Visible Light Path Laser Pointer

Transcription

1 Visible Light Path Laser Pointer Shu Okamoto Nobuchika Sakata Shogo Nishida Osaka University, JAPAN Abstract Under remote collaboration, a teleoperated laser pointer is applied as an instruction tool to point at real-world objects. However, this method has difficulty in identifying the object being pointed at when the laser spot is occluded behind objects. This paper proposes a method that visualizes the light path of a laser by jetting a mist along the light axis to mitigate this difficulty. A worker should be able to estimate the position of the object being pointed at by using the visible light path laser pointer. Experiments were conducted to evaluate the proposed method in a face-to-face situation. The results showed that the estimation accuracy for the position of a laser spot occluded behind an object was improved when the light path of the laser was visualized. 1. Introduction Work conducted by a local worker under the instructions of a remote instructor is called remote collaboration. Using a telecommunication terminal, the remote instructor and the local worker transmit and receive sounds and videos to accomplish their work since they can not share voices and views directly. On the other hand, a worker and an instructor sometimes communicate regarding objects and places in real work spaces in local collaborative works [1-3]. To conduct such communication smoothly, a support system sends the remote instructor's instructions including the place of the work to the local worker. In some researches, a teleoperated laser pointer is adopted as a means of sending such instructions [4-7]. This method enables a remote collaboration system to be realized with a compact device. However, although the instructor can watch the laser spot, this method can be difficult for workers observing the laser spot when it is occluded behind the object being pointed at or other obstacles. Research has shown that laser-pointing enables smooth remote collaboration for simple tasks, such as selecting and specifying real-world targets. Therefore, the remote collaboration cannot be conducted smoothly when the laser spot providing important information is hidden from the local worker's sight. This research aims to support remote collaboration, including instructions identifying real objects with a compact device. This paper proposes a new method for pointing to objects for remote collaboration support devices and provides an assessment of the method. The method uses 3D directional expression by visualizing the light path of a laser pointer. With this method, a worker can be expected to estimate the position of an occluded laser spot from the visualized light path.. 2. Related work The teleoperated laser pointer is adopted in some research as a pointing tool for remote collaboration. Telepointer [4] is a wearable device mounted on the chest. Telepointer is equipped with a fixed camera and a teleoperated laser pointer which enables remote collaboration targeting objects in real-world. A Wearable Active Camera with a Laser pointer (WACL) [5] is a hands, head, and eye-free wearable device mounted on a worker's shoulder. WACL is comprised of a camera and a laser pointer on a biaxial camera platform. The instructor can look around the environment through the camera no matter what the worker looks at. GestureCam [6] is not a wearable device but a device placed in a work space. GestureCam has a camera mounting a laser pointer on it. An instructor can pan and tilt the laser pointer on the camera to point at real-world objects. CTerm [7] is also a device placed in a work space. CTerm is equipped with a camera and a laser pointer which are remotely controlled. GestureMan [9] is a system equipped with not only a teleoperated laser pointer but also a robot head and a robot arm. The robot head and the robot arm trace the motion of the remote instructor. Therefore, GestureMan is assumed to be able to convey the target of instruction even if the laser spot is occluded. However, the robot arm and the robot head prevent the system from being reduced in size because the robot arm and the robot head require a certain size to express gestures and orientation. As described above, remote collaboration systems with rich gestural expressions are large-scale, although they are assumed to be able to convey instructions regarding real-world objects without relying upon a laser spot. 8

2 There has been no research aimed at realizing rich 3D directional expression with a compact device. 