Object Extraction Based On Image Processing Techniques And Its Applications In Crack Detection

Target detection technology is a priority issue in digital image processing research areas. It has achieved automatically detection of special targets in static scenes and intelligent video surveillance based on computer vision. Generally, static image detection inspects the target with specific characteristics by extracting features of the targets first. Automatic target detection technology detects moving target sequence by analysis of the video sequence from a specific camera. Nowadays, target detection technology is widely applied in the field of license plate recognition face recognition, community security, target recognition in the infrared images, as well as the field of traffic incident detection, military affairs and national defense. These applications have greatly improved convenience in our daily life. Currently, research on target detection technology has made many achievements. However, there are still many problems to be solved.Based on existing research results, target detection techniques were studied in this dissertation. Then, the application of material crack detection is studied based on the target detection algorithm proposed in this dissertation. Compared to the traditional physical detection method, such as ultrasonic, eddy current and so on, crack detection technology based on image processing has the advantages of economic, convenient, real-time, wide application range and also getting rid of the artificial conditions. Crack detection based on image processing can solve the problem of determining the life circle of the material, estimating the lifetime of the material, and avoiding catastrophic accidents in the practical application caused by failure of material structure. Meanwhile, it can guide the optimization of material formulation and processing technology. Studying of the growth pattern of crack can assist in designing better formula or process to prolong the lifetime of the material. Therefore, the crack detection technology based on image target detection does not only have a profound theoretical significance, but also has great practical value.The main work and innovations of the research of target detection methods and crack detection method as follows:1.Image binarization is the basis of image target detection, and it determines the final detection result. This dissertation analyzes the characteristics of the traditional binary algorithm first. Then, it combines the object of study in this article with the shortage of traditional binarization method, and proposes an improved binarization method. The improved binarization algorithm achieves the goal of suppressing noise through a local threshold λ and double means filtering method. This dissertation selects the logo images binarization processing as an example, and compares it with the commonly used algorithms. The results demonstrate the effectiveness of the proposed algorithm. Through selecting the entropy of the image and evaluating the effectiveness of binarization method, the improved binarization method has smaller entropy and less noise, which also indicates the good performance of the proposed method.2.This dissertation presents a new background extraction algorithm. Background extraction is a key step in moving object detection, which directly impacts the merits of moving target detection effect. Firstly, the traditional mean background algorithm is described, and then we propose an improved background extraction method based on existing algorithms. The idea of the method is a ccording to two consecutive frames in one time and a given particular threshold value, if the difference between the two frames is greater than the threshold, then the corresponding pixel is detected as a foreground image; on the contrary, if it is less than the threshold, then it is noted as the background. In addition, this dissertation also presents an improved k-means algorithm to extract the background image. Finally, the algorithm is programmed by simulation experiments. The effect of this algorithm is verified, and the results show that the proposed algorithm can accurately detect moving objects in the image, and the object is apparent in the binarization image.3.An improved moving object detection algorithm is proposed. This dissertation discusses the principle of three temporal difference algorithms(TTD) and the mean algorithm(MA), and then it combines these two algorithms to achieve an improved moving object detection algorithm to detect moving targets. The instances used in the experiments are from the highway surveillance video, simulation results show that the improved moving object detection algorithm is stable, reliable, and accurate, which implies it has a high efficiency.4.An improved crack detection algorithm is proposed to study the structural materials of cement. The method comprises three steps: First, it converts the image to grayscale and uses sobel image edge detection algorithm and sobel image filter to detect the cracks. Second, an appropriate threshold is set in the binary image to classify the background and foreground pixels and the area filter is used to remove the so tiny area. Third, sobel filter is employed to eliminate the residual noise, and Otsu method is described to detect the major cracks. The method is simple, easy to implement and has a high computational efficiency. Simulation results show that, the new algorithm could detect cracks accurately and apparently.5.A crack detection algorithm is proposed to study the growth of cracks in cement structure, which could be modeled by continuous observation in the video afterwards. The idea of this algorithm is: First, we use the visual background extractor(VIBE) is used to detect moving cracks in video. Second, an appropriate threshold is set in the binary image to classify the background and foreground pixels, and the area filter is used to remove the so tiny area. The experimental results show that the proposed algorithm performs significantly better than other algorithms.