Research On The Instrument Of The Mass Of 1000 Based On Image Recognition

The mass of one thousand grains is one of important indices to measure the quality of seeds. However, in order to test the mass of one thousand grains, separating, counting and weighing are mostly operated by manpower at present. This way costs a lot of time and manpower. In view of the problem, selecting maize as an object, combing the technology of image processing and recognition, using the technology of electromagnetic vibrating, a device that can separate sundries and broken seeds from a mass of maize was designed. The main contents are as follows:1. Through the force analysis of grain on the bevel, using the pulse of electromagnetic force generated by electromagnetic coil, an electromagnetic vibrating seeding machine was designed. It can make the testing object passed the testing point sequentially. This device made the force–driving system simplified a lot. The structure is simple and the control and adjustment is convenient. It is not only fit for the grains, but also for other granule object.2. A digital camera was used to collect the image of maize. The images were with high quality. Further more, the digital camera is fit for on-spot image collecting and easier to be manipulated.3. Medium filtering was selected in the process of lower image processing because it could satisfy the demand of design well and the calculating speed was fast. The Sobel operator was selected for edge detecting for its better effect and easier to separate the image form the background.4. Arithmetic of improved moment invariants features were used in the process of feature extracting. The moment invariants features were not only invariant in continuous state, but also invariant in shifting and rotary state. The accuracy was improved a lot with the help of them.5. According to the morphologic structure features, 7 parameters were calculated from the morphologic structure, including shape, size and so on, they could distinguish the normal and broken maize. Based on the extracted feature parameters, the arithmetic of BP neural network was used to classify the image. The accuracy of identifying rate was up to 99%.6. The Infrared photoelectricity coupling sensor and the high sensitivity weighing sensor were used to monitor and weigh the measured object. They could work well .The accuracy and sensitivity was also improved a lot.7. The PC's serial port and the single chip microcomputer's (MCU) serial port were used to design the serial communication and the hardware connecting between PC and MCU. 8. A software of image recognition was developed under the environment of MATLAT7.0.With the powerful function of numerical calculation, image processing function and its own instrument toolbox, image filtering, edge detecting, feature parameters calculating, establishment of artificial neural network, manipulation of serial port instrument object could be accomplished.