School of Science and Technology 科技學院
Computing Programmes 電腦學系
| Programme | Bachelor of Science with Honours in Web Technologies |
| Supervisor | Dr. Jeff Tang |
| Areas | Innovative Systems |
| Year of Completion | 2014 |
The aim of the intelligent work desk system (IWDS) is to enhance the efficiency and safety of human who working on a work desk. We are developing an intelligent system with the use of real-time monitoring 3D sensor. The system is expected to allow user to find their missed things and to preventing the things stolen from the work desk. We have concluded the concepts of the project into three core parts:
In order to decrease and enhance the safety of the working process, The IWDS can logs the date, times, coordinates, movements and other related information of different objects so users can locate the lost object easily. Also, IWDS is easy to use so users can interact with the system efficiently.
The structure of The Intelligent Work Desk System (IWDS) is divided into three sections, and each section has also divided into few components. We will describe each components of each section in this chapter, and to show their purposes, roles, and functions. More details will be described in the Component Implementation and Evaluation Document.
Image Visualization is the process of displaying the images captured by the Kinect sensor projected on the screen. It is used to keep extracting the current work desk environment captured by Kinect Sensor.
As for the calculation of later procedures, such as Object Segmentation, Object Recognition, Object Tracking, we need to remove useless points from the point cloud by using the Cartesian equation. If the depth value of a point is smaller than the z value, which is calculated by substituting the x, y coordinates of the point into the equation of the plane, we keep this point otherwise we discard it.
After the Table Points Removal procession, the result image gotten need to be converted into a grayscale image. The Blob Detection consists of three steps, Edge Detection, Contour Detection, and Drawing Bounding Box. The Edge Detection helps finding the edges of each blob. Second, Contour Detection helps finding the coordinates of the edge pixels. Finally, the IWDS draw a bounding box around each blob to tell users the IWDS how many objects have detected. The IWDS also save each blob detected into a database for later implementation of Object Recognition.
As for implementing the user interface, we consider that how to provide a user-friendly interface for users to use. There is a main goal of the interface design, which is that providing a user interface that users feel easy to use. Therefore, we do many references to other user interfaces. Finally, we implement the user interface shown as below:
After the fine tuning, the result can be good. Users can press the “Start Capturing” to start object logging. Each individual object is separated and saved into the database with 10s temporal. The default name of the first object will be “item0” and so on. The IWDS keep detect the movement of the object or even detect the new object, then save the latest object information into the database without flickering.
Object Recognition can help users find the stuff in real-time. When users select the particular item and search it in the recognition mode, the IWDS start the algorithm and match the pixel. As the matching process is required lots of resource, the object recognition mode has a little flickering but it is acceptable. The IWDS highlight the wallet with the bounding box. It only takes one second to match the object. When users move the wallet now, the bounding box will follow the new position of the wallet too.
To evaluate whether we have satisfied the aims of the project, we have conducted a survey to investigate the opinions from users. We invited 30 peoples to test our system and fill in the survey. Level 1 means strongly disagree, level 2 means disagree, level 3 means somewhat disagree, level 4 means neither agree or disagree, level 5 means somewhat agree, level 6 means agree, and level 7 means strongly agree.
In generally, users agree the IWDS can secure the stuff from being stolen. However, we can do better in this part. As the IWDS saved each objects status into the database with 10s temporal. If the certain stuff disappears with in this 10s, the information of this stuff will not be saved into the database.
We have got a very high mark in the second question. Everyone though the IWDS can really help them to remember the location of stuff. The object logging function is satisfied undoubtedly. Especially for the long-term usage, the IWDS log each object with information automatically. Unless the users delete the database, users can remember the location of stuff in long time ago easily.
Most of the respondents agree the user interface of the IWDS is easy to use. We have got some comments about the user interface. Some of the respondents love the UI design; they said it is modern and easy to use. Also, some of them mentioned that the voice command function is accuracy. However, some of the respondents said the initiation part of IWDS is complex; we may need to simplify this part.
All of the respondents thought the user interface of the IWDS is natural.
Most of the respondents agree the IWDS can help to enhance their productivity.
We have satisfied all the project aim including:
The IWDS is capable to recognize and track the locations of things on desk in real-time, such as stationaries, books, and cups. IWDS can log the date, times, and positions of different objects so users can locate the lost object easily. This function could help users to find the lost stuff easily. Also, IWDS can recognize voice so users can simply say the keyword to search the certain item quickly. After the evaluation part, we know that most users are identity and appreciate the IWDS. They agree the IWDS can help to enhance their productivity, help them to remember the location of stuff, help to secure their stuff on desk from being stolen, and they like the user interface.
In future, this system can have lots of potential. We can learn more complicated algorithm or approach to achieve real 3D detection and recognition. Or other powerful function, For example, the system can recognize the degree of risk of the objects so the system can be used in the kindergarten or nursing home for enhance the safety, decrease the chance of accident.
Copyright Ng Wing Hoi, So Chung Hei, Tsui Chun Yi and Jeff Tang 2014
