指针和结构导致分段错误

Pointer and structures cause Segmentation fault

本文关键字:分段 错误 结构 指针      更新时间:2023-10-16

我一直在使用openCV的ColorSegmentationAlgorithm。我完成了分割,但现在我正在按颜色收集图片上的对象。为了做到这一点,我在RLE中压缩了图像的"行",它们用指针相互引用。

在这一点上(我认为)一切都很好,因为我在控制台上打印了RLE中压缩的"集群"的结果,一切似乎都很好。然后我查看结构列表,创建一个新的变量来表示对象(带有边界框、质心等)。但是,问题是,我现在不知道为什么在"随机情况"下调用指针的内容时加载不好(就像变量被移位一样):例如,如果结构有以下内容:

(int i,int je,int js,"pointer"*child)-->{1,1,0x0AB0231A}

其已加载-->{1,0x0AB0231A,1,1}

如果调用下一个"子",则会引发分段故障(核心转储)错误

所以。。。这是代码的一部分:

struct LineObjRLE {
    unsigned int i;
    unsigned int js;
    unsigned int je;
    unsigned int size;
    unsigned int color;
    struct LineObjRLE *parent;
    struct LineObjRLE *child;
};

struct LineObjRLE auxRLE = aRLE[i][j];
vector<struct LineObjRLE> obj;
while (1) {
    obj.push_back(auxRLE);
    printf("auxRLE: parent: %d  --  child: %dn",
            auxRLE.parent, auxRLE.child);
    if (auxRLE.child == NULL)
        break;
    auxRLE = *auxRLE.child;
}
  • aRLE(RLE数组)是在分割时创建的矢量变量

这里我得到了一些结果:(指针显示为int,但它们是正确的)

---------------------------------
k: 1 
auxRLE: parent: 0  --  child: 55035088
auxRLE: parent: 55036080  --  child: 55505600
auxRLE: parent: 55035088  --  child: 0
---------------------------------
k: 2 
auxRLE: parent: 0  --  child: 55035248
auxRLE: parent: 0  --  child: 0
---------------------------------
k: 3 
auxRLE: parent: 0  --  child: 55035288
auxRLE: parent: 55036200  --  child: 55505720
auxRLE: parent: 55035288  --  child: 0
---------------------------------
k: 4 
auxRLE: parent: 0  --  child: 55035528
auxRLE: parent: 0  --  child: 55505840
auxRLE: parent: 55035528  --  child: 0
---------------------------------
k: 5 
auxRLE: parent: 0  --  child: 55035688
auxRLE: parent: 0  --  child: 0
---------------------------------
k: 6 
auxRLE: parent: 0  --  child: 55035808
auxRLE: parent: 18  --  child: 6
Segmentation fault (core dumped)

当我之前描述的"转变"发生时,就会出现分段。当我检查结构的所有内容时,我发现在结构的其他变量中的指针和指针有其他值(所以指向6或18会产生错误)。

PD:我没有保存"parents"指针,而是只使用"child"方向来查看结构。

我希望我能解释我自己,有人能帮助我。提前感谢!

代码:

    while (waitKey(1)) {
        t1 = clock(); // Time counter1
        device >> frame; // Get the frame from the camera
        imshow("Ori", frame); // Show the original frame
        //medianBlur(frame, frame, 5);
        imageBGR2HSV(&frame); // Transform image from BGR to HSV
        int n = frame.channels(); // Count the number of image's channels to use the pointer
        int color;
        short int js, colorRLE = -1, jRLE = -1; // Variables for RLE encode
        for (int i = 0; i < frame.rows; i++) {
            uchar* ptr = frame.ptr<uchar>(i); // Pointer to i row
            vector<struct LineObjRLE> temp;
            for (int j = 0; j < frame.cols; j++) {
                // Proximate the color to a cluster
                color = CS.whichColorHSV(
                        (color3cInt ) { ptr[n * j], ptr[n * j + 1], ptr[n
                                        * j + 2] });
                // RLE encoding
                if (!j) {
                    colorRLE = color;
                    jRLE = 1;
                    js = 0;
                } else {
                    if (j == frame.cols - 1) {
                        temp.push_back((LineObjRLE ) { i, js, j, j - js, color,
                                NULL, NULL });
                    } else if (color == colorRLE) {
                        jRLE = jRLE + 1;
                    } else if (color != colorRLE) {
                        temp.push_back((LineObjRLE ) { i, js, j, j - js, color,
                                NULL, NULL });
                        colorRLE = color;
                        jRLE = 1;
                        js = j;
                    }
                }
                // Change the color (Improve assigning directly the BGR color, save from using imageHSV2BGR)
                if (color != -1) {
                    ptr[n * j] = CS.ColorCluster[color].a;
                    ptr[n * j + 1] = CS.ColorCluster[color].b;
                    ptr[n * j + 2] = CS.ColorCluster[color].c;
                } else {
                    ptr[n * j] = CS.ColorCluster[0].a;
                    ptr[n * j + 1] = CS.ColorCluster[0].b;
                    ptr[n * j + 2] = CS.ColorCluster[0].c;
                }
            }
            aRLE.push_back(temp);
            if (i) { // Except the first line that can't be child of any object (only parent) start joining objects grouped in LineObjRLE variables
                unsigned int j = 0, jp = 0; // Pointer to current and previous LineObjRLE
                unsigned int pp = aRLE[i - 1][jp].size, pc = aRLE[i][j].size; // Pointer to previous and current col
                bool end = false; // Flag to manage the loop
                while (!end) {
                    if ((aRLE[i - 1][jp].je > aRLE[i][j].js
                            && aRLE[i - 1][jp].js < aRLE[i][j].je)
                            && aRLE[i - 1][jp].color == aRLE[i][j].color) {
                        aRLE[i][j].parent = &(aRLE[i - 1][jp]);
                        aRLE[i - 1][jp].child = &(aRLE[i][j]);
                        //printf("Parent is %d and says that child is: %dn", &aRLE[i - 1][jp], aRLE[i - 1][jp].child);
                        //printf("Child is %d and says that parent is: %dn", &aRLE[i][j], aRLE[i][j].parent);
                    }
                    if (j == aRLE[i].size() - 1 || jp == aRLE[i - 1].size() - 1)
                        end = true;
                    if (pp > pc) {
                        pc += aRLE[i][j].size;
                        j++;
                    } else {
                        pp += aRLE[i - 1][jp].size;
                        jp++;
                    }
                }
            }
        }
        // Run vertically to identifies the parents of the objects
        int k = 0; //Index for final object
        for (unsigned int i = 0; i < aRLE.size(); i++) {
            for (unsigned int j = 0; j < aRLE.at(i).size(); j++) {
                if (aRLE[i][j].parent == NULL && aRLE[i][j].child != NULL) {
                    printf("---------------------------------n");
                    printf("k: %d n", k);
                    printf("aRLE: parent: %d  --  child: %dn",
                            aRLE[i][j].parent, aRLE[i][j].child);
                    // Grow from the seed
                    struct LineObjRLE auxRLE = aRLE[i][j];
                    vector<struct LineObjRLE> obj;
                    while (1) {
                        obj.push_back(auxRLE);
                        printf("auxRLE: parent: %d  --  child: %dn",
                                auxRLE.parent, auxRLE.child);
                        if (auxRLE.child == NULL)
                            break;
                        auxRLE = *auxRLE.child;
                    }
                    k = k + 1;
                }
            }
        }
        // Calculate increment of time
        t2 = clock();
        float dif = (((float) t2 - (float) t1) / 1000000.0F) * 1000;
        printf("FINISHED IN %f n", dif);
        imageHSV2BGR(&frame);
        imshow("Viewer", frame); // Show the image segmented
        aRLE.clear();
        objs.clear();
    }
    return 0;
}

查看代码,您似乎在std::vector<RLE>中存储指向对象的指针。例如,我发现:

aRLE[i][j].parent = &(aRLE[i - 1][jp]);

同时,我没有看到任何对aRLE.reserve()的调用,这将确保在添加新对象时不会重新定位对象。我怀疑,向量需要重新分配内存,释放已经指向的内存,然后得到一堆过时的指针。

如果你创建了过时的指针,测试这个理论最简单的方法是:

  • 如果您提前知道数据结构的最终大小,则可以使用std::vector<T>::reserve(size_type n)告诉数据结构分配足够的空间来容纳最多n个元素
  • 如果你不知道大小,但你只需要一个看起来像数组的东西,并且你只在末尾或开头附加,你可以用std::deque<RLE>代替std::vector<RLE>:虽然在两端添加对象可能会改变迭代器,但它不会改变对象的位置