3. Visualization of the light path of a laser pointer The light path of a laser is visualized as a bright line when the laser is scattered by particles. Scattering a high-power laser with dust is a possible method for visualizing the light path of a laser. However, a high-power laser is not suitable for remote collaboration since it can harm a worker. Although scattering a low-power laser by filling up the workspace with mist is harmless, time is required to fill up the workspace with an artificial mist even if a large-size device is applied. Therefore, this paper suggests a harmless method which requires neither a large-scale device, nor a long wait time. The proposed method scatters a laser with an artificially generated mist jetted in the direction of the laser with a fan. This allows a reduction in the necessary amount of mist and reduces the required size of the device compared with filling up the workspace with mist. Moreover, this method has low danger since the method can visualize the light path of the laser even at low power. A visible light path laser pointer was developed experimentally by applying the suggested method and is shown in Figure 1. The device consists of a mist generator and a laser pointer. A mist generated by the mist generator is transferred to a hose by a fan, and then jets from the hose end. A laser pointer is mounted on the hose end along the direction of jetted mists, so the laser goes through the mist. The mist generator is equipped with two ultrasound atomizer HM-2412 (Honda Electronics Co., Ltd.). One HM-2412 can atomize approximately 250 ml/h of water. The size of a mist Fan Oscillation circuit 15cm 15cm Tank Oscillator 22cm Laser pointer Figure 1. Laser pointer with jetting mist along laser light path. particle is approximately 3 m. The size of one circuit board is mm 2, and the size of one oscillator for the mist generator is mm 3. The laser pointer is a commercially available product with an on/off push switch. The laser is green with a wavelength of 532 nm, maximum output of less than 1 mw, and a beam width of 5 mm. The length from the switch to the end of the laser pointer is 46 mm, and the length from the switch to the end of the hose is 55 mm. The light path of the visible light path laser pointer with atomized water could be observed under room lights of 460 lx illuminance. The length of the visible part of the path was approximately 1 m. The length was slightly short, probably caused by the ready vaporization of the water mist. To prevent vaporization, fog liquid (FLG1, Stage Evolution), which is a liquid for smoke machines, was atomized. When fog liquid diluted to 20% with water was atomized, the length of the visible part was extended to approximately 3 m since the fog liquid vaporized less easily than water. However, the amount of mist decreased because FLG1 was not atomized by HM As a result, the brightness of the light path decreased compared with the case of water. Though the light path was bright enough to be easily observed from the front of the laser pointer, the light path was less easy to see from the side. In addition, when the direction of the laser pointer was changed, the visible light path took a short time until the mist followed the movement. The delay was approximately s when water was used. Moreover, in windy circumstances such as outdoors, the mist was blown away, shortening the visible light path. The function of the visible light path is similar to a pointing stick. However, a pointing stick cannot point at targets with shorter distances than the length of the pointing stick because the stick may make contact with the target. Moreover, a pointing stick has the danger of contact with a worker or object since a worker may have to work near a device mounted with a pointing stick. Therefore, a pointing stick is not suitable as a device supporting remote collaboration. 4. Experiment 1 When an instructor facing a worker tries to point to an object between the worker and the instructor with a laser pointer, the laser spot is likely to be out of worker's line of vision. If the worker can estimate the position of the laser spot from the visualized light path even when the worker cannot see the laser spot the remote collaboration targeting real-world objects can be conducted smoothly. Therefore, an experiment was conducted to examine the effectiveness of the visible light path laser pointer in a face-to-face situation. The objective of the experiment was to confirm whether the visualization of a light path of 9

3 a laser improves the accuracy when estimating the position of a laser spot out of the observer's line of vision Experimental settings In the experimental space, a board was stood palewise on the floor as shown in Figure 2. White paper was pasted on both faces of the board so that corresponding projection areas of 150 cm widths and 100 cm heights could be acquired. The thickness of the board was 5.4 cm, the length from the top of the board to the top of each projection area was 2.3 cm, and the length from the floor to the bottom of each projection area was 19.0 cm. An "instructor" stood in front of the board at a distance of 300 cm. On the other side of the board, an "observer" stood at a distance of 100 cm. The observer can see the hand of the observer over the board. Two projectors were placed at certain points from which the projectors could project images on each projection area without interruptions from the instructor or observer. The illuminance around the instructor's hand was 460 lx. The following task was conducted in this environment: 1. A target position is selected randomly in the projection area; a cross mark is then projected on the target position on the instructor's side. 2. After a starting sound, the instructor points at the cross mark. 3. The observer estimates the target position, and then points using a laser pointer in his hand. 4. After a lapse of 3 seconds from