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

Fruit harvesting is a labor-concentrated job.With the senility and lack of labor, cost of labor force in the field is getting higher. Therefore, more and more people pay attention to the research on agriculture robots. But the roboticized level of picking is still very low now in china because of the complicated object and mutable working environment. To adapt to the request and need of precision agriculture, the vision system of apple picking robots aiming at ripe apple 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 apples images were collectioned. according to the experiment environment and the experiment request,methodwas designed.2. The background segmentation of pictures and the contrast research to eleven kinds of segmentation factors on the segmentation effect based on the color characteristic was studied. The experiment showed that it can achieve good segmentation effect with the segmentation factor:the R-G chromatism model.3.Three methods of Image segmentation were investigated.Through the contrast research, finally improved Ostu was selected because the processed effect was good. After images were segmented, the noise was eliminated with the area threshold and low-pass filter method.4. The separation of the contacted or the adhered multi-fruit was also studied preliminary, and three methods were proposed:Watershed, Circle Hough transformation, Morphology operator. The experimental verification was also given finally.5. Two methods of detecting fruit growth posture were proposed based on analyzing the shape features of apples:the shape characteristic method and the Minimum circumscribed rectangular. The theoretical analysis and the experimental verification to the two methods were carried out. The results of the experiments and the error analysis were 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 apples and the development of vision systems of the fruit picking robots.