Research On Objects Image High-speed Detection Algorithm Under Natural Environments

Digital image processing technology,with its flexibility, high processing accuracy,and simple system configuration, has been widely used in many fields such as scientific research, industrial and agricultural production, biomedical, aerospace, military, transportation, robot vision, etc. It has become an attractive and promising new discipline. In digital image processing, objects detection and recognition is the basis for image analysis and image understanding, and also it is the core of digital image processing technology.In the paper, the content of study is to detect objects under natural environment. By the research and development of four projects, a set of objects image detection strategy and algorithm which can be used universally in natural environment with high accuracy and high-speed was summarized. The research laies the foundation for engineering application of objects image detection technology. The main contents are as follows.(1) Detection of ripe peaches in orchards. Considering the color characteristic of a mature peach,color difference enhancement and matching expansion were applied to recognize the entire region of the peach. By the intersection of the perpendicular bisector of the line on the contour, the center point and radius of the fitting circle for the peach were obtained. Finally, the task of recognition and localization of the ripe peache was accomplished quickly and accurately.(2) Detection of navigation lines and field ends for farmland operating machinery. For three different farmland operating environments, wavelet transform, monotony analysis, front and rear frames interrelated and Passing a Known Point Hough Transform were used to detect navigation lines and field ends accurately, steadily and quickly.(3) Detection of on/off states of platens in transformer substations. Different color difference models, traditional projection, interval projection and waveform data analysis to obtain peaks and troughs information were used to recognize the on/off states of platens with various color and shapes on relay screens in transformer substations(4) Detection of markers in traffic accident scene. For the ruler and labels in accident scene,line scan, interval processing, color difference enhancement, geometric parameters measurements and calibration point method were used to achieve high-speed and accurate detection of the ruler and labels,and automatically measurement of the distance as well as area.The main conclusions are:(1) In the RGB color space,it is effective by establishing the appropriate color difference models based on the color characteristics of the objects to be detected, to achieve objects enhancement,background suppression and elimination of the influence of light.(2) By using interval projection proposed in the paper and traditional projection, variation trend of objects in the image was enlarged and the influence of background and light was removed. Wavelet transform and moving average were used to smooth waveform data obtained by projection, then peaks and troughs information was calculated by waveform data analysis method. Eventually, according to the information, objects detection was completed.(3) When objects detection carried out in the image with high resolution and large amounts of data,interval processing based on line scan or converting two-dimensional image data into one-dimensional waveform data by projection can be used to increase the processing speed greatly.