the starting sound, the task ends with the ending sound. Ten training tasks were conducted, followed by ten actual tasks. After each training task, a cross mark was projected on the target position on the observer's side so that he could check the correct position, as shown in Figure 3. The tasks were conducted in four instruction conditions. For each condition, the instructor used a different tool: a visible light path laser pointer atomizing water (water + LP), a visible light path laser pointer atomizing fog liquid (FL + LP), a conventional laser pointer (LP), and the instructor's finger (finger). In prior research, the authors planned to mount a laser pointer on a teleoperated compact device. Therefore, for the conditions using a laser pointer, the instructor showed only his hand and a laser pointer to the observer, hiding his body behind a partition as shown in Figure 2 in order to prevent the observer from estimating the target position from the direction of the instructor's body and head. The observer could see neither the laser spot nor the cross mark since the board was opaque. In each instruction condition except for "finger," the instructor used identical laser pointer and identical hose. In addition, even in the LP condition the laser pointer was connected to the mist generator with the hose. People point at targets with their fingers in different styles. Therefore, the subjective position of the instructor's pointing is different from the observer's estimation of the objective position from the instructor's finger and arm. Therefore, the instruction with the finger condition had a disadvantage in terms of accuracy. In this experiment, the finger condition was added to compare to the other conditions. This experiment was conducted with ten subjects (gender: one female/nine male; age: 22 to 24; height: 162 to 178 cm) as observers. The subjects had no prior experience with the task for this experiment. One instructor (gender: male; age: 22; height: 180 cm) who had prior practice with the task was paired with the ten observers. Each pair conducted the tasks with four instruction conditions. The order of instruction conditions was different for each pair to prevent the order effect. In this experiment, each randomly selected target point was recorded. In addition, the view of the projection area on the observer's side was recorded with a video camera. The position the observer was pointing at the moment each task ended was measured from this video Results Figure 4 shows a box-and-whiskers plot of the error distance: the distance between the target position selected randomly and the position the observer estimated in the actual tasks for each instruction condition. Using the Wilcoxon signed rank test with a significance level of Observer Instructor 1.5m 1m 1m 3m Error distance Figure 2. Configuration of experiment. Figure 3. The correct answer is projected after a training task. 10

4 0.05, a significant difference was found between the FL + LP and LP, FL + LP and finger, water + LP and LP, and water + LP and finger conditions (p < for each pair). Figure 4. Error distance between estimated position and correct position Discussion The error distances for the water + LP and FL + LP conditions were significantly shorter compared to the LP condition. Therefore, the visualization of the light path makes it easy to estimate the position of the occluded laser spot in a face-to-face situation; in other words, the problem caused by the occlusion of a laser spot can be remedied in a face-to-face situation with the visualization of the light path. The error distances for the FL + LP and water + LP conditions were also significantly shorter compared to the finger condition, although both the visible light path laser pointer and a human finger can express 3D direction without a laser spot. This suggests that a visible light path laser pointer may provide better pointing instruction than a humanoid robot. However, the instructor can confirm the exact position he is pointing at by using the laser spot and then making adjustments when using a laser pointer. On the other hand, when the instructor points with his finger, he has to estimate the position by eye. As a result, corrections in the pointing are difficult for the instructor. Therefore, the results for the finger condition are more susceptible to error than the other instruction conditions. 5. Experiment 2 As shown in Section 4, visualizing the light path of the laser pointer improved the accuracy of estimating the position of an occluded laser spot. An experiment was then conducted with a task similar to a selecting operation in actual work. In this task, the observer distinguished certain objects pointed at by the instructor from many objects placed on a table in a face-to-face situation. The laser spot was occluded behind objects similar to the settings of the experiment in Section 4. The objective of the experiment was to confirm whether the visualization of the light path of the laser was effective in such settings. The observer was expected to be able to distinguish the object being pointed at from the others when the estimation error was short compared to the distance between the objects. Based upon the results