1)车牌区域定位完结后，将车牌区域切割成单个字符，字符在直线方向的投影需求在字符之间的空隙处或字符内部取得部分的接近度，并且这个方向要满足字符书写形式、字符、比例限制等一些车牌的条件。多合一车牌识别机选用直线投影法在杂乱环境下切割汽车图片中的字符。

1) After the location of the license plate area is completed, the license plate area is cut into a single character. The projection of the character in the straight direction needs to obtain part of the proximity in the gap between characters or inside characters, and this direction should meet some license plate conditions such as the writing form of characters, characters, proportion restrictions, etc. The multi-in-one license plate reader uses the linear projection method to cut the characters in the car picture in the chaotic environment.

2)车牌定位，定位图片中的车牌方位：车牌识别一体机首要对采集到的视频图片进行大规模的相关查找，找到几个符合汽车车牌特征的区域作为候选区域，然后对这些候选区域进行进一步的剖析和点评，然后挑选合适的区域作为车牌区域，并从图片中分离出来。

2) License plate location, locating the license plate location in the picture: the license plate recognition integrated machine first conducts a large-scale search on the collected video images, finds several regions in line with the characteristics of the car license plate as candidate areas, and then carries out further analysis and comment on these candidate areas, and then selects the appropriate region as the license plate area and separates it from the picture.

3)车牌字符识别：辨认剪切的字符，构成车牌号码。车牌字符识别方法主要根据模板匹配算法和人工神经网络算法。然后与所有模板匹配，挑选匹配的结果。停车场车牌识别系统中根据人工神经网络的算法有两种：一种是先提取字符的特征，然后用取得的特征训练神经网络分配器;另一种方法是直接将图片输入网络，网络自动完结特征提取，直到辨认出结果。

3) License plate character recognition: identify the cut characters, constitute the license plate number. The license plate character recognition method is mainly based on template matching algorithm and artificial neural network algorithm. Then match all the templates and pick the matching results. There are two algorithms based on artificial neural network in car park license plate recognition system. One is to extract character features first, and then train neural network distributor with the obtained features. Another method is to directly input the image into the network, and the network automatically completes the feature extraction until the result is recognized.