Proc. 2011 2d Int. Congress Computer Applications Computer Science, 39 –44 (2012). Google Scholar

nine. 

Grand. Matsuo, K. Ueda and M. Umeda, " Extraction of graphic symbol string region on signboard from scene image using adaptive threshold methods ," IEICE Trans. Inform. Syst., J80-D-two (6), 1617 –1626 (1997). Google Scholar

11. 

Westward. J. Dibble and P. Mitchell, Draught Surveys, North of England P&I Association, Newcastle upon Tyne (2009). Google Scholar

Biography

Takahiro Tsujii received his Be and ME degrees from Osaka University, Osaka, Japan, in 2014 and 2016, respectively. His research interests include paradigm processing and machine vision.

Hiromi Yoshida received his Exist degree from Kobe University, Hyogo, Japan, in 2007, M. maritime sciences degree in 2009 and DEng degree in 2012. Currently, he is an assistant professor in Osaka University, Osaka, Nihon. He is involved in research on design recognition and prototype processing.

Youji Iiguni received his BE and ME degrees in applied mathematics and physics from Kyoto University, Nippon, in 1982 and 1984, respectively, and his DE degree from Kyoto University, Japan, in 1989. He was an assistant professor at Kyoto University from 1984 to 1995, and an associate professor at Osaka University. Since 2003, he has been a professor at Osaka University. His enquiry interest includes systems assay.

hottingerhishmithad.blogspot.com

Source: https://www.spiedigitallibrary.org/journals/optical-engineering/volume-55/issue-10/104104/Automatic-draft-reading-based-on-image-processing/10.1117/1.OE.55.10.104104.full

Share this post