Research And Application Of High-speed Photography Object Extraction And Tracking System

Lack of advanced experimental instrument for studying the object movement in farm machine operations is a serious problem at present. Hence a high-speed photography system is established on soil bin to study the rules of moving objects. It is first time to research the moving track of the maize straw shattered on the planter by using a high-speed photography system in order to resolve the anti-blocking problem of no-tillage planter. The dissertation have made a pioneer study on moving picture acquirement, straw object extraction and moving object tracing in the heavy clutter situation of imbalance lighting, camera and straw moving simultaneously with several similar moving objects on mass background, and obtained following achievements:1. A capturing system for moving pictures is developed, which could capture and save pictures in real time. The special emphasis is on solving the lost frame problem during images transmitting caused by low transmitting speed of peripheral equipments.2. A new background training method is put out which using weight average approach of multiple images to establish the background model. Background subtraction method based on this model can great remove the disturbance of edge of sawtooth disk during the process of target detection.3. The GVF Snake model was introduced to extract the moving targets from the image sequences. A creative modification was added in order to overcoming the shortage of the model by using the watershed segmentation algorithm. The experiment shows that the modified method can extract the outline of straw object in the pictures efficiently.4. To the problem of time exhaustion in using the GVF Snake model, a "open-window" method is put forward which is more pertinence and time saving than the traditional method by restricting the calculating area in a small window. A comparison result is given to prove the high efficiency of the proposed method.5. Two criteria are put out in straw classification process. The first one is based on the proportion of length and width of object's bound box, and the second one is based on the circularity of object. The given method is simple and high efficiency, because most classification problems could be done through the first criterion, and complex solution procedure in getting the object's area could be saved.6. A tracking method by combining Condensation and Snake model is first put forward in the dissertation. A new rectangle box with a center gained by Snake model in stead of the minimum bound box was pointed out, which not only benefits the calculation of factor sampling but also fits the real situation by which the real target is tracing; Further a measure based on the edge features of black-white image corresponded to the original image was given out, and the confidential function was given out as well; A method is advanced to avoid of filter diffusion by using residual error prediction.7. A high-speed photography software system was designed, and was used in the straw throwing experiment. The results show that the trace is similar to the quadratic curve predicted by theory.