of the experiment in Section 4, the instructor was hypothesized to be able to instruct with a visible light path laser pointer more accurately than with a conventional laser pointer. In Section 4, the distances from the instructor to the targets were approximately constant. In contrast, in this section the distances from the instructor to the targets varied. Moreover, in some cases the targets overlapped from the instructor's viewpoint Experimental Settings As shown in Figure 5, the instructor stood in front of a table at a distance of 50 cm, while the observer stood on the other side of the table at a distance of 70 cm. The table dimensions were 180 cm width, 80 cm depth, and 70 cm height. Moreover, fifty blocks with dimensions of 4.0 cm width, 3.2 cm depth, and 6.3 cm height were placed on the table. On the top and on the observer's side of the blocks, sequential numbers were printed. On the instructor's side of the blocks, black papers were pasted so that the observer could not estimate the block being pointed at from the reflected laser light. The illuminance around the hand of the instructor was 980 lx. In such an environment, the following task was conducted: 1. The observer stands with his eyes closed. 2. The instructor points at a target block selected randomly. 3. After hearing a voice announcing the start of the task, the observer opens his eyes and estimates which block is being pointed at. 4. The observer reads out the number printed on the estimated block. In fact, the observer can estimate the object being pointed at from the laser spot moving on the table in actual work. However, a worker does not always watch the laser spot. The worker can sometimes lose sight of the laser spot when he starts searching for a laser spot after his manipulation with watching around his hands, or when the laser spot moves extremely quickly. For this reason, the observer waits with his eyes closed in this Observer 80cm 50cm Instructor 70cm Figure 5. Scene of experiment. 11

5 experiment so as not to watch the laser spot moving on the table. First, the subjects conducted five training tasks to get used to the task for each instruction condition, and then conducted twenty actual tasks. This experiment was designed to clarify how the position of a target block and the distance from a target block to other blocks affected the error rate. Fifty blocks were put on the table; however, only ten blocks were actually targeted, and the other forty blocks were treated as dummy blocks so that many samples were obtained to analyze for each target block. In order to prevent the observer from perceiving the lopsided targeting, whether the answer was right or not was not disclosed. The instruction conditions (water + LP, FL + LP, LP, and finger) were the same as in Section 4. For the water + LP, FL + LP, and LP conditions, the instructor showed only the hand and a laser pointer to the observer, hiding his body behind a partition as shown in Figure 5 when pointing to a target block. For these instruction conditions, the instructor used identical laser pointers and identical hoses. In addition, even in the LP condition the laser pointer was connected to the mist generator with the hose. The laser pointer was used at a height approximately 120 cm above the floor. This experiment was conducted with twelve subjects (gender: two female/ten male; age: 21 to 24; height: 160 to 178 cm) as observers. The subjects had no prior experience with the task of this experiment. One instructor (gender: male; age: 22; height: 180 cm) with prior practice for this task was paired with the twelve observers. Each pair conducted the tasks in four different instruction conditions. The order of instruction conditions was different for each pair to prevent the order effect. In this experiment, the time from the voice announcing the start of a task to the observer's voice reading out the answer was recorded as the task completion time. The number printed on the target block selected randomly was recorded as the correct answer, and the number the observer read out were recorded as the observer's answer Results Figure 6 shows a box-and-whiskers plot of the task completion time for the actual tasks in each instruction condition. Figure 7 shows a box-and-whiskers plot of the error rate for each observer in the actual tasks for each instruction condition. Figure 8 shows a bar graph of the error rate for each target (A-J) in actual tasks for each instruction condition. A Wilcoxon signed rank test with significance level of 0.05 was carried out to evaluate the completion time and error rate for each observer. The results showed that the task completion time for the finger condition was significantly short compared to the other instruction conditions (p < for each instruction condition). Furthermore, the task completion time for the water + LP condition was significantly shorter than the FL + LP condition (p = 0.022). The error rate for the water + LP condition was significantly lower than the finger (p = 0.004) and LP (p = 0.002) conditions. Moreover, the error rate for the FL + LP condition was significantly lower than the finger (p = 0.005) and LP (p = 0.003) conditions Discussion The task completion time for the finger condition was shorter than in the other instruction conditions. Moreover, Observer J Figure 6. Task completion time. Instructor I Figure 7. Error rate. H G F E Figure 8. Error rate for each target. C D B A 12

