Image Recognition And Posture Detection Of Ripe Tomato Based On Machine Vision

Recently, the development of agriculture is heading to mass production,diversification and precision. The cost of labor force in the field is getting higher, and the phenomenon of lack of labor force is getting obvious. Therefore, more and more people pay attention to the research on agriculture robot. But the roboticized level of picking is still very low now in homeland because of the complicated object and non-structural working environment. To adapt to the request and need of precision agriculture, the vision system of tomato picking robot aiming at ripe tomato was researched based on computer machine vision, by using digital image processing and pattern recognition. The major contents and results of the study could be briefly summarized as follows:1. The composition of software and hardware and the experiment environment of this study were introduced; massive tomatoes color images under different natural environment were gathered; the plan of experiments plan according to the experiment environment and the experiment request was designed.2. The background segmentation of pictures and the contrast research to ten kinds of segmentation factors on the segmentation effect based on the color characteristic was researched. The experiment showed that it can achieve good segmentation effect with the segmentation factor:I component of the YIQ color space.3. Image segmentation based on automatic selection of image threshold can enhance the image processing timeliness greatly. Through the contrast research, finally Iteration was selected because the processed effect was good. After images were segmented, the noise was eliminated with the area threshold method.4. Three methods of detecting fruit growth posture were proposed based on analyzing the shape features of tomatoes:the moment characteristic method, the shape characteristic method and the equal division area method. The theoretical analysis and the experimental verification to the three methods were carried out. The results of the experiments and the error analysis were given finally. 5. The separation of the contacted or the adhered multi-fruit was also studied preliminary, and three methods were proposed:Circle Hough transformation, Watershed, Morphology operator. The experimental verification was also given finally.Finally, the prime task was summarized, and the defects and the later research direction were pointed out.The research of this topic is of definite instruction to the automatic picking of tomatoes and the development of vision systems of the fruit picking robots.