Study On The Computer Vision System Of Tomato Quality Testing And Grading

The population of China is very large, but the per capita infield area is far off the average level of the world and the agricultural production value is very low, too. One of the main methods to increase the agricultural production value is to improve the post disposal level of the farm products. In our country the grading still mainly depends on manual grading, and it has many disadvantages such as: taking time, inefficient and subjectivity strong, which results in the fruit bad quality. Therefore an automatic grading system is required urgently. It aims at solving the problem to study the tomato quality detection on computer vision. The inspection items for tomato external quality deal with surface defects, color, shape and size. The tasks of study are briefly summarized as follows:(1) A hardware system used to acquire image was built. It included: Person Computer, PCI/PXI-1411 IMAQ from NI company and a homemade light-box. Panasonic WV-CP240/G camera, annular incandescence lamp 40W and elevator were set up in light-box. After experimenting, white was chosen as background, the acquired images were studied. The distance was maximal between two-peak value in G gray level histogram. A set of computer vision hardware system which adapts for quality evaluation and classification of tomato was set up.(2) Six arithmetics of lower-layer image processing technique were studied. The following methods were determined at last: image transformation, image enhancement, median filtering, image segmentation, image edge detection, etc. The method of lower-layer image processing technique can meet the demand of quality detection and classification.(3) Training multilayer feedforward neural networks with GA for detecting tomato maturation and shape was built, and was compared with BP algorithm .GA was superior to BP algorithm not only in training times but also degree of accuracy. The shape of tomato can be described with roundness, symmetry and height ratio. An equation which reflects the relationship between the area in maximal diameter and weight of tomato was set up to measure the weight of tomato. Its testing precision reached 99%. The color was graded by HIS.(4) The application software of tomato grading was programmed with Labview and IMAQ Vision. Its functions include 'file', 'image acquisition','lower-layer processing of image', 'character pick-up ', 'network training', and'classification'. The software had user-friendly interface and can be operated easily.