經過實際測試,發現使用模糊類神經網路演算法及後續顏色特徵補助,可以在機車行駛路段中判別未戴安全帽者。其中在機車駕駛高亮度未戴安全帽樣本學習辨識率為95%,正常亮度晴天為96%。高亮度晴天戴半罩式安全帽者95%、正常亮度晴天下為98%。高亮度晴天戴全罩式辨識率為97%、正常晴天戴全罩式安全帽為98%。實際測試高亮度晴天未戴安全帽辨識率為94%、正常亮度晴天未戴安全帽96%,雖然未達百分之百辨識率,但已成功進行初步辨識工作。 By the highly Video technology developing within the past several years, the video-based detector can reached more Traffic parameters than tradition version, the video-based detector have the advantage like low cost, and easily to maintain. Therefore, it is a efficiency method to obtain Traffic parameters by image processing.
The number of motorcycle is very higher in Taiwan now, but the discussion of enforcement by video-based detector is less. The driving habits of people generally adverse, although the violation may not cause car accident, it still makes the society pay huge cost. Recently, the violation of driver without helmet is getting higher with the car-hold-rate increasing; it shows that traffic safety is very important.
In our study, first we use the video-based detector getting video frame, next, using Fuzzy Neural Network to identify the violation of driver without helmet, and finally add the HSV technology to increase detection rates.The simulation results show that using the FNN and HSV to increase detection rates is worked. The detection rate of driver without helmet is more than 96%.Although the detection rates were not achieve 100%, but the feasible platform was established successfully.