Study Of The Image Extraction Suitable For The Opisthenar Puncture

With the improvement of living standard, people demand for better medical condition. Hand vein injection is one of the most popular medical procedures. Improving its success rate and realizing automatic vein injcetion have great significance on saving medical resources.Now there are vein display instruments, which only improve the displaying results without extracting the parts suitable for vein injection. Hand vein extraction and recognition are mainly applied in biometric field but little research in medical field. Based on the method in biometric field, this paper has images collected preprocessing at first, and then analyzes features of veins suitable for injection. At last extracts the suitable injection parts successfully and marks them. The main work is as the following:(1) Images collection: According to the absorption property of the hand vein, its absorption rate of near-infrared light is much more than other tissues. Experimental results show that the combination of the incandescent lamp and infrared filter can get relatively uniform near-infrared light source.Then using Raspberry Pi and its matching camera to collect the image.(2) Images preprocessing: Using the centroid method to extract the effective area of the hand vein and carry on the gray normalization for the subsequent processing. Then taking gaussian filter and median filter method to remove the noise. Finally analyzing several typical image segmentation algorithms, the result shows that the effect based on one dimensional gray segmentation method is better.(3) Images extraction: At first, comparing three thinning algorithm and then using improved region growing method to extract vein separately. Making use of the length and curvature to match 3 veins which suitable for the intravenous injection. At last, applying template matching method to obtain suitable injection parts and mark them based on the original segmentation image. Experimental results show that the algorithm in this paper can effectively extract the part for intravenous injection, and has high recognition rate. On the other hand, the total time of the algorithm is less than 1.5 s, which can satisfy the actual needs.