Application In Agricultural Products Fruit Recognition Based On DSP Machine Vision

Current domestic field vegetable and fruit picking tasks are basically completed by hand, in recent years, the aging of the population and the urbanization to speed accelerate led to the agricultural labor force reduction, agricultural production costs and improve labor productivity and reduce, in order to solve these problems, the degree of automation of agricultural machinery potential necessary improved. With the specific working environment of industrial robot, the main job of agricultural robot is to face more complicated situations, and more problems need to be solved. The automatic identification technology of the agricultural picking robot increases the difficulty of the complicated and changeable farmland scene. This experiment is designed to design a machine vision system that can recognize the fruit in real time and effectively in the natural scene.At present, most of the scholars use the PC as the core processor of the machine vision systems, and the advantage of this design is that the recognition speed and reliability are high. However, there are huge volume and power consumption of PC, which can’t be widely used in the actual production of fruit picking robot based on PC. Therefore, the automatic recognition system of agricultural product fruits based on DSP+ light intensity acquisition equipment is proposed in this paper.This thesis is as follows:(1) According to the overall design needs, the system is determined by TMS320DM642, video codec chip, CCD camera, monitor, illumination intensity sensor, FLASH, SDRAM and other hardware components;(2) Fruit recognition algorithm research: in this paper, citrus, for example, through a large number of MATLAB simulation experiment, ultimately determine the mean of the YUV color space for the Cr component information filtering, histogram threshold segmentation, morphological noise, based on the area of feature extraction and Prewitt edge edge detection operation, thus completing the identification of citrus fruit. Provide theoretical basis for the next step of the software design;(3) Software design: according to the function and features of image processing, the program framework of DSP is written data acquisition program, fruit recognition program and driver program;(4) System optimization and experimental results analysis: the operation of the system for a long time, the image recognition algorithm programming optimization, make sure that the execution time of the whole system is less than 40 ms, and the experimental results were analyzed.This paper has completed all the experiments mentioned above, and designed the machine vision system of the fruit picking robot with small size, light weight and low power consumption. The system solves the light is too strong or too weak to the fruit recognition effect is not ideal, reached in natural light according to the real-time and effective identification under the condition of the effect of fruit contour, and the recognition accuracy is above 90%. In addition, the system through changing the value of image recognition algorithm in the corresponding threshold and area features can be lies in the application of other agricultural products(fruit and background color differences) picking robots, with a wide range of versatility.