1.1 Problem Description
Nowadays, disadvantages people such as visually impaired people are around the world. According to World Health Organization (WHO) estimated in 2013, there are 285 million of world population who are visually impaired. Furthermore, 39 millions out of the population are blind. These people require assistance in their daily life such as locating and grabbing objects (i.e. a cup, a key, etc.) which may have been fallen or misplaced. This is because, they have less information about the environment. Hence, developing assistive technology and handheld devices can help them to increase the independence and inconveniences. The final goal is to improve their ...view middle of the document...
Secondary, descriptor is constructed on each feature points detected in the first step. After that, the descriptor on each feature points is compared using the K-Nearest Neighbor (KNN) match and produce the good descriptors among the feature points. And, these good descriptors can be used for object matching. (details in Chapter 4)
In this time, the author will only use the still images to test whether all the extracted SURF features that are scale invariant, rotation invariant and illumination invariant. Hence, this project needs a wearable web camera and a computer. The images captured from the web camera act as the reference images for object recognition to compare with the query images in the dataset.
1.4 Objectives and Contributions
In this project, the author uses computer vision technology to assist the visually impaired. Here, the system is mainly focused on invariant feature points of objects by employing SURF detection and description. Invariant feature points of the objects will be extracted for matching purpose. After matching the object, the system will be able to tell user about what the object is and the visually impaired eventually can notice what is the object even they have less information about the surrounding environment. Therefore, this is enabling them to achieve higher independence. They can find the intended object by using this system instead of others' help.
Additionally, this system can reduce their daily inconvenience and improve their life in order to complete their tasks efficiently. They can use this system to solve their problem of finding a specific object. Other than that, this system can also compensate for their limitation. For instance, increase their independences access to information and effective communication with their environment.
3 System Design
This proposed system focuses on how SURF is being used to detect and match the feature points on the frame images for object recognition. There are two steps involved in SURF technique. First step is to detect the feature points, second is to use the descriptor to describe the feature with points within a frame image. After extracted the features descriptors, KNN with K=2 are found among the descriptors for object matching. Details of the SURF technique and object matching are described in Chapter 4. Fig 3.1 is the flow chart of the entire system whereas Fig 3.2 shows the steps involved in Object Detection using SURF.
Figure 3.1 Flow Chart of entire system.
This proposed system is designed in such a way to help visually impaired people. In the first place, all the objects will be detected and showing the type of objects on screen such as cup, key, etc. The steps involved in object detection as shows in Fig 3.2 are discussed in Chapter 4. Meanwhile, the system will tell user the all detected objects with number. If the intended object is detected, the user will voice out the number of the object. Afterwards, hand of the user will be detected in order...