| With the increasing aging population and the continuous improvement of the living standards,in high strength, high pressure and high risk human labor has urgent need to be replaced bymachines. Thus the human can escape from the excessive consumption of life. Compared with thetraditional machine, modern robot can adapt to more complex and changeable environment,instead of people to accomplish more complex tasks. In the process of robot research anddevelopment, intelligent is the theme of the development. Some advanced robots imitating humanbehavior are edging towards our vision. But as an agricultural country, the development of theagricultural robots is not very fast, intelligent level needs to improve.To realize the autonomous mobile robot depends on the advanced navigation system, and inthe navigation system vision navigation is the most direct the most effective way. So robot visualnavigation is of great significance. Because the characteristics of agricultural robot workingenvironment are wide, difference obvious and easily affected by surroundings, study ofagricultural robot vision navigation has profound significance. At present, due to the robot visualsystem needs a lot of data calculations, the requirements for hardware and software resource arerelatively high. General microprocessor can’t meet the requirements of recommend suite andreal-time. So this system used the advanced digital signal processor TMS320DM642produced byTI company as the core to build a vision system.In this paper, the research object is robot visual navigation system in farmland environment.It contains the software and hardware design of the system and the selection of real-time systems,the image processing algorithm simulation, etc.The concrete research content of this paper is showed as follows:（1）The hardware design of Visual navigation system: Acquisition system consists of videoimage acquisition module, data storage module, DM642information processing module, imagedisplay module and power module. On the basis of the hardware platform visual navigationsoftware is developed.（2） The design of visual navigation system software: The operating system based onreal-time DSP/BIOS and CSL library image acquisition program. DSP/BIOS and CSL library cansimplify software development, the library function configuration and control hardware, so as toshorten the development cycle and also reduce the unknown error probability.（3）Image acquisition: Using CCD high-definition cameras to collect image information inPAL, analyses the collection procedures. Collected images are in YUV color space. This color space interference between the components is smaller, suitable for the next step of imageprocessing.（4）System optimization: The resampling method according to the actual situation of furrowin farmland is put forward to ensure the real-time processing. Because environmental variation offarmland furrow is not large, the sampling frequency can be reduced, that leaves a more adequateprocessing time on image processing.（5）Experiment simulation: Study the furrow image processing algorithm through MATLABsimulation processing of the collected pictures, including gray-scale transformation, thresholdsegmentation, background removing, smooth filtering, corrosion, effective area extraction, edgedetection. A path midpoint is found from the edge information, finally using the least squares to fitout walking direction information as the robot motion.The robot visual platform studied in this paper can accomplish the image acquisition andprocessing. Using DSP/BIOS real-time operating system kernel, you can define all attributes ofDM642, also can complete program online debugging and system scheduling functions. UsingCSL library can facilitate the configuration and control peripherals, saves to use DSP peripheraldevice programming work, also can improve the standardization and compatibility of process.Based on the investigation to the furrow environment analysis, the collected video image25frames per second is sampled again, reducing the amount of data processing, so in the same timealgorithm on image processing is increasing to ensure the reliability of the image characteristics.Image processing algorithms result can make the agricultural robot identify the environment offarmland clearly. Using the least squares estimate and recognize motion path. Least squares canpartly filter noise. It has some certain discriminated function for interference from openagricultural practical environment (such as garbage, weed, trees, etc.). It also has an importantpractical significance to the agricultural production. |
PDF Full Text Request