Research on Object Recognition Based on Image Segmentation

1.Background

Image segmentation refers to the technology and process of dividing an image into several specific areas and then extracting the desired object target. Since image segmentation is an important step from image processing to image analysis, image segmentation has been highly valued since its generation. In addition, the results of image segmentation are the basis for Image Feature Extraction and Recognition. The research on image segmentation has always been the focus and focus of digital image processing technology. It mainly originated from the film industry in the 1970s s and has evolved from Image Segmentation technology to the present decades. With the continuous advancement of our technology, we have also proposed many image segmentation algorithms, which have also achieved remarkable results. In addition, image segmentation has been widely used in all areas of our lives, for example, online product inspection, production process control, document image processing, industrial automation, remote sensing, production process control and biomedical image analysis, as well as sports, agriculture, military engineering, etc. In general, on the one hand, researchers constantly improve the original segmentation method and apply it to other fields. On the other hand, more and more scholars begin to apply the research results of mathematical morphology, fuzzy theory, genetic algorithm theory, fragment theory and wavelet transform theory to image segmentation, and propose some more advanced methods. However, so far, we have not provided a more general segmentation method [2]. The methods we have currently proposed are as follows, most of them are raised after analyzing specific problems, and they are not necessarily applicable to other problems. In addition, because of this, we have not yet developed a suitable image segmentation algorithm standard. Therefore, our target recognition method is also uncertain. In our real life, there are still many problems that need to be solved.

In fact, as early as the 1960s s, some people proposed edge detection methods, and therefore produced many classic image segmentation algorithms [3]. As image segmentation plays an important role in image processing and analysis, more and more scholars have begun to focus on finding new theories and methods to improve the quality of image segmentation, to meet people's needs in real life. At present, many scholars have tried to apply the genetic algorithm theory, mathematical morphology, wavelet transform theory, fragment theory and fuzzy theory to image segmentation, an advanced image segmentation technology combining specific mathematical methods and specific image segmentation, such as multiple feature fusion and multiple segmentation methods.

2.Significance of Research

In the decades of emergence and development of image segmentation technology, more and more scholars have begun to apply the research results of mathematical morphology, fuzzy theory, genetic algorithm theory, fragment theory and wavelet transform theory to image segmentation, A more advanced image segmentation technology combining specific mathematical methods and specific image segmentation is produced. However, there is still no common segmentation method at this stage. There are many problems that need to be solved. Because there is no common segmentation algorithm, when we use a segmentation algorithm, we also need to identify the target, so what kind of recognition method can be used to better and more easily identify the target? This is what I want to study. After selecting a segmentation method, compare several different object recognition methods and select a relatively good object recognition solution.

The Automatic Color *** image segmentation algorithm integrates the edge gradient texture plug-in for extraction and seed area growth in the YUV color space. The edge of Y, U, and V is detected by an edge detector of the same nature. The three parts are combined to obtain the edge. The center of gravity of the adjacent edge area is used as the initial seed.

The Hausdorff distance is a measure of the similarity between two sets of points. It is a definition of the distance between two sets of points. Its advantage is that the calculation is relatively easy, but the matching degree is not good. If there is too much noise in the target image, the matching accuracy will be affected.

Principal Component Analysis (PCA) is a statistical technology that is often used in human facial recognition, image compression, signal denoising, and other fields, it is a common technology used to extract patterns from high-dimensional data. We can use Principal Component Analysis (PCA) to reduce the dimensionality of data and multiple variables so that we can calculate and analyze images.

3.Contents of Research

1. design the Image Segmentation Algorithm Based on the regional growth technology;

2. Evaluate the performance indicators of image segmentation algorithms;

3. Learn and compare several object identification methods, such as Hausdorff distance, imagecontext, and principal component analysis;

4. Use MATLAB for experimental verification, draw the corresponding conclusions based on the experimental simulation results, and the shortcomings and improvements of this method.

4.Research methods

We can use MATLAB software to simulate the image segmentation algorithm, and we can also use the image library test (Berkeley data set) [17]. We will use video clips during the experiment. We can use the existing one or ourselves to record video clips. Finally, there will be a series of performance indicators for different target recognition methods, including the target recognition rate, recognition error, and noise impact, then, we can use these performance indicators to compare the advantages and disadvantages of Hausdorff distance, image context, principal component analysis, and Other Object Recognition Methods. Finally, we can draw our conclusion.