6 the task completion time for the water + LP condition was shorter than for the FL + LP condition. These results may have been influenced by the visibility of the laser light path. The narrow light path is more difficult to see than pointing with a finger. Furthermore, the light path in the FL + LP condition is not as easy to see as that in the water + LP condition. The observer should take more time to estimate the object being pointed at when the light path is difficult to see. Therefore, improving the visibility of the light path may shorten the time to distinguish the target. One of the methods to improve the visibility of the light path is modifying the method to generate or jet the mist. Another is to modify the width of the laser. The width of the laser can be widened or narrowed by vibrating the light path with a simple vibration motor as well as optical means. In Figure 7, the error rates in the water + LP and FL + LP conditions are lower than the LP condition. This result may have been brought about by the 3D direction expression of the visible light path laser pointer. In addition, the error rates for the water + LP and FL + LP conditions were lower than for the finger condition. This result shows that the visible light path laser pointer is better at expressing a 3D direction than pointing with finger in terms of accuracy. However, it must be noted that the results can be affected by the personal pointing skill of the instructor, especially in the finger condition. Although there was no significant difference in error rate between the water + LP and FL + LP conditions, FL + LP had a slightly lower error rate. This suggests the reasonable assumption that a longer visible light path enables better comprehension of the instruction. However, one subject who had bad eyesight had a very high error rate with the FL + LP condition because the light path was darker than the water + LP condition and was difficult to watch. The error rate can be expected to increase in a brighter lighting environment as well as the case that the subject has bad eyesight. In Figure 8, the error rates for blocks close to the instructor and distant from other blocks (such as D, G, and J) for the water + LP and FL + LP conditions were lower than for the LP condition. Moreover, the error rates for blocks away from the instructor (such as B and C) for the FL + LP condition were lower than for the water + LP condition. These results showed that the observer can easily distinguish the block being pointed at from the other blocks if the distances from the block being pointed at to the other blocks are long compared with the length of the visible light path. 6. Conclusion This paper suggests a method to visualize the light path of a laser pointer in order to remedy problems caused when the laser spot is occluded. Experiment results showed that the proposed method improves the accuracy when estimating the position of an occluded laser spot. In addition, the proposed method was effective in terms of accuracy when distinguishing an object being pointed at from other objects, even when the laser spot is occluded. For future work, a device, named the Visible Light path Laser Projector (VLLP), for remote collaboration is being constructed. VLLP is equipped with a laser projector and a mist generator. With the laser projector, VLLP can instruct with not only a laser spot but also simple line drawings. Acknowledgements This research was partially supported by the Ministry of Education, Science, Sports and Culture, Grant-in-Aid for Young Scientists (B), , 2009 and the special grant from SECOM Science and Technology Foundation. References [1] H. Kuzuoka, Spatial workspace collaboration: a Sharedview video support system for remote collaboration capability, Proc. CHI 1992, , [2] S.R. Fussell, L.D. Setlock, and R.E. Kraut, Effects of head-mounted and scene-oriented video systems on remote collaboration on physical tasks, Proc. CHI 2003, , [3] R.E. Kraut, M.D. Miler, and J. Siegal, Collaboration in performance of physical tasks: Effects on outcomes and communication, Proc. CSCW 96, 57-66, [4] S. Mann, Telepointer: Hands-free completely self-contained wearable visual augmented reality without headwear and without any infrastructural reliance, Proc. ISWC00, , [5] WACL: Supporting telecommunications using wearable active camera with laser pointer, Proc. ISWC2003, NY, USA, 53-56, 2003 [6] H. Kuzuoka, T. Kousuge, and M. Tanaka, GestureCam: a video communication system for sympathetic remote collaboration, Proc. CSCW 94, 35-43, 1994 [7] M. Mikawa and M. Matsumoto, Smooth and easy telecommunication using CTerm, Proc. IEEE SMC '99, , 1999 [8] N. Sakata, T. Kurata, and H. Kuzuoka, Visual assist with a laser pointer and wearable display for remote collaboration, Proc. 2nd international conference on Collaboration Technologies (CollabTech 2006) in Tsukuba, Japan, 66-71, 2006 [9] H. Kuzuoka, Y. Furusawa, N. Kobayashi, and K. Yamazaki, Effect on Displaying a Remote Operator's Face on a Media Robot, Proc. ICCAS 2007, ,